Publications of Bilge Karaçalı
Journal Articles |
Çakı, O; Karaçalı, B Quasi‐supervised Strategies for Compound‐protein Interaction Prediction Journal Article Molecular Informatics, 41 (4), pp. 2100118–2100118, 2022. @article{pop00001w, title = {Quasi‐supervised Strategies for Compound‐protein Interaction Prediction}, author = {O Çakı and B Karaçalı}, year = {2022}, date = {2022-01-01}, journal = {Molecular Informatics}, volume = {41}, number = {4}, pages = {2100118--2100118}, keywords = {}, pubstate = {published}, tppubtype = {article} } |
Olcay, BO; Özgören, M; Karaçalı, B On the characterization of cognitive tasks using activity-specific short-lived synchronization between electroencephalography channels Journal Article Neural Networks, 143 , pp. 452–474, 2021. @article{pop00001l, title = {On the characterization of cognitive tasks using activity-specific short-lived synchronization between electroencephalography channels}, author = {BO Olcay and M Özgören and B Karaçalı}, year = {2021}, date = {2021-01-01}, journal = {Neural Networks}, volume = {143}, pages = {452--474}, keywords = {}, pubstate = {published}, tppubtype = {article} } |
Çakı, O; Karaçalı, B Quasi‐Supervised Strategies for Compound‐Protein Interaction Prediction Journal Article Molecular Informatics, 2100118 , 2021. @article{pop00002g, title = {Quasi‐Supervised Strategies for Compound‐Protein Interaction Prediction}, author = {O Çakı and B Karaçalı}, year = {2021}, date = {2021-01-01}, journal = {Molecular Informatics}, volume = {2100118}, keywords = {}, pubstate = {published}, tppubtype = {article} } |
Güzel, BEK; Karaçalı, B Identification and visualization of cell subgroups in uncompensated flow cytometry data Journal Article Chemometrics and Intelligent Laboratory Systems, 196 , pp. 103892–103892, 2020. @article{pop00003l, title = {Identification and visualization of cell subgroups in uncompensated flow cytometry data}, author = {BEK Güzel and B Karaçalı}, year = {2020}, date = {2020-01-01}, journal = {Chemometrics and Intelligent Laboratory Systems}, volume = {196}, pages = {103892--103892}, keywords = {}, pubstate = {published}, tppubtype = {article} } |
Olcay, BO; Karaçalı, B Evaluation of synchronization measures for capturing the lagged synchronization between EEG channels: A cognitive task recognition approach Journal Article Computers in biology and medicine, 114 , pp. 103441–103441, 2019. @article{pop00004j, title = {Evaluation of synchronization measures for capturing the lagged synchronization between EEG channels: A cognitive task recognition approach}, author = {BO Olcay and B Karaçalı}, year = {2019}, date = {2019-01-01}, journal = {Computers in biology and medicine}, volume = {114}, pages = {103441--103441}, keywords = {}, pubstate = {published}, tppubtype = {article} } |
Köktürk, Başak Esin; Karacalı, Bilge Annealing-based model-free expectation maximisation for multi-colour flow cytometry data clustering Journal Article International Journal of Data Mining and Bioinformatics, 2016, ISSN: 1748-5673. @article{Kokturk2016, title = {Annealing-based model-free expectation maximisation for multi-colour flow cytometry data clustering}, author = {Başak Esin Köktürk and Bilge Karacalı}, doi = {10.1504/ijdmb.2016.073365}, issn = {1748-5673}, year = {2016}, date = {2016-01-01}, journal = {International Journal of Data Mining and Bioinformatics}, abstract = {Copyright textcopyright 2016 Inderscience Enterprises Ltd.This paper proposes an optimised model-free expectation maximisation method for automated clustering of high-dimensional datasets. The method is based on a recursive binary division strategy that successively divides an original dataset into distinct clusters. Each binary division is carriedout using a model-free expectation maximisation scheme that exploits the posterior probability computation capability of the quasi-supervised learningalgorithm subjected to a line-search optimisation over the reference set size parameter analogous to a simulated annealing approach. The divisions arecontinued until a division cost exceeds an adaptively determined limit. Experiment results on synthetic as well as real multi-colour flow cytometrydatasets showed that the proposed method can accurately capture the prominent clusters without requiring any prior knowledge on the number of clusters ortheir distribution models.}, keywords = {}, pubstate = {published}, tppubtype = {article} } Copyright textcopyright 2016 Inderscience Enterprises Ltd.This paper proposes an optimised model-free expectation maximisation method for automated clustering of high-dimensional datasets. The method is based on a recursive binary division strategy that successively divides an original dataset into distinct clusters. Each binary division is carriedout using a model-free expectation maximisation scheme that exploits the posterior probability computation capability of the quasi-supervised learningalgorithm subjected to a line-search optimisation over the reference set size parameter analogous to a simulated annealing approach. The divisions arecontinued until a division cost exceeds an adaptively determined limit. Experiment results on synthetic as well as real multi-colour flow cytometrydatasets showed that the proposed method can accurately capture the prominent clusters without requiring any prior knowledge on the number of clusters ortheir distribution models. |
Doty, Richard L; Tourbier, Isabelle A; Pham, Dzung L; Cuzzocreo, Jennifer L; Udupa, Jayaram K; Karacali, Bilge; Beals, Evan; Fabius, Laura; Leon-Sarmiento, Fidias E; Moonis, Gul; Kim, Taehoon; Mihama, Toru; Geckle, Rena J; Yousem, David M Taste dysfunction in multiple sclerosis Journal Article Journal of Neurology, 2016, ISSN: 14321459. @article{Doty2016, title = {Taste dysfunction in multiple sclerosis}, author = {Richard L Doty and Isabelle A Tourbier and Dzung L Pham and Jennifer L Cuzzocreo and Jayaram K Udupa and Bilge Karacali and Evan Beals and Laura Fabius and Fidias E Leon-Sarmiento and Gul Moonis and Taehoon Kim and Toru Mihama and Rena J Geckle and David M Yousem}, doi = {10.1007/s00415-016-8030-6}, issn = {14321459}, year = {2016}, date = {2016-01-01}, journal = {Journal of Neurology}, abstract = {textcopyright 2016, Springer-Verlag Berlin Heidelberg. Empirical studies of taste function in multiple sclerosis (MS) are rare. Moreover, a detailed assessment of whether quantitative measures of taste function correlate with the punctate and patchy myelin-related lesions found throughout the CNS of MS patients has not been made. We administered a 96-trial test of sweet (sucrose), sour (citric acid), bitter (caffeine) and salty (NaCl) taste perception to the left and right anterior (CN VII) and posterior (CN IX) tongue regions of 73 MS patients and 73 matched controls. The number and volume of lesions were assessed using quantitative MRI in 52 brain regions of 63 of the MS patients. Taste identification scores were significantly lower in the MS patients for sucrose (p = 0.0002), citric acid (p = 0.0001), caffeine (p = 0.0372) and NaCl (p = 0.0004) and were present in both anterior and posterior tongue regions. The percent of MS patients with identification scores falling below the 5th percentile of controls was 15.07 % for caffeine, 21.9 % for citric acid, 24.66 % for sucrose, and 31.50 % for NaCl. Such scores were inversely correlated with lesion volumes in the temporal, medial frontal, and superior frontal lobes, and with the number of lesions in the left and right superior frontal lobes, right anterior cingulate gyrus, and left parietal operculum. Regardless of the subject group, women outperformed men on the taste measures. These findings indicate that a sizable number of MS patients exhibit taste deficits that are associated with MS-related lesions throughout the brain.}, keywords = {}, pubstate = {published}, tppubtype = {article} } textcopyright 2016, Springer-Verlag Berlin Heidelberg. Empirical studies of taste function in multiple sclerosis (MS) are rare. Moreover, a detailed assessment of whether quantitative measures of taste function correlate with the punctate and patchy myelin-related lesions found throughout the CNS of MS patients has not been made. We administered a 96-trial test of sweet (sucrose), sour (citric acid), bitter (caffeine) and salty (NaCl) taste perception to the left and right anterior (CN VII) and posterior (CN IX) tongue regions of 73 MS patients and 73 matched controls. The number and volume of lesions were assessed using quantitative MRI in 52 brain regions of 63 of the MS patients. Taste identification scores were significantly lower in the MS patients for sucrose (p = 0.0002), citric acid (p = 0.0001), caffeine (p = 0.0372) and NaCl (p = 0.0004) and were present in both anterior and posterior tongue regions. The percent of MS patients with identification scores falling below the 5th percentile of controls was 15.07 % for caffeine, 21.9 % for citric acid, 24.66 % for sucrose, and 31.50 % for NaCl. Such scores were inversely correlated with lesion volumes in the temporal, medial frontal, and superior frontal lobes, and with the number of lesions in the left and right superior frontal lobes, right anterior cingulate gyrus, and left parietal operculum. Regardless of the subject group, women outperformed men on the taste measures. These findings indicate that a sizable number of MS patients exhibit taste deficits that are associated with MS-related lesions throughout the brain. |
Karacalı, Bilge An efficient algorithm for large-scale quasi-supervised learning Journal Article Pattern Analysis and Applications, 2016, ISSN: 14337541. @article{Karacal2016, title = {An efficient algorithm for large-scale quasi-supervised learning}, author = {Bilge Karacalı}, doi = {10.1007/s10044-014-0401-y}, issn = {14337541}, year = {2016}, date = {2016-01-01}, journal = {Pattern Analysis and Applications}, keywords = {}, pubstate = {published}, tppubtype = {article} } |
Bozkurt, Bariş; Karacali, Bilge A computational analysis of Turkish makam music based on a probabilistic characterization of segmented phrases Journal Article Journal of Mathematics and Music, 2015, ISSN: 17459745. @article{Bozkurt2015, title = {A computational analysis of Turkish makam music based on a probabilistic characterization of segmented phrases}, author = {Bariş Bozkurt and Bilge Karacali}, doi = {10.1080/17459737.2014.927012}, issn = {17459745}, year = {2015}, date = {2015-01-01}, journal = {Journal of Mathematics and Music}, abstract = {textcopyright 2014 Taylor & Francis. This study targets automatic analysis of Turkish makam music pieces on the phrase level. While makam is most simply defined as an organization of melodic phrases, there has been very little effort to computationally study melodic structure in makam music pieces. In this work, we propose an automatic analysis algorithm that takes as input symbolic data in the form of machine-readable scores that are segmented into phrases. Using a measure of makam membership for phrases, our method outputs for each phrase the most likely makam the phrase comes from. The proposed makam membership definition is based on Bayesian classification and the algorithm is specifically designed to process the data with overlapping classes. The proposed analysis system is trained and tested on a large data set of phrases obtained by transferring phrase boundaries manually written by experts of makam music on printed scores, to machine-readable data. For the task of classifying all phrases, or only the beginning phrases to come from the main makam of the piece, the corresponding F-measures are.52 and.60 respectively.}, keywords = {}, pubstate = {published}, tppubtype = {article} } textcopyright 2014 Taylor & Francis. This study targets automatic analysis of Turkish makam music pieces on the phrase level. While makam is most simply defined as an organization of melodic phrases, there has been very little effort to computationally study melodic structure in makam music pieces. In this work, we propose an automatic analysis algorithm that takes as input symbolic data in the form of machine-readable scores that are segmented into phrases. Using a measure of makam membership for phrases, our method outputs for each phrase the most likely makam the phrase comes from. The proposed makam membership definition is based on Bayesian classification and the algorithm is specifically designed to process the data with overlapping classes. The proposed analysis system is trained and tested on a large data set of phrases obtained by transferring phrase boundaries manually written by experts of makam music on printed scores, to machine-readable data. For the task of classifying all phrases, or only the beginning phrases to come from the main makam of the piece, the corresponding F-measures are.52 and.60 respectively. |
Onder, Devrim; Sarioglu, Sulen; Karacali, Bilge Automated labeling of cancer textures in larynx histopathology slides using Quasi-supervised learning Journal Article Analytical and Quantitative Cytology and Histology, 2014, ISSN: 08846812. @article{Onder2014, title = {Automated labeling of cancer textures in larynx histopathology slides using Quasi-supervised learning}, author = {Devrim Onder and Sulen Sarioglu and Bilge Karacali}, issn = {08846812}, year = {2014}, date = {2014-01-01}, journal = {Analytical and Quantitative Cytology and Histology}, abstract = {Quasi-supervised learning is a statistical learning algorithm that contrasts two datasets by computing estimate for the posterior probability of each sample in either dataset. This method has not been applied to histopathological images before. The purpose of this study is to evaluate the performance of the method to identify colorectal tissues with or without adenocarcinoma. Light microscopic digital images from histopathological sections were obtained from 30 colorectal radical surgery materials including adenocarcinoma and non-neoplastic regions. The texture features were extracted by using local histograms and co-occurrence matrices. The quasi-supervised learning algorithm operates on two datasets, one containing samples of normal tissues labelled only indirectly, and the other containing an unlabeled collection of samples of both normal and cancer tissues. As such, the algorithm eliminates the need for manually labelled samples of normal and cancer tissues for conventional supervised learning and significantly reduces the expert intervention. Several texture feature vector datasets corresponding to different extraction parameters were tested within the proposed framework. The Independent Component Analysis dimensionality reduction approach was also identified as the one improving the labelling performance evaluated in this series. In this series, the proposed method was applied to the dataset of 22,080 vectors with reduced dimensionality 119 from 132. Regions containing cancer tissue could be identified accurately having false and true positive rates up to 19% and 88% respectively without using manually labelled ground-truth datasets in a quasi-supervised strategy. The resulting labelling performances were compared to that of a conventional powerful supervised classifier using manually labelled ground-truth data. The supervised classifier results were calculated as 3.5% and 95% for the same case. The results in this series in comparison with the benchmark classifier, suggest that quasi-supervised image texture labelling may be a useful method in the analysis and classification of pathological slides but further study is required to improve the results. textcopyright 2013 Elsevier Ltd.}, keywords = {}, pubstate = {published}, tppubtype = {article} } Quasi-supervised learning is a statistical learning algorithm that contrasts two datasets by computing estimate for the posterior probability of each sample in either dataset. This method has not been applied to histopathological images before. The purpose of this study is to evaluate the performance of the method to identify colorectal tissues with or without adenocarcinoma. Light microscopic digital images from histopathological sections were obtained from 30 colorectal radical surgery materials including adenocarcinoma and non-neoplastic regions. The texture features were extracted by using local histograms and co-occurrence matrices. The quasi-supervised learning algorithm operates on two datasets, one containing samples of normal tissues labelled only indirectly, and the other containing an unlabeled collection of samples of both normal and cancer tissues. As such, the algorithm eliminates the need for manually labelled samples of normal and cancer tissues for conventional supervised learning and significantly reduces the expert intervention. Several texture feature vector datasets corresponding to different extraction parameters were tested within the proposed framework. The Independent Component Analysis dimensionality reduction approach was also identified as the one improving the labelling performance evaluated in this series. In this series, the proposed method was applied to the dataset of 22,080 vectors with reduced dimensionality 119 from 132. Regions containing cancer tissue could be identified accurately having false and true positive rates up to 19% and 88% respectively without using manually labelled ground-truth datasets in a quasi-supervised strategy. The resulting labelling performances were compared to that of a conventional powerful supervised classifier using manually labelled ground-truth data. The supervised classifier results were calculated as 3.5% and 95% for the same case. The results in this series in comparison with the benchmark classifier, suggest that quasi-supervised image texture labelling may be a useful method in the analysis and classification of pathological slides but further study is required to improve the results. textcopyright 2013 Elsevier Ltd. |
Bozkurt, Barış; Karaosmanoğlu, Kemal M; Karacalı, Bilge; Ünal, Erdem Usul and Makam driven automatic melodic segmentation for Turkish music Journal Article Journal of New Music Research, 2014, ISSN: 17445027. @article{Bozkurt2014, title = {Usul and Makam driven automatic melodic segmentation for Turkish music}, author = {Barış Bozkurt and Kemal M Karaosmanoğlu and Bilge Karacalı and Erdem Ünal}, doi = {10.1080/09298215.2014.924535}, issn = {17445027}, year = {2014}, date = {2014-01-01}, journal = {Journal of New Music Research}, abstract = {textcopyright 2014, textcopyright 2014 Taylor & Francis. Automatic melodic segmentation is a topic studied extensively, aiming at developing systems that perform grouping of musical events. Here, we consider the problem of automatic segmentation via supervised learning from a dataset containing segmentation labels of an expert. We present a statistical classification-based segmentation system developed specifically for Turkish makam music. The proposed system uses two novel features, a makam-based and an usul-based feature, together with features commonly used in literature. The makam-based feature is defined as the probability of a note to appear at the phrase boundary, computed from the distributions of boundaries with respect to the piece's makam pitches. Likewise, the usul-based feature is computed from the distributions of boundaries with respect to beats in the rhythmic cycle, usul of the piece. Several experimental setups using different feature groups are designed to test the contribution of the proposed features on three datasets. The results show that the new features carry complementary information to existing features in the literature within the Turkish makam music segmentation context and that the inclusion of new features resulted in statistically significant performance improvement.}, keywords = {}, pubstate = {published}, tppubtype = {article} } textcopyright 2014, textcopyright 2014 Taylor & Francis. Automatic melodic segmentation is a topic studied extensively, aiming at developing systems that perform grouping of musical events. Here, we consider the problem of automatic segmentation via supervised learning from a dataset containing segmentation labels of an expert. We present a statistical classification-based segmentation system developed specifically for Turkish makam music. The proposed system uses two novel features, a makam-based and an usul-based feature, together with features commonly used in literature. The makam-based feature is defined as the probability of a note to appear at the phrase boundary, computed from the distributions of boundaries with respect to the piece's makam pitches. Likewise, the usul-based feature is computed from the distributions of boundaries with respect to beats in the rhythmic cycle, usul of the piece. Several experimental setups using different feature groups are designed to test the contribution of the proposed features on three datasets. The results show that the new features carry complementary information to existing features in the literature within the Turkish makam music segmentation context and that the inclusion of new features resulted in statistically significant performance improvement. |
Bansal, Mukesh; Yang, Jichen; Karan, Charles; Menden, Michael P; Costello, James C; Tang, Hao; Xiao, Guanghua; Li, Yajuan; Allen, Jeffrey; Zhong, Rui; Chen, Beibei; Kim, Minsoo; Wang, Tao; Heiser, Laura M; Realubit, Ronald; Mattioli, Michela; Alvarez, Mariano J; Shen, Yao; Gallahan, Daniel; Singer, Dinah; Saez-Rodriguez, Julio; Xie, Yang; Stolovitzky, Gustavo; Califano, Andrea; Abbuehl, Jean Paul; Altman, Russ B; Balcome, Shawn; Bell, Ana; Bender, Andreas; Berger, Bonnie; Bernard, Jonathan; Bieberich, Andrew A; Borboudakis, Giorgos; Chan, Christina; Chen, Ting Huei; Choi, Jaejoon; Coelho, Luis Pedro; Creighton, Chad J; Dampier, Will; Davisson, Jo V; Deshpande, Raamesh; Diao, Lixia; Camillo, Barbara Di; Dundar, Murat; Ertel, Adam; Goswami, Chirayu P; Gottlieb, Assaf; Gould, Michael N; Goya, Jonathan; Grau, Michael; Gray, Joe W; Hejase, Hussein A; Hoffmann, Michael F; Homicsko, Krisztian; Homilius, Max; Hwang, Woochang; Ijzerman, Adriaan P; Kallioniemi, Olli; Karacali, Bilge; Kaski, Samuel; Kim, Junho; Krishnan, Arjun; Lee, Junehawk; Lee, Young Suk; Lenselink, Eelke B; Lenz, Peter; Li, Lang; Li, Jun; Liang, Han; Mpindi, John Patrick; Myers, Chad L; Newton, Michael A; Overington, John P; Parkkinen, Juuso; Prill, Robert J; Peng, Jian; Pestell, Richard; Qiu, Peng; Rajwa, Bartek; Sadanandam, Anguraj; Sambo, Francesco; Sridhar, Arvind; Sun, Wei; Toffolo, Gianna M; Tozeren, Aydin; Troyanskaya, Olga G; Tsamardinos, Ioannis; Vlijmen, Herman Van W T; Wang, Wen; Wegner, Joerg K; Wennerberg, Krister; Westen, Gerard Van J P; Xia, Tian; Yang, Yang; Yao, Victoria; Yuan, Yuan; Zeng, Haoyang; Zhang, Shihua; Zhao, Junfei; Zhou, Jian A community computational challenge to predict the activity of pairs of compounds Journal Article Nature Biotechnology, 2014, ISSN: 15461696. @article{Bansal2014, title = {A community computational challenge to predict the activity of pairs of compounds}, author = {Mukesh Bansal and Jichen Yang and Charles Karan and Michael P Menden and James C Costello and Hao Tang and Guanghua Xiao and Yajuan Li and Jeffrey Allen and Rui Zhong and Beibei Chen and Minsoo Kim and Tao Wang and Laura M Heiser and Ronald Realubit and Michela Mattioli and Mariano J Alvarez and Yao Shen and Daniel Gallahan and Dinah Singer and Julio Saez-Rodriguez and Yang Xie and Gustavo Stolovitzky and Andrea Califano and Jean Paul Abbuehl and Russ B Altman and Shawn Balcome and Ana Bell and Andreas Bender and Bonnie Berger and Jonathan Bernard and Andrew A Bieberich and Giorgos Borboudakis and Christina Chan and Ting Huei Chen and Jaejoon Choi and Luis Pedro Coelho and Chad J Creighton and Will Dampier and Jo V Davisson and Raamesh Deshpande and Lixia Diao and Barbara Di Camillo and Murat Dundar and Adam Ertel and Chirayu P Goswami and Assaf Gottlieb and Michael N Gould and Jonathan Goya and Michael Grau and Joe W Gray and Hussein A Hejase and Michael F Hoffmann and Krisztian Homicsko and Max Homilius and Woochang Hwang and Adriaan P Ijzerman and Olli Kallioniemi and Bilge Karacali and Samuel Kaski and Junho Kim and Arjun Krishnan and Junehawk Lee and Young Suk Lee and Eelke B Lenselink and Peter Lenz and Lang Li and Jun Li and Han Liang and John Patrick Mpindi and Chad L Myers and Michael A Newton and John P Overington and Juuso Parkkinen and Robert J Prill and Jian Peng and Richard Pestell and Peng Qiu and Bartek Rajwa and Anguraj Sadanandam and Francesco Sambo and Arvind Sridhar and Wei Sun and Gianna M Toffolo and Aydin Tozeren and Olga G Troyanskaya and Ioannis Tsamardinos and Herman Van W T Vlijmen and Wen Wang and Joerg K Wegner and Krister Wennerberg and Gerard Van J P Westen and Tian Xia and Yang Yang and Victoria Yao and Yuan Yuan and Haoyang Zeng and Shihua Zhang and Junfei Zhao and Jian Zhou}, doi = {10.1038/nbt.3052}, issn = {15461696}, year = {2014}, date = {2014-01-01}, journal = {Nature Biotechnology}, abstract = {Recent therapeutic successes have renewed interest in drug combinations, but experimental screening approaches are costly and often identify only small numbers of synergistic combinations. The DREAM consortium launched an open challenge to foster the development of in silico methods to computationally rank 91 compound pairs, from the most synergistic to the most antagonistic, based on gene-expression profiles of human B cells treated with individual compounds at multiple time points and concentrations. Using scoring metrics based on experimental dose-response curves, we assessed 32 methods (31 community-generated approaches and SynGen), four of which performed significantly better than random guessing. We highlight similarities between the methods. Although the accuracy of predictions was not optimal, we find that computational prediction of compound-pair activity is possible, and that community challenges can be useful to advance the field of in silico compound-synergy prediction. Recent success in the study of synergistic combinations, such as the use of CHK1 inhibitors in combination with several DNA damaging agents 1 or of the PARP inhibitor olaparib in combination with the PI3K inhibitor BKM120 (ref. 2), have generated significant interest in the systematic screening of compound pairs to identify synergistic pairs for combination therapy. Compound synergy can be measured by multiple endpoints, including reducing or delaying the devel-opment of resistance to treatment 3 (for instance by abrogating the emergence of resistant clones 4–6), improving overall survival 7,8 or lowering toxicity by decreasing individual compound dose 9 . Similarly, at the molecular level, synergistic interactions can be implemented by several distinct mechanisms. For instance, a compound may sensitize cells to another compound by regulating its absorption and distribution, modulating the cell's growth prop-erties 10 , inhibiting compound degradation 11 , inhibiting pathways that induce resistance 6 or reducing the other compound's toxicity 12 . When used in combination, two compounds may elicit one of three distinct responses: (i) additive, when the combined effect is equivalent to the sum of the independent effects; (ii) synergistic, when the combined effect is greater than additive; and (iii) antago-nistic, when the combined effect is smaller than additive. The goal of combination therapy is thus to attain a synergistic or at least an additive yet complementary effect. Most approaches to identify synergistic compound pairs are still exploratory 13,14 . In cancer research, synergy assays are usually performed by treating cell lines in vitro with all possible compound combinations from a diverse library or with candidate combinations selected on the basis of mechanistic principles. Unfortunately, such experimental screens impose severe limits on the practical size of compound diversity libraries. Computational methods to predict compound synergy can potentially complement high-throughput synergy screens, but the few that have been published lack rigorous experimental validation or are appropriate only for compounds that modulate well-studied molecular pathways 15 or that are equivalent to previously established combinations 16 . Current algorithms are not generalizable to arbitrary compound combinations unless molecular profile data following compound-pair treatment are available 17 , which is clearly impractical. Thus, there is a need for new methods to predict compound synergy from molecular profiles of single compound activity, as well as for assays designed to objectively and systematically evaluate the accuracy and specificity of such predictions. To address this issue, the DREAM Challenges initiative (an effort run by a community of researchers that poses fundamental ques-tions in systems biology and translational science in the form of a community computational challenge to predict the activity of pairs of compounds}, keywords = {}, pubstate = {published}, tppubtype = {article} } Recent therapeutic successes have renewed interest in drug combinations, but experimental screening approaches are costly and often identify only small numbers of synergistic combinations. The DREAM consortium launched an open challenge to foster the development of in silico methods to computationally rank 91 compound pairs, from the most synergistic to the most antagonistic, based on gene-expression profiles of human B cells treated with individual compounds at multiple time points and concentrations. Using scoring metrics based on experimental dose-response curves, we assessed 32 methods (31 community-generated approaches and SynGen), four of which performed significantly better than random guessing. We highlight similarities between the methods. Although the accuracy of predictions was not optimal, we find that computational prediction of compound-pair activity is possible, and that community challenges can be useful to advance the field of in silico compound-synergy prediction. Recent success in the study of synergistic combinations, such as the use of CHK1 inhibitors in combination with several DNA damaging agents 1 or of the PARP inhibitor olaparib in combination with the PI3K inhibitor BKM120 (ref. 2), have generated significant interest in the systematic screening of compound pairs to identify synergistic pairs for combination therapy. Compound synergy can be measured by multiple endpoints, including reducing or delaying the devel-opment of resistance to treatment 3 (for instance by abrogating the emergence of resistant clones 4–6), improving overall survival 7,8 or lowering toxicity by decreasing individual compound dose 9 . Similarly, at the molecular level, synergistic interactions can be implemented by several distinct mechanisms. For instance, a compound may sensitize cells to another compound by regulating its absorption and distribution, modulating the cell's growth prop-erties 10 , inhibiting compound degradation 11 , inhibiting pathways that induce resistance 6 or reducing the other compound's toxicity 12 . When used in combination, two compounds may elicit one of three distinct responses: (i) additive, when the combined effect is equivalent to the sum of the independent effects; (ii) synergistic, when the combined effect is greater than additive; and (iii) antago-nistic, when the combined effect is smaller than additive. The goal of combination therapy is thus to attain a synergistic or at least an additive yet complementary effect. Most approaches to identify synergistic compound pairs are still exploratory 13,14 . In cancer research, synergy assays are usually performed by treating cell lines in vitro with all possible compound combinations from a diverse library or with candidate combinations selected on the basis of mechanistic principles. Unfortunately, such experimental screens impose severe limits on the practical size of compound diversity libraries. Computational methods to predict compound synergy can potentially complement high-throughput synergy screens, but the few that have been published lack rigorous experimental validation or are appropriate only for compounds that modulate well-studied molecular pathways 15 or that are equivalent to previously established combinations 16 . Current algorithms are not generalizable to arbitrary compound combinations unless molecular profile data following compound-pair treatment are available 17 , which is clearly impractical. Thus, there is a need for new methods to predict compound synergy from molecular profiles of single compound activity, as well as for assays designed to objectively and systematically evaluate the accuracy and specificity of such predictions. To address this issue, the DREAM Challenges initiative (an effort run by a community of researchers that poses fundamental ques-tions in systems biology and translational science in the form of a community computational challenge to predict the activity of pairs of compounds |
Costello, James C; Heiser, Laura M; Georgii, Elisabeth; Gönen, Mehmet; Menden, Michael P; Wang, Nicholas J; Bansal, Mukesh; Ammad-Ud-Din, Muhammad; Hintsanen, Petteri; Khan, Suleiman A; Mpindi, John Patrick; Kallioniemi, Olli; Honkela, Antti; Aittokallio, Tero; Wennerberg, Krister; Collins, James J; Gallahan, Dan; Singer, Dinah; Saez-Rodriguez, Julio; Kaski, Samuel; Gray, Joe W; Stolovitzky, Gustavo; Abbuehl, Jean Paul; Allen, Jeffrey; Altman, Russ B; Balcome, Shawn; Battle, Alexis; Bender, Andreas; Berger, Bonnie; Bernard, Jonathan; Bhattacharjee, Madhuchhanda; Bhuvaneshwar, Krithika; Bieberich, Andrew A; Boehm, Fred; Califano, Andrea; Chan, Christina; Chen, Beibei; Chen, Ting Huei; Choi, Jaejoon; Coelho, Luis Pedro; Cokelaer, Thomas; Collins, James C; Creighton, Chad J; Cui, Jike; Dampier, Will; Davisson, Jo V; Baets, Bernard De; Deshpande, Raamesh; DiCamillo, Barbara; Dundar, Murat; Duren, Zhana; Ertel, Adam; Fan, Haoyang; Fang, Hongbin; Gauba, Robinder; Gottlieb, Assaf; Grau, Michael; Gusev, Yuriy; Ha, Min Jin; Han, Leng; Harris, Michael; Henderson, Nicholas; Hejase, Hussein A; Homicsko, Krisztian; Hou, Jack P; Hwang, Woochang; IJzerman, Adriaan P; Karacali, Bilge; Keles, Sunduz; Kendziorski, Christina; Kim, Junho; Kim, Min; Kim, Youngchul; Knowles, David A; Koller, Daphne; Lee, Junehawk; Lee, Jae K; Lenselink, Eelke B; Li, Biao; Li, Bin; Li, Jun; Liang, Han; Ma, Jian; Madhavan, Subha; Mooney, Sean; Myers, Chad L; Newton, Michael A; Overington, John P; Pal, Ranadip; Peng, Jian; Pestell, Richard; Prill, Robert J; Qiu, Peng; Rajwa, Bartek; Sadanandam, Anguraj; Sambo, Francesco; Shin, Hyunjin; Song, Jiuzhou; Song, Lei; Sridhar, Arvind; Stock, Michiel; Sun, Wei; Ta, Tram; Tadesse, Mahlet; Tan, Ming; Tang, Hao; Theodorescu, Dan; Toffolo, Gianna Maria; Tozeren, Aydin; Trepicchio, William; Varoquaux, Nelle; Vert, Jean Philippe; Waegeman, Willem; Walter, Thomas; Wan, Qian; Wang, Difei; Wang, Wen; Wang, Yong; Wang, Zhishi; Wegner, Joerg K; Wu, Tongtong; Xia, Tian; Xiao, Guanghua; Xie, Yang; Xu, Yanxun; Yang, Jichen; Yuan, Yuan; Zhang, Shihua; Zhang, Xiang Sun; Zhao, Junfei; Zuo, Chandler; Vlijmen, Herman Van W T; Westen, Gerard Van J P A community effort to assess and improve drug sensitivity prediction algorithms Journal Article Nature Biotechnology, 2014, ISSN: 15461696. @article{Costello2014, title = {A community effort to assess and improve drug sensitivity prediction algorithms}, author = {James C Costello and Laura M Heiser and Elisabeth Georgii and Mehmet Gönen and Michael P Menden and Nicholas J Wang and Mukesh Bansal and Muhammad Ammad-Ud-Din and Petteri Hintsanen and Suleiman A Khan and John Patrick Mpindi and Olli Kallioniemi and Antti Honkela and Tero Aittokallio and Krister Wennerberg and James J Collins and Dan Gallahan and Dinah Singer and Julio Saez-Rodriguez and Samuel Kaski and Joe W Gray and Gustavo Stolovitzky and Jean Paul Abbuehl and Jeffrey Allen and Russ B Altman and Shawn Balcome and Alexis Battle and Andreas Bender and Bonnie Berger and Jonathan Bernard and Madhuchhanda Bhattacharjee and Krithika Bhuvaneshwar and Andrew A Bieberich and Fred Boehm and Andrea Califano and Christina Chan and Beibei Chen and Ting Huei Chen and Jaejoon Choi and Luis Pedro Coelho and Thomas Cokelaer and James C Collins and Chad J Creighton and Jike Cui and Will Dampier and Jo V Davisson and Bernard De Baets and Raamesh Deshpande and Barbara DiCamillo and Murat Dundar and Zhana Duren and Adam Ertel and Haoyang Fan and Hongbin Fang and Robinder Gauba and Assaf Gottlieb and Michael Grau and Yuriy Gusev and Min Jin Ha and Leng Han and Michael Harris and Nicholas Henderson and Hussein A Hejase and Krisztian Homicsko and Jack P Hou and Woochang Hwang and Adriaan P IJzerman and Bilge Karacali and Sunduz Keles and Christina Kendziorski and Junho Kim and Min Kim and Youngchul Kim and David A Knowles and Daphne Koller and Junehawk Lee and Jae K Lee and Eelke B Lenselink and Biao Li and Bin Li and Jun Li and Han Liang and Jian Ma and Subha Madhavan and Sean Mooney and Chad L Myers and Michael A Newton and John P Overington and Ranadip Pal and Jian Peng and Richard Pestell and Robert J Prill and Peng Qiu and Bartek Rajwa and Anguraj Sadanandam and Francesco Sambo and Hyunjin Shin and Jiuzhou Song and Lei Song and Arvind Sridhar and Michiel Stock and Wei Sun and Tram Ta and Mahlet Tadesse and Ming Tan and Hao Tang and Dan Theodorescu and Gianna Maria Toffolo and Aydin Tozeren and William Trepicchio and Nelle Varoquaux and Jean Philippe Vert and Willem Waegeman and Thomas Walter and Qian Wan and Difei Wang and Wen Wang and Yong Wang and Zhishi Wang and Joerg K Wegner and Tongtong Wu and Tian Xia and Guanghua Xiao and Yang Xie and Yanxun Xu and Jichen Yang and Yuan Yuan and Shihua Zhang and Xiang Sun Zhang and Junfei Zhao and Chandler Zuo and Herman Van W T Vlijmen and Gerard Van J P Westen}, doi = {10.1038/nbt.2877}, issn = {15461696}, year = {2014}, date = {2014-01-01}, journal = {Nature Biotechnology}, abstract = {Predicting the best treatment strategy from genomic information is a core goal of precision medicine. Here we focus on predicting drug response based on a cohort of genomic, epigenomic and proteomic profiling data sets measured in human breast cancer cell lines. Through a collaborative effort between the National Cancer Institute (NCI) and the Dialogue on Reverse Engineering Assessment and Methods (DREAM) project, we analyzed a total of 44 drug sensitivity prediction algorithms. The top-performing approaches modeled nonlinear relationships and incorporated biological pathway information. We found that gene expression microarrays consistently provided the best predictive power of the individual profiling data sets; however, performance was increased by including multiple, independent data sets. We discuss the innovations underlying the top-performing methodology, Bayesian multitask MKL, and we provide detailed descriptions of all methods. This study establishes benchmarks for drug sensitivity prediction and identifies approaches that can be leveraged for the development of new methods.}, keywords = {}, pubstate = {published}, tppubtype = {article} } Predicting the best treatment strategy from genomic information is a core goal of precision medicine. Here we focus on predicting drug response based on a cohort of genomic, epigenomic and proteomic profiling data sets measured in human breast cancer cell lines. Through a collaborative effort between the National Cancer Institute (NCI) and the Dialogue on Reverse Engineering Assessment and Methods (DREAM) project, we analyzed a total of 44 drug sensitivity prediction algorithms. The top-performing approaches modeled nonlinear relationships and incorporated biological pathway information. We found that gene expression microarrays consistently provided the best predictive power of the individual profiling data sets; however, performance was increased by including multiple, independent data sets. We discuss the innovations underlying the top-performing methodology, Bayesian multitask MKL, and we provide detailed descriptions of all methods. This study establishes benchmarks for drug sensitivity prediction and identifies approaches that can be leveraged for the development of new methods. |
Doǧan, Tunca; Karacali, Bilge Automatic Identification of Highly Conserved Family Regions and Relationships in Genome Wide Datasets Including Remote Protein Sequences Journal Article PLoS ONE, 2013, ISSN: 19326203. @article{Dogan2013, title = {Automatic Identification of Highly Conserved Family Regions and Relationships in Genome Wide Datasets Including Remote Protein Sequences}, author = {Tunca Doǧan and Bilge Karacali}, doi = {10.1371/journal.pone.0075458}, issn = {19326203}, year = {2013}, date = {2013-01-01}, journal = {PLoS ONE}, abstract = {Identifying shared sequence segments along amino acid sequences generally requires a collection of closely related proteins, most often curated manually from the sequence datasets to suit the purpose at hand. Currently developed statistical methods are strained, however, when the collection contains remote sequences with poor alignment to the rest, or sequences containing multiple domains. In this paper, we propose a completely unsupervised and automated method to identify the shared sequence segments observed in a diverse collection of protein sequences including those present in a smaller fraction of the sequences in the collection, using a combination of sequence alignment, residue conservation scoring and graph-theoretical approaches. Since shared sequence fragments often imply conserved functional or structural attributes, the method produces a table of associations between the sequences and the identified conserved regions that can reveal previously unknown protein families as well as new members to existing ones. We evaluated the biological relevance of the method by clustering the proteins in gold standard datasets and assessing the clustering performance in comparison with previous methods from the literature. We have then applied the proposed method to a genome wide dataset of 17793 human proteins and generated a global association map to each of the 4753 identified conserved regions. Investigations on the major conserved regions revealed that they corresponded strongly to annotated structural domains. This suggests that the method can be useful in predicting novel domains on protein sequences.}, keywords = {}, pubstate = {published}, tppubtype = {article} } Identifying shared sequence segments along amino acid sequences generally requires a collection of closely related proteins, most often curated manually from the sequence datasets to suit the purpose at hand. Currently developed statistical methods are strained, however, when the collection contains remote sequences with poor alignment to the rest, or sequences containing multiple domains. In this paper, we propose a completely unsupervised and automated method to identify the shared sequence segments observed in a diverse collection of protein sequences including those present in a smaller fraction of the sequences in the collection, using a combination of sequence alignment, residue conservation scoring and graph-theoretical approaches. Since shared sequence fragments often imply conserved functional or structural attributes, the method produces a table of associations between the sequences and the identified conserved regions that can reveal previously unknown protein families as well as new members to existing ones. We evaluated the biological relevance of the method by clustering the proteins in gold standard datasets and assessing the clustering performance in comparison with previous methods from the literature. We have then applied the proposed method to a genome wide dataset of 17793 human proteins and generated a global association map to each of the 4753 identified conserved regions. Investigations on the major conserved regions revealed that they corresponded strongly to annotated structural domains. This suggests that the method can be useful in predicting novel domains on protein sequences. |
Unlu, Mehtat; Cetinayak, Hasan Oguz; Onder, Devrim; Ecevit, Cenk; Akman, Fadime; Ikiz, Ahmet Omer; Ada, Emel; Karacali, Bilge; Sarioglu, Sulen The prognostic value of tumor-stroma proportion in laryngeal squamous cell carcinoma Journal Article Turkish Journal of Pathology, 2013, ISSN: 1018-5615. @article{Unlu2013, title = {The prognostic value of tumor-stroma proportion in laryngeal squamous cell carcinoma}, author = {Mehtat Unlu and Hasan Oguz Cetinayak and Devrim Onder and Cenk Ecevit and Fadime Akman and Ahmet Omer Ikiz and Emel Ada and Bilge Karacali and Sulen Sarioglu}, doi = {10.5146/tjpath.2013.01144}, issn = {1018-5615}, year = {2013}, date = {2013-01-01}, journal = {Turkish Journal of Pathology}, abstract = {OBJECTIVE: Tumor-stroma proportion of tumor has been presented as a prognostic factor in some types of adenocarcinomas, but there is no information about squamous cell carcinomas and laryngeal carcinomas.$backslash$n$backslash$nMATERIAL AND METHOD: Five digital images of the tumor sections were obtained from 85 laryngeal carcinomas. Proportion of epithelial tumor component and stroma were measured by a software tool, allowing the pathologists to mark 205.6 $mu$m2 blocks on areas as carcinomatous/stromal, by clicking at the image. Totally, 3.451 mm2 tumor areas have been marked to 16.785 small square blocks for each case.$backslash$n$backslash$nRESULTS: Median follow up was 48 months (range 3-194). The mean tumor-stroma proportion was 48.63+18.18. There was no difference for tumor-stroma proportion when tumor location, grade, stage and perinodal invasion were considered. Although the following results were statistically insignificant, the mean tumor-stroma proportion was the lowest (37.46±12.49) for subglottic carcinomas, and it was 52.41±37.47, 50.86+19.84 and 44.56±16.91 for supraglottic, transglottic and glottic cases. The tumor-stroma proportion was lowest in cases with perinodal invasion and the highest in cases without lymph node metastasis (44.72±20.23, 47.77±17.37, 50.05±17.34). Tumor-stroma proportion was higher in the basaloid subtype compared with the classical squamous cell carcinoma (53.76±14.70 and 48.63±18.38 respectively). The overall and disease-free survival analysis did not reveal significance for tumor-stroma proportion (p=0.08}, keywords = {}, pubstate = {published}, tppubtype = {article} } OBJECTIVE: Tumor-stroma proportion of tumor has been presented as a prognostic factor in some types of adenocarcinomas, but there is no information about squamous cell carcinomas and laryngeal carcinomas.$backslash$n$backslash$nMATERIAL AND METHOD: Five digital images of the tumor sections were obtained from 85 laryngeal carcinomas. Proportion of epithelial tumor component and stroma were measured by a software tool, allowing the pathologists to mark 205.6 $mu$m2 blocks on areas as carcinomatous/stromal, by clicking at the image. Totally, 3.451 mm2 tumor areas have been marked to 16.785 small square blocks for each case.$backslash$n$backslash$nRESULTS: Median follow up was 48 months (range 3-194). The mean tumor-stroma proportion was 48.63+18.18. There was no difference for tumor-stroma proportion when tumor location, grade, stage and perinodal invasion were considered. Although the following results were statistically insignificant, the mean tumor-stroma proportion was the lowest (37.46±12.49) for subglottic carcinomas, and it was 52.41±37.47, 50.86+19.84 and 44.56±16.91 for supraglottic, transglottic and glottic cases. The tumor-stroma proportion was lowest in cases with perinodal invasion and the highest in cases without lymph node metastasis (44.72±20.23, 47.77±17.37, 50.05±17.34). Tumor-stroma proportion was higher in the basaloid subtype compared with the classical squamous cell carcinoma (53.76±14.70 and 48.63±18.38 respectively). The overall and disease-free survival analysis did not reveal significance for tumor-stroma proportion (p=0.08 |
Ünlü, Mehtat; cetinayak, Hasan Oǧuz; Önder, Devrim; Ecevit, Cenk; Akman, Fadime; Ikiz, Ahmet Ömer; Ada, Emel; Karacali, Bilge; Sarioǧlu, Sülen Laringeal skuamöz hücreli karsinomlarda tümör-stroma oranidotlessnidotlessn prognostik deǧeri Journal Article Turk Patoloji Dergisi/Turkish Journal of Pathology, 2013, ISSN: 10185615. @article{Unlu2013a, title = {Laringeal skuamöz hücreli karsinomlarda tümör-stroma oranidotlessnidotlessn prognostik deǧeri}, author = {Mehtat Ünlü and Hasan Oǧuz cetinayak and Devrim Önder and Cenk Ecevit and Fadime Akman and Ahmet Ömer Ikiz and Emel Ada and Bilge Karacali and Sülen Sarioǧlu}, doi = {10.5146/tjpath.2013.01144}, issn = {10185615}, year = {2013}, date = {2013-01-01}, journal = {Turk Patoloji Dergisi/Turkish Journal of Pathology}, abstract = {Objective: Tumor-stroma proportion of tumor has been presented as a prognostic factor in some types of adenocarcinomas, but there is no information about squamous cell carcinomas and laryngeal carcinomas. Material and Method: Five digital images of the tumor sections were obtained from 85 laryngeal carcinomas. Proportion of epithelial tumor component and stroma were measured by a software tool, allowing the pathologists to mark 205.6 $mu$m2 blocks on areas as carcinomatous/stromal, by clicking at the image. Totally, 3.451 mm2 tumor areas have been marked to 16.785 small square blocks for each case. Results: Median follow up was 48 months (range 3-194). The mean tumor-stroma proportion was 48.63+18.18. There was no difference for tumor-stroma proportion when tumor location, grade, stage and perinodal invasion were considered. Although the following results were statistically insignificant, the mean tumor-stroma proportion was the lowest (37.46±12.49) for subglottic carcinomas, and it was 52.41±37.47, 50.86+19.84 and 44.56±16.91 for supraglottic, transglottic and glottic cases. The tumor-stroma proportion was lowest in cases with perinodal invasion and the highest in cases without lymph node metastasis (44.72±20.23, 47.77±17.37, 50.05±17.34). Tumor-stroma proportion was higher in the basaloid subtype compared with the classical squamous cell carcinoma (53.76±14.70 and 48.63±18.38 respectively). The overall and disease-free survival analysis did not reveal significance for tumor-stroma proportion (p=0.08}, keywords = {}, pubstate = {published}, tppubtype = {article} } Objective: Tumor-stroma proportion of tumor has been presented as a prognostic factor in some types of adenocarcinomas, but there is no information about squamous cell carcinomas and laryngeal carcinomas. Material and Method: Five digital images of the tumor sections were obtained from 85 laryngeal carcinomas. Proportion of epithelial tumor component and stroma were measured by a software tool, allowing the pathologists to mark 205.6 $mu$m2 blocks on areas as carcinomatous/stromal, by clicking at the image. Totally, 3.451 mm2 tumor areas have been marked to 16.785 small square blocks for each case. Results: Median follow up was 48 months (range 3-194). The mean tumor-stroma proportion was 48.63+18.18. There was no difference for tumor-stroma proportion when tumor location, grade, stage and perinodal invasion were considered. Although the following results were statistically insignificant, the mean tumor-stroma proportion was the lowest (37.46±12.49) for subglottic carcinomas, and it was 52.41±37.47, 50.86+19.84 and 44.56±16.91 for supraglottic, transglottic and glottic cases. The tumor-stroma proportion was lowest in cases with perinodal invasion and the highest in cases without lymph node metastasis (44.72±20.23, 47.77±17.37, 50.05±17.34). Tumor-stroma proportion was higher in the basaloid subtype compared with the classical squamous cell carcinoma (53.76±14.70 and 48.63±18.38 respectively). The overall and disease-free survival analysis did not reveal significance for tumor-stroma proportion (p=0.08 |
Karacali, Bilge Hierarchical motif vectors for prediction of functional sites in amino acid sequences using quasi-supervised learning Journal Article IEEE/ACM Transactions on Computational Biology and Bioinformatics, 2012, ISSN: 15455963. @article{Karacali2012, title = {Hierarchical motif vectors for prediction of functional sites in amino acid sequences using quasi-supervised learning}, author = {Bilge Karacali}, doi = {10.1109/TCBB.2012.68}, issn = {15455963}, year = {2012}, date = {2012-01-01}, journal = {IEEE/ACM Transactions on Computational Biology and Bioinformatics}, abstract = {We propose hierarchical motif vectors to represent local amino acid sequence configurations for predicting the functional attributes of amino acid sites on a global scale in a quasi-supervised learning framework. The motif vectors are constructed via wavelet decomposition on the variations of physico-chemical amino acid properties along the sequences. We then formulate a prediction scheme for the functional attributes of amino acid sites in terms of the respective motif vectors using the quasi-supervised learning algorithm that carries out predictions for all sites in consideration using only the experimentally verified sites. We have carried out comparative performance evaluation of the proposed method on the prediction of N-glycosylation of 55,184 sites possessing the consensus N-glycosylation sequon identified over 15,104 human proteins, out of which only 1,939 were experimentally verified N-glycosylation sites. In the experiments, the proposed method achieved better predictive performance than the alternative strategies from the literature. In addition, the predicted N-glycosylation sites showed good agreement with existing potential annotations, while the novel predictions belonged to proteins known to be modified by glycosylation.}, keywords = {}, pubstate = {published}, tppubtype = {article} } We propose hierarchical motif vectors to represent local amino acid sequence configurations for predicting the functional attributes of amino acid sites on a global scale in a quasi-supervised learning framework. The motif vectors are constructed via wavelet decomposition on the variations of physico-chemical amino acid properties along the sequences. We then formulate a prediction scheme for the functional attributes of amino acid sites in terms of the respective motif vectors using the quasi-supervised learning algorithm that carries out predictions for all sites in consideration using only the experimentally verified sites. We have carried out comparative performance evaluation of the proposed method on the prediction of N-glycosylation of 55,184 sites possessing the consensus N-glycosylation sequon identified over 15,104 human proteins, out of which only 1,939 were experimentally verified N-glycosylation sites. In the experiments, the proposed method achieved better predictive performance than the alternative strategies from the literature. In addition, the predicted N-glycosylation sites showed good agreement with existing potential annotations, while the novel predictions belonged to proteins known to be modified by glycosylation. |
Karacali, Bilge Quasi-supervised learning for biomedical data analysis Journal Article Pattern Recognition, 2010, ISSN: 00313203. @article{Karacali2010, title = {Quasi-supervised learning for biomedical data analysis}, author = {Bilge Karacali}, doi = {10.1016/j.patcog.2010.04.024}, issn = {00313203}, year = {2010}, date = {2010-01-01}, journal = {Pattern Recognition}, abstract = {We present a novel formulation for pattern recognition in biomedical data. We adopt a binary recognition scenario where a control dataset contains samples of one class only, while a mixed dataset contains an unlabeled collection of samples from both classes. The mixed dataset samples that belong to the second class are identified by estimating posterior probabilities of samples for being in the control or the mixed datasets. Experiments on synthetic data established a better detection performance against possible alternatives. The fitness of the method in biomedical data analysis was further demonstrated on real multi-color flow cytometry and multi-channel electroencephalography data. textcopyright 2010 Elsevier Ltd. All rights reserved.}, keywords = {}, pubstate = {published}, tppubtype = {article} } We present a novel formulation for pattern recognition in biomedical data. We adopt a binary recognition scenario where a control dataset contains samples of one class only, while a mixed dataset contains an unlabeled collection of samples from both classes. The mixed dataset samples that belong to the second class are identified by estimating posterior probabilities of samples for being in the control or the mixed datasets. Experiments on synthetic data established a better detection performance against possible alternatives. The fitness of the method in biomedical data analysis was further demonstrated on real multi-color flow cytometry and multi-channel electroencephalography data. textcopyright 2010 Elsevier Ltd. All rights reserved. |
Karacali, Bilge; Tözeren, Aydin Automated detection of regions of interest for tissue microarray experiments: An image texture analysis Journal Article BMC Medical Imaging, 2007, ISSN: 14712342. @article{Karacali2007, title = {Automated detection of regions of interest for tissue microarray experiments: An image texture analysis}, author = {Bilge Karacali and Aydin Tözeren}, doi = {10.1186/1471-2342-7-2}, issn = {14712342}, year = {2007}, date = {2007-01-01}, journal = {BMC Medical Imaging}, abstract = {BACKGROUND: Recent research with tissue microarrays led to a rapid progress toward quantifying the expressions of large sets of biomarkers in normal and diseased tissue. However, standard procedures for sampling tissue for molecular profiling have not yet been established. METHODS: This study presents a high throughput analysis of texture heterogeneity on breast tissue images for the purpose of identifying regions of interest in the tissue for molecular profiling via tissue microarray technology. Image texture of breast histology slides was described in terms of three parameters: the percentage of area occupied in an image block by chromatin (B), percentage occupied by stroma-like regions (P), and a statistical heterogeneity index H commonly used in image analysis. Texture parameters were defined and computed for each of the thousands of image blocks in our dataset using both the gray scale and color segmentation. The image blocks were then classified into three categories using the texture feature parameters in a novel statistical learning algorithm. These categories are as follows: image blocks specific to normal breast tissue, blocks specific to cancerous tissue, and those image blocks that are non-specific to normal and disease states. RESULTS: Gray scale and color segmentation techniques led to identification of same regions in histology slides as cancer-specific. Moreover the image blocks identified as cancer-specific belonged to those cell crowded regions in whole section image slides that were marked by two pathologists as regions of interest for further histological studies. CONCLUSION: These results indicate the high efficiency of our automated method for identifying pathologic regions of interest on histology slides. Automation of critical region identification will help minimize the inter-rater variability among different raters (pathologists) as hundreds of tumors that are used to develop an array have typically been evaluated (graded) by different pathologists. The region of interest information gathered from the whole section images will guide the excision of tissue for constructing tissue microarrays and for high throughput profiling of global gene expression.}, keywords = {}, pubstate = {published}, tppubtype = {article} } BACKGROUND: Recent research with tissue microarrays led to a rapid progress toward quantifying the expressions of large sets of biomarkers in normal and diseased tissue. However, standard procedures for sampling tissue for molecular profiling have not yet been established. METHODS: This study presents a high throughput analysis of texture heterogeneity on breast tissue images for the purpose of identifying regions of interest in the tissue for molecular profiling via tissue microarray technology. Image texture of breast histology slides was described in terms of three parameters: the percentage of area occupied in an image block by chromatin (B), percentage occupied by stroma-like regions (P), and a statistical heterogeneity index H commonly used in image analysis. Texture parameters were defined and computed for each of the thousands of image blocks in our dataset using both the gray scale and color segmentation. The image blocks were then classified into three categories using the texture feature parameters in a novel statistical learning algorithm. These categories are as follows: image blocks specific to normal breast tissue, blocks specific to cancerous tissue, and those image blocks that are non-specific to normal and disease states. RESULTS: Gray scale and color segmentation techniques led to identification of same regions in histology slides as cancer-specific. Moreover the image blocks identified as cancer-specific belonged to those cell crowded regions in whole section image slides that were marked by two pathologists as regions of interest for further histological studies. CONCLUSION: These results indicate the high efficiency of our automated method for identifying pathologic regions of interest on histology slides. Automation of critical region identification will help minimize the inter-rater variability among different raters (pathologists) as hundreds of tumors that are used to develop an array have typically been evaluated (graded) by different pathologists. The region of interest information gathered from the whole section images will guide the excision of tissue for constructing tissue microarrays and for high throughput profiling of global gene expression. |
Karacali, Bilge Information theoretic deformable registration using local image information Journal Article International Journal of Computer Vision, 2007, ISSN: 09205691. @article{Karacali2007a, title = {Information theoretic deformable registration using local image information}, author = {Bilge Karacali}, doi = {10.1007/s11263-006-8704-0}, issn = {09205691}, year = {2007}, date = {2007-01-01}, journal = {International Journal of Computer Vision}, abstract = {We present a deformable registration algorithm for multi-modality images based on information theoretic similarity measures at the scale of individual image voxels. We derive analytical expressions for the mutual information, the joint entropy, and the sum of marginal entropies of two images over a small neighborhood in terms of image gradients. Using these expressions, we formulate image registration algorithms maximizing local similarity over the whole image domain in an energy minimization framework. This strategy produces highly elastic image alignment as the registration is driven by voxel similarities between the images, the algorithms are easily implementable using the closed-form expressions for the derivative of the optimization function with respect to the deformation, and avoid estimation of joint and marginal probability densities governing the image intensities essential to conventional information theoretic image registration methods.}, keywords = {}, pubstate = {published}, tppubtype = {article} } We present a deformable registration algorithm for multi-modality images based on information theoretic similarity measures at the scale of individual image voxels. We derive analytical expressions for the mutual information, the joint entropy, and the sum of marginal entropies of two images over a small neighborhood in terms of image gradients. Using these expressions, we formulate image registration algorithms maximizing local similarity over the whole image domain in an energy minimization framework. This strategy produces highly elastic image alignment as the registration is driven by voxel similarities between the images, the algorithms are easily implementable using the closed-form expressions for the derivative of the optimization function with respect to the deformation, and avoid estimation of joint and marginal probability densities governing the image intensities essential to conventional information theoretic image registration methods. |
Karacali, Bilge; Vamvakidou, Alexandra P; Tözeren, Aydin Automated recognition of cell phenotypes in histology images based on membrane- and nuclei-targeting biomarkers. Journal Article BMC medical imaging, 2007, ISSN: 14712342. @article{Karacali2007b, title = {Automated recognition of cell phenotypes in histology images based on membrane- and nuclei-targeting biomarkers.}, author = {Bilge Karacali and Alexandra P Vamvakidou and Aydin Tözeren}, doi = {10.1186/1471-2342-7-7}, issn = {14712342}, year = {2007}, date = {2007-01-01}, journal = {BMC medical imaging}, abstract = {BACKGROUND: Three-dimensional in vitro culture of cancer cells are used to predict the effects of prospective anti-cancer drugs in vivo. In this study, we present an automated image analysis protocol for detailed morphological protein marker profiling of tumoroid cross section images.$backslash$n$backslash$nMETHODS: Histologic cross sections of breast tumoroids developed in co-culture suspensions of breast cancer cell lines, stained for E-cadherin and progesterone receptor, were digitized and pixels in these images were classified into five categories using k-means clustering. Automated segmentation was used to identify image regions composed of cells expressing a given biomarker. Synthesized images were created to check the accuracy of the image processing system.$backslash$n$backslash$nRESULTS: Accuracy of automated segmentation was over 95% in identifying regions of interest in synthesized images. Image analysis of adjacent histology slides stained, respectively, for Ecad and PR, accurately predicted regions of different cell phenotypes. Image analysis of tumoroid cross sections from different tumoroids obtained under the same co-culture conditions indicated the variation of cellular composition from one tumoroid to another. Variations in the compositions of cross sections obtained from the same tumoroid were established by parallel analysis of Ecad and PR-stained cross section images.$backslash$n$backslash$nCONCLUSION: Proposed image analysis methods offer standardized high throughput profiling of molecular anatomy of tumoroids based on both membrane and nuclei markers that is suitable to rapid large scale investigations of anti-cancer compounds for drug development.}, keywords = {}, pubstate = {published}, tppubtype = {article} } BACKGROUND: Three-dimensional in vitro culture of cancer cells are used to predict the effects of prospective anti-cancer drugs in vivo. In this study, we present an automated image analysis protocol for detailed morphological protein marker profiling of tumoroid cross section images.$backslash$n$backslash$nMETHODS: Histologic cross sections of breast tumoroids developed in co-culture suspensions of breast cancer cell lines, stained for E-cadherin and progesterone receptor, were digitized and pixels in these images were classified into five categories using k-means clustering. Automated segmentation was used to identify image regions composed of cells expressing a given biomarker. Synthesized images were created to check the accuracy of the image processing system.$backslash$n$backslash$nRESULTS: Accuracy of automated segmentation was over 95% in identifying regions of interest in synthesized images. Image analysis of adjacent histology slides stained, respectively, for Ecad and PR, accurately predicted regions of different cell phenotypes. Image analysis of tumoroid cross sections from different tumoroids obtained under the same co-culture conditions indicated the variation of cellular composition from one tumoroid to another. Variations in the compositions of cross sections obtained from the same tumoroid were established by parallel analysis of Ecad and PR-stained cross section images.$backslash$n$backslash$nCONCLUSION: Proposed image analysis methods offer standardized high throughput profiling of molecular anatomy of tumoroids based on both membrane and nuclei markers that is suitable to rapid large scale investigations of anti-cancer compounds for drug development. |
Gormley, Michael; Dampier, William; Ertel, Adam; Karacali, Bilge; Tozeren, Aydin Prediction potential of candidate biomarker sets identified and validated on gene expression data from multiple datasets Journal Article BMC Bioinformatics, 2007, ISSN: 14712105. @article{Gormley2007, title = {Prediction potential of candidate biomarker sets identified and validated on gene expression data from multiple datasets}, author = {Michael Gormley and William Dampier and Adam Ertel and Bilge Karacali and Aydin Tozeren}, doi = {10.1186/1471-2105-8-415}, issn = {14712105}, year = {2007}, date = {2007-01-01}, journal = {BMC Bioinformatics}, abstract = {Background: Independently derived expression profiles of the same biological condition often have few genes in common. In this study, we created populations of expression profiles from publicly available microarray datasets of cancer (breast, lymphoma and renal) samples linked to clinical information with an iterative machine learning algorithm. ROC curves were used to assess the prediction error of each profile for classification. We compared the prediction error of profiles correlated with molecular phenotype against profiles correlated with relapse-free status. Prediction error of profiles identified with supervised univariate feature selection algorithms were compared to profiles selected randomly from a) all genes on the microarray platform and b) a list of known disease-related genes (a priori selection). We also determined the relevance of expression profiles on test arrays from independent datasets, measured on either the same or different microarray platforms. Results: Highly discriminative expression profiles were produced on both simulated gene expression data and expression data from breast cancer and lymphoma datasets on the basis of ER and BCL-6 expression, respectively. Use of relapse -free status to identify profiles for prognosis prediction resulted in poorly discriminative decision rules. Supervised feature selection resulted in more accurate classifications than random or a priori selection, however, the difference in prediction error decreased as the number of features increased. These results held when decision rules were applied across-datasets to samples profiled on the same microarray platform. Conclusion: Our results show that many gene sets predict molecular phenotypes accurately. Given this, expression profiles identified using different training datasets should be expected to show little agreement. In addition, we demonstrate the difficulty in predicting relapse directly from microarray data using supervised machine learning approaches. These findings are relevant to the use of molecular profiling for the identification of candidate biomarker panels. textcopyright 2007 Gormley et al; licensee BioMed Central Ltd.}, keywords = {}, pubstate = {published}, tppubtype = {article} } Background: Independently derived expression profiles of the same biological condition often have few genes in common. In this study, we created populations of expression profiles from publicly available microarray datasets of cancer (breast, lymphoma and renal) samples linked to clinical information with an iterative machine learning algorithm. ROC curves were used to assess the prediction error of each profile for classification. We compared the prediction error of profiles correlated with molecular phenotype against profiles correlated with relapse-free status. Prediction error of profiles identified with supervised univariate feature selection algorithms were compared to profiles selected randomly from a) all genes on the microarray platform and b) a list of known disease-related genes (a priori selection). We also determined the relevance of expression profiles on test arrays from independent datasets, measured on either the same or different microarray platforms. Results: Highly discriminative expression profiles were produced on both simulated gene expression data and expression data from breast cancer and lymphoma datasets on the basis of ER and BCL-6 expression, respectively. Use of relapse -free status to identify profiles for prognosis prediction resulted in poorly discriminative decision rules. Supervised feature selection resulted in more accurate classifications than random or a priori selection, however, the difference in prediction error decreased as the number of features increased. These results held when decision rules were applied across-datasets to samples profiled on the same microarray platform. Conclusion: Our results show that many gene sets predict molecular phenotypes accurately. Given this, expression profiles identified using different training datasets should be expected to show little agreement. In addition, we demonstrate the difficulty in predicting relapse directly from microarray data using supervised machine learning approaches. These findings are relevant to the use of molecular profiling for the identification of candidate biomarker panels. textcopyright 2007 Gormley et al; licensee BioMed Central Ltd. |
Karacali, Bilge; Davatzikos, Christos Simulation of tissue atrophy using a topology preserving transformation model Journal Article IEEE Transactions on Medical Imaging, 2006, ISSN: 02780062. @article{Karacali2006, title = {Simulation of tissue atrophy using a topology preserving transformation model}, author = {Bilge Karacali and Christos Davatzikos}, doi = {10.1109/TMI.2006.873221}, issn = {02780062}, year = {2006}, date = {2006-01-01}, journal = {IEEE Transactions on Medical Imaging}, abstract = {We propose a method to simulate atrophy and other similar volumetric change effects on medical images. Given a desired level of atrophy, we find a dense warping deformation that produces the corresponding levels of volumetric loss on the labeled tissue using an energy minimization strategy. Simulated results on a real brain image indicate that the method generates realistic images of tissue loss. The method does not make assumptions regarding the mechanics of tissue deformation, and provides a framework where a pre-specified pattern of atrophy can readily be simulated. Furthermore, it provides exact correspondences between images prior and posterior to the atrophy that can be used to evaluate provisional image registration and atrophy quantification algorithms.}, keywords = {}, pubstate = {published}, tppubtype = {article} } We propose a method to simulate atrophy and other similar volumetric change effects on medical images. Given a desired level of atrophy, we find a dense warping deformation that produces the corresponding levels of volumetric loss on the labeled tissue using an energy minimization strategy. Simulated results on a real brain image indicate that the method generates realistic images of tissue loss. The method does not make assumptions regarding the mechanics of tissue deformation, and provides a framework where a pre-specified pattern of atrophy can readily be simulated. Furthermore, it provides exact correspondences between images prior and posterior to the atrophy that can be used to evaluate provisional image registration and atrophy quantification algorithms. |
Xue, Zhong; Shen, Dinggang; Karacali, Bilge; Stern, Joshua; Rottenberg, David; Davatzikos, Christos Simulating deformations of MR brain images for validation of atlas-based segmentation and registration algorithms Journal Article NeuroImage, 2006, ISSN: 10538119. @article{Xue2006, title = {Simulating deformations of MR brain images for validation of atlas-based segmentation and registration algorithms}, author = {Zhong Xue and Dinggang Shen and Bilge Karacali and Joshua Stern and David Rottenberg and Christos Davatzikos}, doi = {10.1016/j.neuroimage.2006.08.007}, issn = {10538119}, year = {2006}, date = {2006-01-01}, journal = {NeuroImage}, abstract = {Simulated deformations and images can act as the gold standard for evaluating various template-based image segmentation and registration algorithms. Traditional deformable simulation methods, such as the use of analytic deformation fields or the displacement of landmarks followed by some form of interpolation, are often unable to construct rich (complex) and/or realistic deformations of anatomical organs. This paper presents new methods aiming to automatically simulate realistic inter- and intra-individual deformations. The paper first describes a statistical approach to capturing inter-individual variability of high-deformation fields from a number of examples (training samples). In this approach, Wavelet-Packet Transform (WPT) of the training deformations and their Jacobians, in conjunction with a Markov random field (MRF) spatial regularization, are used to capture both coarse and fine characteristics of the training deformations in a statistical fashion. Simulated deformations can then be constructed by randomly sampling the resultant statistical distribution in an unconstrained or a landmark-constrained fashion. The paper also describes a model for generating tissue atrophy or growth in order to simulate intra-individual brain deformations. Several sets of simulated deformation fields and respective images are generated, which can be used in the future for systematic and extensive validation studies of automated atlas-based segmentation and deformable registration methods. The code and simulated data are available through our Web site. textcopyright 2006 Elsevier Inc. All rights reserved.}, keywords = {}, pubstate = {published}, tppubtype = {article} } Simulated deformations and images can act as the gold standard for evaluating various template-based image segmentation and registration algorithms. Traditional deformable simulation methods, such as the use of analytic deformation fields or the displacement of landmarks followed by some form of interpolation, are often unable to construct rich (complex) and/or realistic deformations of anatomical organs. This paper presents new methods aiming to automatically simulate realistic inter- and intra-individual deformations. The paper first describes a statistical approach to capturing inter-individual variability of high-deformation fields from a number of examples (training samples). In this approach, Wavelet-Packet Transform (WPT) of the training deformations and their Jacobians, in conjunction with a Markov random field (MRF) spatial regularization, are used to capture both coarse and fine characteristics of the training deformations in a statistical fashion. Simulated deformations can then be constructed by randomly sampling the resultant statistical distribution in an unconstrained or a landmark-constrained fashion. The paper also describes a model for generating tissue atrophy or growth in order to simulate intra-individual brain deformations. Several sets of simulated deformation fields and respective images are generated, which can be used in the future for systematic and extensive validation studies of automated atlas-based segmentation and deformable registration methods. The code and simulated data are available through our Web site. textcopyright 2006 Elsevier Inc. All rights reserved. |
Karacali, Bilge; Davatzikos, Christos Estimating topology preserving and smooth displacement fields Journal Article IEEE Transactions on Medical Imaging, 2004, ISSN: 02780062. @article{Karacali2004, title = {Estimating topology preserving and smooth displacement fields}, author = {Bilge Karacali and Christos Davatzikos}, doi = {10.1109/TMI.2004.827963}, issn = {02780062}, year = {2004}, date = {2004-01-01}, journal = {IEEE Transactions on Medical Imaging}, abstract = {We propose a method for enforcing topology preservation and smoothness onto a given displacement field. We first analyze the conditions for topology preservation on two- and three-dimensional displacement fields over a discrete rectangular grid. We then pose the problem of finding the closest topology preserving displacement field in terms of its complete set of gradients, which we later solve using a cyclic projections framework. Adaptive smoothing of a displacement field is then formulated as an extension of topology preservation, via constraints imposed on the Jacobian of the displacement field. The simulation results indicate that this technique is a fast and reliable method to estimate a topology preserving displacement field from a noisy observation that does not necessarily preserve topology. They also show that the proposed smoothing method can render morphometric analysis methods that are based on displacement field of shape transformations more robust to noise without removing important morphologic characteristics.}, keywords = {}, pubstate = {published}, tppubtype = {article} } We propose a method for enforcing topology preservation and smoothness onto a given displacement field. We first analyze the conditions for topology preservation on two- and three-dimensional displacement fields over a discrete rectangular grid. We then pose the problem of finding the closest topology preserving displacement field in terms of its complete set of gradients, which we later solve using a cyclic projections framework. Adaptive smoothing of a displacement field is then formulated as an extension of topology preservation, via constraints imposed on the Jacobian of the displacement field. The simulation results indicate that this technique is a fast and reliable method to estimate a topology preserving displacement field from a noisy observation that does not necessarily preserve topology. They also show that the proposed smoothing method can render morphometric analysis methods that are based on displacement field of shape transformations more robust to noise without removing important morphologic characteristics. |
Lao, Zhiqiang; Shen, Dinggang; Xue, Zhong; Karacali, Bilge; Resnick, Susan M; Davatzikos, Christos Morphological classification of brains via high-dimensional shape transformations and machine learning methods Journal Article NeuroImage, 2004, ISSN: 10538119. @article{Lao2004, title = {Morphological classification of brains via high-dimensional shape transformations and machine learning methods}, author = {Zhiqiang Lao and Dinggang Shen and Zhong Xue and Bilge Karacali and Susan M Resnick and Christos Davatzikos}, doi = {10.1016/j.neuroimage.2003.09.027}, issn = {10538119}, year = {2004}, date = {2004-01-01}, journal = {NeuroImage}, abstract = {A high-dimensional shape transformation posed in a mass-preserving framework is used as a morphological signature of a brain image. Population differences with complex spatial patterns are then determined by applying a nonlinear support vector machine (SVM) pattern classification method to the morphological signatures. Significant reduction of the dimensionality of the morphological signatures is achieved via wavelet decomposition and feature reduction methods. Applying the method to MR images with simulated atrophy shows that the method can correctly detect subtle and spatially complex atrophy, even when the simulated atrophy represents only a 5% variation from the original image. Applying this method to actual MR images shows that brains can be correctly determined to be male or female with a successful classification rate of 97%, using the leave-one-out method. This proposed method also shows a high classification rate for old adults' age classification, even under difficult test scenarios. The main characteristic of the proposed methodology is that, by applying multivariate pattern classification methods, it can detect subtle and spatially complex patterns of morphological group differences which are often not detectable by voxel-based morphometric methods, because these methods analyze morphological measurements voxel-by-voxel and do not consider the entirety of the data simultaneously. textcopyright 2003 Elsevier Inc. All rights reserved.}, keywords = {}, pubstate = {published}, tppubtype = {article} } A high-dimensional shape transformation posed in a mass-preserving framework is used as a morphological signature of a brain image. Population differences with complex spatial patterns are then determined by applying a nonlinear support vector machine (SVM) pattern classification method to the morphological signatures. Significant reduction of the dimensionality of the morphological signatures is achieved via wavelet decomposition and feature reduction methods. Applying the method to MR images with simulated atrophy shows that the method can correctly detect subtle and spatially complex atrophy, even when the simulated atrophy represents only a 5% variation from the original image. Applying this method to actual MR images shows that brains can be correctly determined to be male or female with a successful classification rate of 97%, using the leave-one-out method. This proposed method also shows a high classification rate for old adults' age classification, even under difficult test scenarios. The main characteristic of the proposed methodology is that, by applying multivariate pattern classification methods, it can detect subtle and spatially complex patterns of morphological group differences which are often not detectable by voxel-based morphometric methods, because these methods analyze morphological measurements voxel-by-voxel and do not consider the entirety of the data simultaneously. textcopyright 2003 Elsevier Inc. All rights reserved. |
Karacali, Bilge; Snyder, Wesley Reconstructing discontinuous surfaces from a given gradient field using partial integrability Journal Article Computer Vision and Image Understanding, 2003, ISSN: 10773142. @article{Karacali2003, title = {Reconstructing discontinuous surfaces from a given gradient field using partial integrability}, author = {Bilge Karacali and Wesley Snyder}, doi = {10.1016/S1077-3142(03)00095-X}, issn = {10773142}, year = {2003}, date = {2003-01-01}, journal = {Computer Vision and Image Understanding}, abstract = {This paper describes an adaptive surface reconstruction method from a given gradient field that allows discontinuities in the solution. We first formalize a vector space projection technique to reconstruct a surface with a uniformly integrable gradient field that corresponds to the minimum norm solution in the gradient space over discrete imaging settings. Next, we generalize this technique to reconstruct minimum norm solution surfaces with partially integrable gradient fields, where partial integrability is characterized adaptively from the given gradient field using multi-scale gradient space expansions. The simulations on synthesized and real data using block processing techniques indicate that the proposed method provides fast and reliable surface reconstruction through accurate characterization of embedded partial integrability in a given gradient field. textcopyright 2003 Elsevier Inc. All rights reserved.}, keywords = {}, pubstate = {published}, tppubtype = {article} } This paper describes an adaptive surface reconstruction method from a given gradient field that allows discontinuities in the solution. We first formalize a vector space projection technique to reconstruct a surface with a uniformly integrable gradient field that corresponds to the minimum norm solution in the gradient space over discrete imaging settings. Next, we generalize this technique to reconstruct minimum norm solution surfaces with partially integrable gradient fields, where partial integrability is characterized adaptively from the given gradient field using multi-scale gradient space expansions. The simulations on synthesized and real data using block processing techniques indicate that the proposed method provides fast and reliable surface reconstruction through accurate characterization of embedded partial integrability in a given gradient field. textcopyright 2003 Elsevier Inc. All rights reserved. |
Karacali, Bilge; Krim, Hamid Fast minimization of structural risk by nearest neighbor rule Journal Article IEEE Transactions on Neural Networks, 2003, ISSN: 10459227. @article{Karacali2003a, title = {Fast minimization of structural risk by nearest neighbor rule}, author = {Bilge Karacali and Hamid Krim}, doi = {10.1109/TNN.2002.804315}, issn = {10459227}, year = {2003}, date = {2003-01-01}, journal = {IEEE Transactions on Neural Networks}, abstract = {In this paper, we present a novel nearest neighbor rule-based implementation of the structural risk minimization principle to address a generic classification problem. We propose a fast reference set thinning algorithm on the training data set similar to a support vector machine (SVM) approach. We then show that the nearest neighbor rule based on the reduced set implements the structural risk minimization principle, in a manner which does not involve selection of a convenient feature space. Simulation results on real data indicate that this method significantly reduces the computational cost of the conventional SVMs, and achieves a nearly comparable test error performance.}, keywords = {}, pubstate = {published}, tppubtype = {article} } In this paper, we present a novel nearest neighbor rule-based implementation of the structural risk minimization principle to address a generic classification problem. We propose a fast reference set thinning algorithm on the training data set similar to a support vector machine (SVM) approach. We then show that the nearest neighbor rule based on the reduced set implements the structural risk minimization principle, in a manner which does not involve selection of a convenient feature space. Simulation results on real data indicate that this method significantly reduces the computational cost of the conventional SVMs, and achieves a nearly comparable test error performance. |
Incollections |
Karacalı, Bilge; Davatzikos, Christos Topology Preservation and Regularity in Estimated Deformation Fields Incollection 2010. @incollection{Karacal2010, title = {Topology Preservation and Regularity in Estimated Deformation Fields}, author = {Bilge Karacalı and Christos Davatzikos}, doi = {10.1007/978-3-540-45087-0_36}, year = {2010}, date = {2010-01-01}, abstract = {A general formalism to impose topology preserving regularity on a given irregular deformation field is presented. The topology preservation conditions are derived with regard to the discrete approximations to the deformation field Jacobian in a two-dimensional image registration problem. The problem of enforcing topology preservation onto a given deformation field is formulated in terms of the deformation gradients, and solved using a cyclic projections approach. The generalization of the developed algorithm leads to a deformation field regularity control by limiting the per voxel volumetric change within a prescribed interval. Extension of the topology preservation conditions onto a three-dimensional registration problem is also presented, together with a comparative analysis of the proposed algorithm with respect to a Gaussian regularizer that enforces the same topology preservation conditions.}, keywords = {}, pubstate = {published}, tppubtype = {incollection} } A general formalism to impose topology preserving regularity on a given irregular deformation field is presented. The topology preservation conditions are derived with regard to the discrete approximations to the deformation field Jacobian in a two-dimensional image registration problem. The problem of enforcing topology preservation onto a given deformation field is formulated in terms of the deformation gradients, and solved using a cyclic projections approach. The generalization of the developed algorithm leads to a deformation field regularity control by limiting the per voxel volumetric change within a prescribed interval. Extension of the topology preservation conditions onto a three-dimensional registration problem is also presented, together with a comparative analysis of the proposed algorithm with respect to a Gaussian regularizer that enforces the same topology preservation conditions. |
Inproceedings |
caǧdaş, Serhat; Karacali, Bilge Novel techniques for model-free and fast computation of mutual information Inproceedings 26th IEEE Signal Processing and Communications Applications Conference, SIU 2018, 2018, ISBN: 9781538615010. @inproceedings{Cagdas2018, title = {Novel techniques for model-free and fast computation of mutual information}, author = {Serhat caǧdaş and Bilge Karacali}, doi = {10.1109/SIU.2018.8404637}, isbn = {9781538615010}, year = {2018}, date = {2018-01-01}, booktitle = {26th IEEE Signal Processing and Communications Applications Conference, SIU 2018}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } |
Olcay, Bilal Orkan; Karacali, Bilge; Ozgoren, Murat; Guducu, Cagdas Brain activity characterization by entropic clustering of EEG signals Inproceedings 2017. @inproceedings{Olcay2017, title = {Brain activity characterization by entropic clustering of EEG signals}, author = {Bilal Orkan Olcay and Bilge Karacali and Murat Ozgoren and Cagdas Guducu}, doi = {10.1109/siu.2017.7960503}, year = {2017}, date = {2017-01-01}, abstract = {textcopyright 2017 IEEE. In this study, two novel entropy and mutual information based algorithms have been proposed to characterize the stimulus specific brain activity. In the first method, inter-channel mutual information of electroencephalography signals has been calculated and the channels that exhibit synchronized behaivour during stimulus. In the second method, the responsiveness of the individual channels has been characterized in an entropic manner and then, the channels which demonstrates stimulus specific entropic behavior have been obtained. The performance of the proposed methods has been simulated on a real dataset obtained from Dokuz Eylul University Brain Biophysics laboratory. The results demonstrate that different regions of the brain exhibit a coherent activity during stimulus.}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } textcopyright 2017 IEEE. In this study, two novel entropy and mutual information based algorithms have been proposed to characterize the stimulus specific brain activity. In the first method, inter-channel mutual information of electroencephalography signals has been calculated and the channels that exhibit synchronized behaivour during stimulus. In the second method, the responsiveness of the individual channels has been characterized in an entropic manner and then, the channels which demonstrates stimulus specific entropic behavior have been obtained. The performance of the proposed methods has been simulated on a real dataset obtained from Dokuz Eylul University Brain Biophysics laboratory. The results demonstrate that different regions of the brain exhibit a coherent activity during stimulus. |
Güzel, Başak Esin Köktürk; Karacalı, Bilge Fisher's Linear Discriminant Analysis Based Prediction using Transient Features of Seismic Events in Coal Mines Inproceedings Proceedings of the 2016 Federated Conference on Computer Science and Information Systems, 2016. @inproceedings{Guzel2016, title = {Fisher's Linear Discriminant Analysis Based Prediction using Transient Features of Seismic Events in Coal Mines}, author = {Başak Esin Köktürk Güzel and Bilge Karacalı}, doi = {10.15439/2016f116}, year = {2016}, date = {2016-01-01}, booktitle = {Proceedings of the 2016 Federated Conference on Computer Science and Information Systems}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } |
Olcay, Orkan B; Ozgoren, Murat; Karacali, Bilge Imaginary activity recognition using inter-channel time coherence profiles in EEG data Inproceedings 2016. @inproceedings{Olcay2016, title = {Imaginary activity recognition using inter-channel time coherence profiles in EEG data}, author = {Orkan B Olcay and Murat Ozgoren and Bilge Karacali}, doi = {10.1109/siu.2016.7496002}, year = {2016}, date = {2016-01-01}, abstract = {textcopyright 2016 IEEE. In this study, we have carried out a brain-computer interface study that uses time delays between electrodes. As features, we have calculated the time delays that maximizes the absolute value of the cross-covariance between the chosen reference channel and the remaining channels. Performance results of 3 out of 5 participants that were nearly at a %100 accuracy level along with a relatively smaller number of training data, and a lack of similar studies indicate that the proposed approach is open to the further improvements.}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } textcopyright 2016 IEEE. In this study, we have carried out a brain-computer interface study that uses time delays between electrodes. As features, we have calculated the time delays that maximizes the absolute value of the cross-covariance between the chosen reference channel and the remaining channels. Performance results of 3 out of 5 participants that were nearly at a %100 accuracy level along with a relatively smaller number of training data, and a lack of similar studies indicate that the proposed approach is open to the further improvements. |
Köktürk, Başak Esin; Karacali, Bilge Model-free expectation maximization for divisive hierarchical clustering of multicolor flow cytometry data Inproceedings Proceedings - 2014 IEEE International Conference on Bioinformatics and Biomedicine, IEEE BIBM 2014, 2014, ISBN: 9781479956692. @inproceedings{Kokturk2014, title = {Model-free expectation maximization for divisive hierarchical clustering of multicolor flow cytometry data}, author = {Başak Esin Köktürk and Bilge Karacali}, doi = {10.1109/BIBM.2014.6999166}, isbn = {9781479956692}, year = {2014}, date = {2014-01-01}, booktitle = {Proceedings - 2014 IEEE International Conference on Bioinformatics and Biomedicine, IEEE BIBM 2014}, abstract = {textcopyright 2014 IEEE.This paper proposes a new method for automated clustering of high dimensional datasets. The method is based on a recursive binary division strategy that successively divides an original dataset into distinct clusters. Each binary division is carried out using a model-free expectation maximization scheme that exploits the posterior probability computation capability of the quasi-supervised learning algorithm. The divisions are carried out until a division cost exceeds an adaptively determined limit. Experiment results on synthetic as well as real multi-color flow cytometry datasets showed that the proposed method can accurately capture the prominent clusters without requiring any knowledge on the number of clusters or their distribution models.}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } textcopyright 2014 IEEE.This paper proposes a new method for automated clustering of high dimensional datasets. The method is based on a recursive binary division strategy that successively divides an original dataset into distinct clusters. Each binary division is carried out using a model-free expectation maximization scheme that exploits the posterior probability computation capability of the quasi-supervised learning algorithm. The divisions are carried out until a division cost exceeds an adaptively determined limit. Experiment results on synthetic as well as real multi-color flow cytometry datasets showed that the proposed method can accurately capture the prominent clusters without requiring any knowledge on the number of clusters or their distribution models. |
Bozkurt, Baris; Karaosmanoglu, Kemal M; Karacali, Bilge; Unal, Erdem Automatic melodic segmentation of Turkish makam music scores Inproceedings 2014. @inproceedings{Bozkurt2014a, title = {Automatic melodic segmentation of Turkish makam music scores}, author = {Baris Bozkurt and Kemal M Karaosmanoglu and Bilge Karacali and Erdem Unal}, doi = {10.1109/siu.2014.6830262}, year = {2014}, date = {2014-01-01}, abstract = {Automatic melodic segmentation is one of the important steps in computational analysis of melodic content from symbolic data. This widely studied research problem has been very rarely considered for Turkish makam music. In this paper we first present test results for state-of-the-art techniques from literature on Turkish makam music data. Then, we present a statistical classification-based segmentation system that exploits the link between makam melodies and usul and makam scale hierarchies together with the well-known features in literature. We show through tests on a large dataset that the proposed system has a higher accuracy. textcopyright 2014 IEEE.}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } Automatic melodic segmentation is one of the important steps in computational analysis of melodic content from symbolic data. This widely studied research problem has been very rarely considered for Turkish makam music. In this paper we first present test results for state-of-the-art techniques from literature on Turkish makam music data. Then, we present a statistical classification-based segmentation system that exploits the link between makam melodies and usul and makam scale hierarchies together with the well-known features in literature. We show through tests on a large dataset that the proposed system has a higher accuracy. textcopyright 2014 IEEE. |
Bozkurt, Baris; Karaosmanoglu, Kemal M; Karacali, Bilge; Unal, Erdem Türk makam müziǧi notalari icin otomatik ezgi bölütleme Inproceedings 2014 22nd Signal Processing and Communications Applications Conference, SIU 2014 - Proceedings, 2014, ISBN: 9781479948741. @inproceedings{Bozkurt2014b, title = {Türk makam müziǧi notalari icin otomatik ezgi bölütleme}, author = {Baris Bozkurt and Kemal M Karaosmanoglu and Bilge Karacali and Erdem Unal}, doi = {10.1109/SIU.2014.6830262}, isbn = {9781479948741}, year = {2014}, date = {2014-01-01}, booktitle = {2014 22nd Signal Processing and Communications Applications Conference, SIU 2014 - Proceedings}, abstract = {Automatic melodic segmentation is one of the important steps in computational analysis of melodic content from symbolic data. This widely studied research problem has been very rarely considered for Turkish makam music. In this paper we first present test results for state-of-the-art techniques from literature on Turkish makam music data. Then, we present a statistical classification-based segmentation system that exploits the link between makam melodies and usul and makam scale hierarchies together with the well-known features in literature. We show through tests on a large dataset that the proposed system has a higher accuracy.}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } Automatic melodic segmentation is one of the important steps in computational analysis of melodic content from symbolic data. This widely studied research problem has been very rarely considered for Turkish makam music. In this paper we first present test results for state-of-the-art techniques from literature on Turkish makam music data. Then, we present a statistical classification-based segmentation system that exploits the link between makam melodies and usul and makam scale hierarchies together with the well-known features in literature. We show through tests on a large dataset that the proposed system has a higher accuracy. |
Bozkurt, Barış; Karaosmanoğlu, Kemal M; Karacalı, Bilge; Ünal, Erdem Türk makam müziği notaları icin otomatik ezgi bölütleme Inproceedings IEEE 22. Sinyal İşleme ve İletişim Uygulamaları Kurultayı (SIU), 2014, ISBN: 9781479948741. @inproceedings{Bozkurt2014c, title = {Türk makam müziği notaları icin otomatik ezgi bölütleme}, author = {Barış Bozkurt and Kemal M Karaosmanoğlu and Bilge Karacalı and Erdem Ünal}, isbn = {9781479948741}, year = {2014}, date = {2014-01-01}, booktitle = {IEEE 22. Sinyal İşleme ve İletişim Uygulamaları Kurultayı (SIU)}, abstract = {Automatic melodic segmentation is one of the important steps in computational analysis of melodic content from symbolic data. This widely studied research problem has been very rarely considered for Turkish makam music. In this paper we first present test results for state-of-the-art techniques from literature on Turkish makam music data. Then, we present a statistical classification-based segmentation system that exploits the link between makam melodies and usul and makam scale hierarchies together with the well-known features in literature. We show through tests on a large dataset that the proposed system has a higher accuracy.}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } Automatic melodic segmentation is one of the important steps in computational analysis of melodic content from symbolic data. This widely studied research problem has been very rarely considered for Turkish makam music. In this paper we first present test results for state-of-the-art techniques from literature on Turkish makam music data. Then, we present a statistical classification-based segmentation system that exploits the link between makam melodies and usul and makam scale hierarchies together with the well-known features in literature. We show through tests on a large dataset that the proposed system has a higher accuracy. |
Doğan, Tunca; Karacalı, Bilge 2-D Thresholding of the Connectivity Map Following the Multiple Sequence Alignments of Diverse Datasets Inproceedings 2013. @inproceedings{Dogan2013a, title = {2-D Thresholding of the Connectivity Map Following the Multiple Sequence Alignments of Diverse Datasets}, author = {Tunca Doğan and Bilge Karacalı}, doi = {10.2316/p.2013.791-092}, year = {2013}, date = {2013-01-01}, abstract = {Multiple sequence alignment (MSA) is a widely used method to uncover the relationships between the biomolecular sequences. One essential prerequisite to apply this procedure is to have a considerable amount of similarity between the test sequences. It's usually not possible to obtain reliable results from the multiple alignments of large and diverse datasets. Here we propose a method to obtain sequence clusters of significant intragroup similarities and make sense out of the multiple alignments containing remote sequences. This is achieved by thresholding the pairwise connectivity map over 2 parameters. The first one is the inferred pairwise evolutionary distances and the second parameter is the number of gapless positions on the pairwise comparisons of the alignment. Threshold curves are generated regarding the statistical parameter values obtained from a shuffled dataset and probability distribution techniques are employed to select an optimum threshold curve that eliminate as much of the unreliable connectivities while keeping the reliable ones. We applied the method on a large and diverse dataset composed of nearly 18000 human proteins and measured the biological relevance of the recovered connectivities. Our precision measure (0.981) was nearly 20% higher than the one for the connectivities left after a classical thresholding procedure displaying a significant improvement. Finally we employed the method for the functional clustering of protein sequences in a gold standard dataset. We have also measured the performance, obtaining a higher F-measure (0.882) compared to a conventional clustering operation (0.827).}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } Multiple sequence alignment (MSA) is a widely used method to uncover the relationships between the biomolecular sequences. One essential prerequisite to apply this procedure is to have a considerable amount of similarity between the test sequences. It's usually not possible to obtain reliable results from the multiple alignments of large and diverse datasets. Here we propose a method to obtain sequence clusters of significant intragroup similarities and make sense out of the multiple alignments containing remote sequences. This is achieved by thresholding the pairwise connectivity map over 2 parameters. The first one is the inferred pairwise evolutionary distances and the second parameter is the number of gapless positions on the pairwise comparisons of the alignment. Threshold curves are generated regarding the statistical parameter values obtained from a shuffled dataset and probability distribution techniques are employed to select an optimum threshold curve that eliminate as much of the unreliable connectivities while keeping the reliable ones. We applied the method on a large and diverse dataset composed of nearly 18000 human proteins and measured the biological relevance of the recovered connectivities. Our precision measure (0.981) was nearly 20% higher than the one for the connectivities left after a classical thresholding procedure displaying a significant improvement. Finally we employed the method for the functional clustering of protein sequences in a gold standard dataset. We have also measured the performance, obtaining a higher F-measure (0.882) compared to a conventional clustering operation (0.827). |
Kokturk, Basak Esin; Karacali, Bilge Automated labeling of electroencephalography data using quasi-supervised learning Inproceedings 2012. @inproceedings{Kokturk2012, title = {Automated labeling of electroencephalography data using quasi-supervised learning}, author = {Basak Esin Kokturk and Bilge Karacali}, doi = {10.1109/siu.2012.6204600}, year = {2012}, date = {2012-01-01}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } |
Karacalı, Bilge Hierarchical Motif Vectors for Amino Acid Sequence Alignment Inproceedings 2012. @inproceedings{Karacal2012, title = {Hierarchical Motif Vectors for Amino Acid Sequence Alignment}, author = {Bilge Karacalı}, doi = {10.2316/p.2012.764-055}, year = {2012}, date = {2012-01-01}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } |
Karacali, Bilge Analytical and predictive quasi-supervised learning for cancer recognition in digital cytology Inproceedings 2012. @inproceedings{Karacali2012a, title = {Analytical and predictive quasi-supervised learning for cancer recognition in digital cytology}, author = {Bilge Karacali}, doi = {10.1109/siu.2012.6204467}, year = {2012}, date = {2012-01-01}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } |
Karacali, Bilge Identification and evaluation of landmarks for deformable alignment of multi-modality medical images Inproceedings 2011. @inproceedings{Karacali2011, title = {Identification and evaluation of landmarks for deformable alignment of multi-modality medical images}, author = {Bilge Karacali}, doi = {10.1109/siu.2011.5929611}, year = {2011}, date = {2011-01-01}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } |
Onder, Devrim; Sarioglu, Sulen; Karacali, Bilge Automated classification of cancerous textures in histology images using quasi-supervised learning algorithm Inproceedings 2010. @inproceedings{Onder2010, title = {Automated classification of cancerous textures in histology images using quasi-supervised learning algorithm}, author = {Devrim Onder and Sulen Sarioglu and Bilge Karacali}, doi = {10.1109/biyomut.2010.5479863}, year = {2010}, date = {2010-01-01}, abstract = {The aim of this work is to perform automated texture classification of histology slide images in health and cancerous conditions using quasi-supervised statistical learning method. Tissue images were acquired from histological slides of human colon and were separated into two groups in terms of normal and disease conditions. Texture feature vectors corresponding to tissue segments of each image were calculated using co-occurrence matrices. Different texture regions were determined by the quasi-supervised statistical learning method using texture features of normal and cancerous groups.}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } The aim of this work is to perform automated texture classification of histology slide images in health and cancerous conditions using quasi-supervised statistical learning method. Tissue images were acquired from histological slides of human colon and were separated into two groups in terms of normal and disease conditions. Texture feature vectors corresponding to tissue segments of each image were calculated using co-occurrence matrices. Different texture regions were determined by the quasi-supervised statistical learning method using texture features of normal and cancerous groups. |
Karacali, Bilge Deformation field interpolation using rotational landmarks Inproceedings 2010. @inproceedings{Karacali2010a, title = {Deformation field interpolation using rotational landmarks}, author = {Bilge Karacali}, doi = {10.1109/biyomut.2010.5479857}, year = {2010}, date = {2010-01-01}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } |
Onder, Devrim; Karacali, Bilge Automated clustering of histology slide texture using co-occurrence based grayscale image features and manifold learning Inproceedings 2009. @inproceedings{Onder2009, title = {Automated clustering of histology slide texture using co-occurrence based grayscale image features and manifold learning}, author = {Devrim Onder and Bilge Karacali}, doi = {10.1109/biyomut.2009.5130342}, year = {2009}, date = {2009-01-01}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } |
Makrogiannis, Sokratis; Verma, Ragini; Karacali, Bilge; Davatzikos, Christos A joint transformation and residual image descriptor for morphometric image analysis using an equivalence class formulation Inproceedings Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2006, ISSN: 10636919. @inproceedings{Makrogiannis2006, title = {A joint transformation and residual image descriptor for morphometric image analysis using an equivalence class formulation}, author = {Sokratis Makrogiannis and Ragini Verma and Bilge Karacali and Christos Davatzikos}, doi = {10.1109/CVPRW.2006.17}, issn = {10636919}, year = {2006}, date = {2006-01-01}, booktitle = {Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition}, abstract = {Existing computational anatomy methodologies for morphometric analysis of medical images are often based solely on the shape transformation, typically being a diffeomorphism, that warps these images to a common template or vice versa. However, anatomical differences as well as changes induced by pathology, prevent the warping transformation from producing an exact correspondence. The residual image captures information that is not reflected by the diffeomorphism, and therefore allows us to maintain the entire morphological profile for analysis. In this paper we present a morphological descriptor which combines the warping transformation with the residual image in an equivalence class formulation, to characterize morphology of anatomical structures. Equivalence classes are formed by pairs of transformation and residual, for different levels of smoothness of the warping transformation. These pairs belong to the same equivalence class, since they jointly reconstruct the exact same morphology. Moreover, pattern classification methods are trained on the entire equivalence class, instead of a single pair, in order to become more robust to a variety of factors that affect the warping transformation, including the anatomy being measured. This joint descriptor is evaluated by statistical testing and estimation of class separation by classification, initially for 2-D synthetic images with simulated atrophy and subsequently for a volumetric dataset consisting of schizophrenia patients and healthy controls. Results of class separation indicate that this joint descriptor produces generally better and more robust class separation than using each of the components separately.}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } Existing computational anatomy methodologies for morphometric analysis of medical images are often based solely on the shape transformation, typically being a diffeomorphism, that warps these images to a common template or vice versa. However, anatomical differences as well as changes induced by pathology, prevent the warping transformation from producing an exact correspondence. The residual image captures information that is not reflected by the diffeomorphism, and therefore allows us to maintain the entire morphological profile for analysis. In this paper we present a morphological descriptor which combines the warping transformation with the residual image in an equivalence class formulation, to characterize morphology of anatomical structures. Equivalence classes are formed by pairs of transformation and residual, for different levels of smoothness of the warping transformation. These pairs belong to the same equivalence class, since they jointly reconstruct the exact same morphology. Moreover, pattern classification methods are trained on the entire equivalence class, instead of a single pair, in order to become more robust to a variety of factors that affect the warping transformation, including the anatomy being measured. This joint descriptor is evaluated by statistical testing and estimation of class separation by classification, initially for 2-D synthetic images with simulated atrophy and subsequently for a volumetric dataset consisting of schizophrenia patients and healthy controls. Results of class separation indicate that this joint descriptor produces generally better and more robust class separation than using each of the components separately. |
Xue, Zhong; Shen, Dinggang; Karacali, Bilge; Davatzikos, Christos Statistical representation and simulation of high-dimensional deformations: Application to synthesizing brain deformations Inproceedings Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 2005, ISSN: 03029743. @inproceedings{Xue2005, title = {Statistical representation and simulation of high-dimensional deformations: Application to synthesizing brain deformations}, author = {Zhong Xue and Dinggang Shen and Bilge Karacali and Christos Davatzikos}, doi = {10.1007/11566489_62}, issn = {03029743}, year = {2005}, date = {2005-01-01}, booktitle = {Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)}, abstract = {This paper proposes an approach to effectively representing the statistics of high-dimensional deformations, when relatively few training samples are available, and conventional methods, like PCA, fail due to insufficient training. Based on previous work on scale-space decomposition of deformation fields, herein we represent the space of "valid deformations" as the intersection of three subspaces: one that satisfies constraints on deformations themselves, one that satisfies constraints on Jacobian determinants of deformations, and one that represents smooth deformations via a Markov Random Field (MRF). The first two are extensions of PCA-based statistical shape models. They are based on a wavelet packet basis decomposition that allows for more accurate estimation of the covariance structure of deformation or Jacobian fields, and they are used jointly due to their complementary strengths and limitations. The third is a nested MRF regularization aiming at eliminating potential discontinuities introduced by assumptions in the statistical models. A randomly sampled deformation field is projected onto the space of valid deformations via iterative projections on each of these subspaces until convergence, i.e. all three constraints are met. A deformation field simulator uses this process to generate random samples of deformation fields that are not only realistic but also representative of the full range of anatomical variability. These simulated deformations can be used for validation of deformable registration methods. Other potential uses of this approach include representation of shape priors in statistical shape models as well as various estimation and hypothesis testing paradigms in the general fields of computational anatomy and pattern recognition.}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } This paper proposes an approach to effectively representing the statistics of high-dimensional deformations, when relatively few training samples are available, and conventional methods, like PCA, fail due to insufficient training. Based on previous work on scale-space decomposition of deformation fields, herein we represent the space of "valid deformations" as the intersection of three subspaces: one that satisfies constraints on deformations themselves, one that satisfies constraints on Jacobian determinants of deformations, and one that represents smooth deformations via a Markov Random Field (MRF). The first two are extensions of PCA-based statistical shape models. They are based on a wavelet packet basis decomposition that allows for more accurate estimation of the covariance structure of deformation or Jacobian fields, and they are used jointly due to their complementary strengths and limitations. The third is a nested MRF regularization aiming at eliminating potential discontinuities introduced by assumptions in the statistical models. A randomly sampled deformation field is projected onto the space of valid deformations via iterative projections on each of these subspaces until convergence, i.e. all three constraints are met. A deformation field simulator uses this process to generate random samples of deformation fields that are not only realistic but also representative of the full range of anatomical variability. These simulated deformations can be used for validation of deformable registration methods. Other potential uses of this approach include representation of shape priors in statistical shape models as well as various estimation and hypothesis testing paradigms in the general fields of computational anatomy and pattern recognition. |
Hamza, Ben A; Krim, Hamid; Karacali, Bilge Structural risk minimization using nearest neighbor rule Inproceedings Proceedings - IEEE International Conference on Multimedia and Expo, 2003, ISSN: 1945788X. @inproceedings{Hamza2003, title = {Structural risk minimization using nearest neighbor rule}, author = {Ben A Hamza and Hamid Krim and Bilge Karacali}, doi = {10.1109/ICME.2003.1221046}, issn = {1945788X}, year = {2003}, date = {2003-01-01}, booktitle = {Proceedings - IEEE International Conference on Multimedia and Expo}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } |
Karacali, Bilge; Krim, Hamid A; Schick, Irvin C textlesstitletextgreaterWavelet-based methods in Global Positioning System signal trackingtextless/titletextgreater Inproceedings Wavelet Applications VII, 2003. @inproceedings{Karacali2003b, title = {textlesstitletextgreaterWavelet-based methods in Global Positioning System signal trackingtextless/titletextgreater}, author = {Bilge Karacali and Hamid A Krim and Irvin C Schick}, doi = {10.1117/12.381675}, year = {2003}, date = {2003-01-01}, booktitle = {Wavelet Applications VII}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } |
Karacali, Bilge; Snyder, Wesley Automatic target detection using multispectral imaging Inproceedings Proceedings - Applied Imagery Pattern Recognition Workshop, 2002, ISSN: 21642516. @inproceedings{Karacali2002, title = {Automatic target detection using multispectral imaging}, author = {Bilge Karacali and Wesley Snyder}, doi = {10.1109/AIPR.2002.1182255}, issn = {21642516}, year = {2002}, date = {2002-01-01}, booktitle = {Proceedings - Applied Imagery Pattern Recognition Workshop}, abstract = {We propose using multispectral imaging for on-the-fly target detection and classification instead of hyperspectral imaging. We initially pose the target detection problem as a classification problem with classes identified as target and clutter. The classification data consists of multispectral observations of the region of interest, focusing on visual and infrared wavelengths. We then solve this classification problem using nearest neighbor rule, support vector machines, and maximum likelihood classification. Simulation results on real data indicate that information from a multispectral sensor can offer better performance than both single band and hyperspectral sensors, also showing that costly hyperspectral analysis performance can be attained onboard a small airborne platform such as a smart missile using cost-effective multispectral sensors.}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } We propose using multispectral imaging for on-the-fly target detection and classification instead of hyperspectral imaging. We initially pose the target detection problem as a classification problem with classes identified as target and clutter. The classification data consists of multispectral observations of the region of interest, focusing on visual and infrared wavelengths. We then solve this classification problem using nearest neighbor rule, support vector machines, and maximum likelihood classification. Simulation results on real data indicate that information from a multispectral sensor can offer better performance than both single band and hyperspectral sensors, also showing that costly hyperspectral analysis performance can be attained onboard a small airborne platform such as a smart missile using cost-effective multispectral sensors. |