LABORATORY OF COMPUTATIONAL IMAGING AND BIOINFORMATICS Annual Report: 2011 Director: Dr. Anant Madabhushi Associate Professor, Department of Biomedical Engineering Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 LCIB Website @ http://lcib.rutgers.edu Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 LAB MEMBERS April 2011 Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 LAB MEMBERS Lab Director: Anant Madabhushi, PhD Administrative Assistant: Rhonda Breen-Simone Postdocs: • Mirabela Rusu, PhD • Tao Wan, PhD Graduate Students • Sahir Ali • Ajay Basavanhally • Andrew Janowczyk (IIT, Bombay) • George Lee • Shoshana Rosskamm • Rachel Sparks • Pallavi Tiwari • Rob Toth • Satish Viswanath Research Faculty • James Monaco, PhD Undergraduate Students • Ronak Chawla • Joe Galero • Abhishek Golugula • Aparma Kannan • Sudha Karthigeyan • Ross Kleiman • Prateeka Koul • Eileen Hwuang • Pratik Patel • Srivathsan Prabu • Elaine Yu Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 LAB MEMBERS Assistant Faculty Lab Director: Anant Madabhushi, PhD Research Faculty James Monaco, PhD Rhonda Breen-Simone PostDocs Mirabela Rusu, PhD Tao Wan, PhD Graduate Students Sahir Ali Ajay Basavanhally Andrew Janowczyk (IIT, Bombay) George Lee Shoshana Rosskamm Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 LAB MEMBERS Graduate Students (Contd) Rachel Sparks Pallavi Tiwari Rob Toth Satish Viswanath Undergraduate Students • Ronak Chawla • Aparma Kannan • Sudha Karthigeyan • Ross Kleiman • Prateeka Koul • Eileen Hwuang • Srivathsan Prabu Joe Galero Elaine Yu Pratik Patel Abhishek Golugula Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 RECENT ALUMNI PostDocs Gaoyu Xiao, PhD Jun Xu, PhD Professor at Nanjing University Master Students PhD Students Shannon Agner Jon Chappelow Research 3rd Year Medical Scientist at Student at UMDNJ-RWJMS Accuray, Inc. Scott Doyle Director of Research at Ibris, Inc. Akshay Shridar Assistant Project Manager at Integra Life Sciences Najeeb Chowdhury Healthcare Market Research Analyst at AlphaDetail Inc. 7 Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 SOME OF OUR COLLABORATORS John E .Tomaszewski Mark Rosen, Michael D. Feldman, Shridar Ganesan 8 Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 CONFERENCE PARTICIPATION - 2011 SPIE, Finalist Student Paper Award: Shannon Agner MICCAI: Sahir Ali, Satish Viswanath, Ajay Basavanhally 9 Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 SUMMARY OF ACCOMPLISHMENTS 2011 Lab Members: 25 (2 faculty, 2 post-docs, 9 graduate, 1 admin, 11 undergrads) Papers Books: 1 Theses: 3 PhD+2 MS Journal Papers: 24 Peer-Reviewed Conference Papers: 31 Abstracts: 18 5 new PCT Patents Filed 5 new grants 1 NIH 1 DOD On going projects: 45 10 Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 AWARDS AND ACCOMPLISHMENTS Editorial, Committees, Membership - Anant Madabhushi Editor (Guest), Special Issue on IEEE Transactions on Biomedical Engineering Letters: Multi-Scale/Resolution Signal and Image Analysis, 2011. Co-Chair, Session on “Image Analysis in Cancer Detection”, IEEE Engineering in Medicine and Biology Conference 2011, Boston, MA, September 1, 2011. Chair, Workshop on Prostate Cancer Imaging, MICCAI 2011. Chair, Workshop on Histopathology Image Analysis, MICCAI 2011. Co-Organizer, Tutorial on Manifold Learning for Medical Images, MICCAI 2011. Co-Chair, Special Session on Prostate Cancer Imaging, IEEE International Symposium on Biomedical Imaging, March 2011. Program Committee Member, Computer-Aided Diagnosis conference at SPIE Medical Imaging, 2011. Program Committee Member, MICCAI 2011. Panelist and Invited Speaker, “Preparing for Campus Interviews & Negotiating Academic Job Offers”, University of Pennsylvania, January, 2011. Program Committee on IEEE Conference on Bioinformatics and Bioengineering (BIBE), 2011. Technical Advisory Board Member, IEEE EMBS/ISBI, 2010-Present Co-Chair, Workshop on Trends in Experimental Pathology: Imaging Organisms/Experimental Trends – “Brave New World”, In conjunction with 2011 Annual Meeting at Experimental Biology, American Society for Investigative Pathology, April 9-13, 2011, Washington, DC. Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 AWARDS AND ACCOMPLISHMENTS Awards - Anant Madabhushi First Prize, SNR Denton Elevator Pitch Competition, Rutgers Entrepreneurship Day, 2011 Media Recognition “Entrepreneurs flock to Rutgers Entrepreneurship Conference”, New Jersey Tech Weekly, November 22nd, 2011. “University Startup receives NIH SBIR Funding”, Daily Targum, October 11, 2011. “Rutgers-Affiliated Start-Up Company Receives Funding for Technology to Help Choose Breast Cancer Treatments”, Rutgers Today, October 6th, 2011. Story featured on the following news sites AAAS EurekAlert! Noodls.com NJBIZ.com www.njcrea.com www.biomedicalproducts.com www.bioprodmag.com www.physorg.com www.news-medical.net www.biosciencetechnology.com www.biosciencetechnology.com www.bionj.org “The Race for Another Cure”, Rutgers Magazine, Pg. 31, Winter 2011 Issue. Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 NEW GRANTS - 2011 Madabhushi, Anant (PI) Date: 07/01/11-12/31/11 Tech Commercialization Fund, ORSP, Rutgers Image based risk score predictors for ER+ Breast Cancers Role: PI Madabhushi, Anant (MPI) Date: 09/03/11-08/31/12 NIH R43EB015199-01 Decision Support System for Predicting outcome of ER+ breast cancers Role: PI with Shridar Ganesan (CINJ), David Harding (IbRiS Inc.) Madabhushi, Anant (PI) Date: 07/01/11- 06/30/12 Pre-doctoral Training Fellowship Society for Imaging Informatics in Medicine Computerized Decision Support for Prostate Cancer Detection from Multi-parametric MRI Role: Mentor for Satish Viswanath Madabhushi, Anant (PI) Date: 01/01/12 -12/31/13 Pre-doctoral Training Fellowship in Health Disparity Department of Defense Comparing Proteomic, Histological Biomarkers for Biochemical Failure Among African Americans and Caucasians Following Radical Prostatectomy Role: Mentor for George Lee Monaco, James (PI) Date: 06/09/11 - 07/15/11 Burroughs Wellcome Fund Pathology Training and Collaboration between Rutgers and University of Michigan Role: Collaborator Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 STUDENTS AWARDS AND FELLOWSHIPS Pallavi Tiwari, Travel Award ($250) from SciWomen, The Office for the promotion of Women in Science, Engineering and Mathematics (STEM), 2011 Sahirzeeshan Ali, Travel Award ($600) from Omnyx to attend Workshop on Histopathology Image Analysis, Toronto, Canada, 2011 Ajay Basavanhally, Travel Award ($600) from Omnyx to attend Workshop on Histopathology Image Analysis, Toronto, Canada, 2011 Joseph Galero, Elaine Yu, BMEStart Competition, Honorable Mention, NCIAA, 2011. Shannon Agner, “Textural Kinetics: A Novel Dynamic Contrast-Enhanced (DCE)-MRI Feature for Breast Lesion Classification” nominated for best paper of 2010 by the Journal of Digital Imaging editorial board. Joe Galero, Best Poster Award for “Integrated Texton and Bag of Words Classifier for Identifying Anaplastic Medulloblastomas" at BME Senior Design Conference 2011 Elaine Yu, Admitted to Pittsburgh Tissue Engineering Initiative Summer Internship program with a research scholarship ($3,850), April 2011. Shannon Agner, Outstanding Graduate Student Award, School of Engineering, April 15, 2011. Pratik Patel, Selected as an intern to (RISE) Research in Science and Engineering Program ($2500), Germany, 2011. Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 STUDENTS AWARDS AND FELLOWSHIPS (CONTD) Ajay Basavanhally, Travel Award ($300) from IEEE Intl Symposium on Biomedical Imaging, 2011 Shannon Agner, Departmental Travel Award ($200), 2011 Satish Viswanath, Departmental Travel Award ($100), 2011 Akshay Shridar, Departmental Travel Award ($100), 2011 Ajay Basavanhally, Departmental Travel Award ($100), 2011 Scott Doyle, Departmental Travel Award ($100), 2011 George Lee, Departmental Travel Award ($100), 2011 Satish Viswanath, Interview featured in NJN TV Channel, March 12th, 2011. Satish Viswanath, Interview featured in CP News Story, “Study Finds most Indian graduate students in the US want to use their degrees back home”, 2011. Sudha Karthigeyan, Selected as an intern to (UROP) Undergraduate Research Opportunities Program ($2500), 2011. Shannon Agner, Michael B. Merickel Best Student Paper Award Finalist, SPIE Medical Imaging, 2011. Shannon Agner, Rachel Sparks, BMEIdea Competition Finalist, NCIAA ($500), 2011. Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 PEER REVIEWED PUBLICATIONS FOR 2011 Summary 35 30 25 20 15 10 5 0 Theses Journal Papers Conference Papers Abstracts 16 Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 BOOKS Madabhushi, A, Dowling, J, Huisman, H, Barratt, D, Prostate Cancer Imaging: Image analysis and image guided interventions, International Workshop Held in Conjunction with MICCAI 2011, Proceedings Springer Verlag ISBN-10: 3642239439, 2011 17 Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 THESES PhD Shannon Agner, A Computerized Image Analysis Framework for Dynamic Contrast Enhanced Magnetic Resonance Imaging (DCE-MRI): Applications to Breast Cancer (04/2011) Jon Chappelow , Multimodal Image Registration using Multivariate Information Theoretic Similarity Measures: Applications to Prostate Cancer Diagnosis and Targeted Treatment (04/2011) Scott Doyle, Computerized Detection, Segmentation and Classification of Digital Pathology: Case Study in Prostate Cancer (04/2011) MS Najeeb Chowdhury, Concurrent segmentation of the prostate on MRI and CT via linked statistical shape models for radiotherapy planning (07/2011) Akshay Shridar, Content-Based Image Retrieval of Digitized Histopathology via Boosted Spectral Embedding (BoSE) (10/2011) 18 Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 PEER REVIEWED PUBLICATIONS JOURNAL PAPERS Hipp, J, Monaco, J, Kunju, P Cheng, P, Yagi, Y, Rodriguez-Canales, J, Emmert-Buck, M, Hewitt, S, Feldman, M, Tomaszewski, J, Shih, N, Toner, M, Tompkins, R, Flotte, T, Lucas, D, Gilbertson, Kunju, LP, J, Balis, U, Madabhushi, A, “Integration of architectural and cytologic driven image algorithms for prostate adenocarcinoma identification”, Analytical Cellular Pathology, Accepted. Chowdhury, N, Toth, R, Chappelow, J, Kim, S, Motwani, S, Punekar, S, Lin, H, Both, S, Vapiwala, N, Hahn, S, Madabhushi, A, “Concurrent segmentation of the prostate on MRI and CT via linked statistical shape models for radiotherapy planning”, Medical Physics, Accepted. Golugula, A, Lee, G, Master, S, Feldman, M, Tomaszewski, J, Speicher, D, Madabhushi, A, “Supervised Regularized Canonical Correlation Analysis: Integrating Histologic and Proteomic measurements for predicting Biochemical Failures following Prostate Surgery”, BMC Bioinformatics, Accepted. Viswanath, S, Bloch, B, Chappelow, J, Toth, R, Rofsky, N, Genega, E, Lenkinski, R, Madabhushi, A, “Central Gland and Peripheral Zone Prostate Tumors have Significantly Different Quantitative Imaging Signatures on 3 Tesla Endorectal, In Vivo T2-Weighted Magnetic Resonance Imagery”, Journal of Magnetic Resonance Imaging, Accepted. Janowczyk, A, Chandran, S, Singh, R, Sasaroli, D, Coukos, G, Feldman, M, Madabhushi, A, “HighThroughput Biomarker Segmentation on Ovarian Cancer Tissue Microarrays via Hierarchical Normalized Cuts”, IEEE Trans on Biomedical Engineering, In Press. Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 PEER REVIEWED PUBLICATIONS JOURNAL PAPERS (CONTD) Hipp, J, Monaco, J, Kunju, P Cheng, P, Yagi, Y, Rodriguez-Canales, J, Emmert-Buck, M, Hewitt, S, Feldman, M, Tomaszewski, J, Shih, N, Toner, M, Tompkins, R, Flotte, T, Lucas, D, Gilbertson, Madabhushi, A, Kunju, LP, J, Balis, U, “Optimization of complex cancer morphology detection using the SIVQ pattern recognition algorithm”, Analytical Cellular Pathology, 2011 Oct 11 [Epub ahead of print] (PMID: 21988838) Hipp, J, Cheng, J, Pantanowitz, L, Hewitt, S, Yagi, Y, Monaco, J, Madabhushi, A, Rodriguez-canales, J, Hanson, J, Roy-Chowdhuri, S, Filie, A, Feldman, M, Tomaszewski, J, Shih, N, Gilbertson, J, EmmertBuck, M, Balis, U, “Image Microarrays (IMA) Digital Pathology's Missing Tool”, Journal of Pathology Informatics, Accepted Ali, S, Madabhushi, A, "GPU Implementation of an Integrated Shape Based Active Contour: Application to Digital Pathology", Journal of Pathology Informatics, Accepted. Basavanhally, A, Ganesan, S, Shih, N, Feldman, M, Tomaszewski, J, Madabhushi, A, “Multi-Field-ofView Strategy for Image-Based Outcome Prediction of ER+ Breast Cancer Histopathology Using Spatio-Architectural and Vascular Features”, Journal of Pathology Informatics, Accepted. Ghaznavi, D, Evans, A, Madabhushi, A, Feldman, M, “Digital Imaging in Pathology: Whole-Slide Imaging and Beyond”, Annual Review of Pathology: Mechanisms of Disease, Accepted. Doyle, S, Monaco, J, Tomaszewski, J, Feldman, M, Madabhushi, A, "An Active Learning Based Classification Strategy for the Minority Class Problem: Application to Histopathology Annotation", BMC Bioinformatics, 12:424, 2011. Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 PEER REVIEWED PUBLICATIONS JOURNAL PAPERS (CONTD) Hipp, J, Monaco, J, Cheng, J, Lucas, D, Madabhushi, A, Balis, U, Automated Vector Selection of SIVQ and Parallel Computing Integration MATLABTM: Innovations supporting large-scale and highthroughput image analysis studies, Journal of Pathology Informatics, 2:37, 2011 (PMID: 21886893). Tiwari, P, Viswanath, S, Kurhanewicz, J, Shridar, A, Madabhushi, A, “Multimodal Wavelet Embedding Representation for data Combination (MaWERiC): Integrating Magnetic Resonance Imaging and Spectroscopy for Prostate Cancer Detection”, NMR in Biomedicine, 2011 doi: 10.1002/nbm.1777, (PMID: 21960175). Hipp, J, Sica, J, McKenna, B, Monaco, J, Madabhushi, A, Cheng, J, Balis, U, The need for the pathology community to sponsor a WSI repository with technical guidance from the pathology informatics community, Journal of Pathology Informatics, 2:31, 2011 (PMID: 21845229). Bulman, J, Toth, R, Patel, AD, Bloch, BN, McMahon, CJ, Ngo, L, Madabhushi, A, Rofsky, N, Automated Computer-Derived Prostate Volumes from MRI Data: Comparison to RadiologistDerived MRI Volumes and Pathology Specimen Volumes, Radiology, Accepted. Hipp, J, Flotte, T, Monaco, J, Cheng, J, Madabhushi, A, Yagi, Y, Rodriguez-Canales, J, Emmert-Buck, M, Dugan, MC, Hewitt, S, Toner, M, Tompkins, R, Lucas, D, Gilbertson, JR, Balis, U, Computer Aided Diagnostic (CAD) tools aim to empower rather than replace Pathologists: Lessons learned from computational chess, Journal of Pathology Informatics, 2:25, 2011 (PMID: 21773056). Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 PEER REVIEWED PUBLICATIONS JOURNAL PAPERS (CONTD) Xu, J, Janowcyzk, A., Tomaszewski, J., Feldman, M., and Madabhushi, A, “A High-throughput Active Contour Scheme for Segmentation of Histopathological Imagery”, Medical Image Analysis, vol. 15, pp. 851-62, 2011 (PMID: 21570336). Chappelow, J, Bloch, N., Rofsky, N, Genega, E, Lenkinski, R, DeWolf, W, Madabhushi, A, Elastic Registration of Multimodal Prostate MRI and Histology via Multi-Attribute Combined Mutual Information, Medical Physics, vol. 38[4], pp. 2005-2018, 2011 (PMID: 21626933). Toth, R, Bloch, N, Genega, E, Rofsky, N, Lenkinsky, R, Rosen, M, Kalyanpur, A, Pungavkar, S, Madabhushi, A, Accurate Prostate Volume Estimation Using Multi-Feature Active Shape Models on T2-Weighted MRI, Academic Radiology, vol. 18[6], pp. 745-54, 2011 (PMID: 21549962). Chappelow, J, Feldman, M, Tomaszewski, J, Shih, N, Madabhushi, A, HistoStitcher©: An Interactive Program for Accurate and Rapid Reconstruction of Digitized Whole Histological Sections from Tissue Fragments, Computerized Medical Imaging and Graphics, vol. 35[7-8], pp. 542-56, 2011 (PMID: 21397459). Monaco, J, Madabhushi, A, “Weighted Maximum Posterior Marginals for Random Fields Using an Ensemble of Conditional Densities from Multiple Markov Chain Monte Carlo Simulations”, IEEE Transactions on Medical Imaging, vol. 30[7], pp. 1353-64, 2011 (PMID: 21335309). Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 PEER REVIEWED PUBLICATIONS JOURNAL PAPERS (CONTD) Madabhushi, A, Agner, S, Doyle, S, Basavanhally, A, Lee, G, Computer-Aided Prognosis: Predicting patient and disease outcome via quantitative fusion of multi-scale, multi-modal data, Special Issue of Computerized Medical Imaging and Graphics on Whole Slide Microscopic Image Processing, vol. 35[7-8], pp. 506-14, 2011 (PMID: 21333490). Xiao, G, Bloch, N, Chappelow, J, Genega, E, Rofsky, N, Lenkinsky, R, Tomaszewski, J, Feldman, M, Rosen, M, Madabhushi, A, Determining Histology-MRI Slice Correspondences for Defining MRIbased Disease Signatures of Prostate Cancer, Special Issue of Computerized Medical Imaging and Graphics on Whole Slide Microscopic Image Processing, vol. 35[7-8], pp. 568-78, 2011 (PMID: 21255974). Toth, R, Tiwari, P, Rosen, M, Reed, G, Kurhanewicz, J, Kalyanpur, A, Pungavkar, S, Madabhushi, A, A Magnetic Resonance Spectroscopy Driven Initialization Scheme for Active Shape Model Based Prostate Segmentation, Medical Image Analysis, vol. 15(2), pp. 214-225, 2011 (PMID: 21195016). Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 PEER REVIEWED PUBLICATIONS PEER-REVIEWED CONFERENCE PROCEEDINGS Tiwari, P, Viswanath, S, Kurhanewicz, J, Madabhushi, A, Weighted Combination of Multi-parametric MR Imaging markers for Evaluating Radiation Therapy changes in the prostate, Workshop on Prostate Cancer Imaging, In Conjunction with MICCAI 2011, pp. 80-91. Ginsburg, S, Tiwari, P, Kurhanewicz, J, Madabhushi, A, Variable Ranking with PCA: Finding Multiparametric MR Imaging Biomarkers for Prostate Cancer Diagnosis and Grading, Workshop on Prostate Cancer Imaging, In Conjunction with MICCAI 2011, pp. 146-157. Palumbo, D, Yee, B, Leedy, S, O’Dea, P, Viswanath, S, Madabhushi, A, Interplay between Bias Field Correction, Intensity Standardization, and Noise Filtering for T2-weighted MRI, IEEE Engineering in Medicine and Biology Conference, 2011, Accepted. Golugula, A, Lee, G, Master, S, Feldman, M, Tomaszewski, J, Madabhushi, A, Supervised Regularized Canonical Correlation Analysis: Integrating Histologic and Proteomic Data for Predicting Biochemical Failures, IEEE Engineering in Medicine and Biology Conference, 2011, Accepted Galero, J, Judkins, A, Bacon, J, Ellison, D, Madabhushi, A, An Integrated Texton and Bag of Words Classifier for Identifying Anaplastic Medulloblastomas, IEEE Engineering in Medicine and Biology Conference, 2011, Accepted Patel, P, Chappelow, J, Tomaszewski, J, Feldman, M, Rosen, M, Madabhushi, A, Spatially Weighted Mutual Information (SWMI) for Registration of Digitally Reconstructed ex vivo Whole Mount Histology and in vivo prostate MRI, IEEE Engineering in Medicine and Biology Conference, 2011, Accepted Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 PEER REVIEWED PUBLICATIONS PEER-REVIEWED CONFERENCE PROCEEDINGS (CONTD) Yu, E, Monaco, J, Shih, N, Feldman, M, Tomaszewski, J, Madabhushi, A, Detection of Prostate Cancer on Histopathology using Color Fractals and Probabilistic Pairwise Markov Models, IEEE Engineering in Medicine and Biology Conference, 2011, Accepted Golugula, A, Lee, G, Madabhushi, A, Evaluating Feature Selection Strategies for High Dimensional, Small Sample Size Datasets, IEEE Engineering in Medicine and Biology Conference, 2011, Accepted Ali, S, Veltri, R, Epstein, J, Christudass, C, Madabhushi, A, “Adaptive Energy Selective active contour with Shape Priors For Nuclear Segmentation and Gleason grading of prostate cancer”, In Proc of Medical Image Computing and Computer Assisted Interventions (MICCAI), pp. 661-669, 2011 (PMID: 22003675). Xiao, G, Madabhushi, A, “Aggregated Distance Metric learning (ADM) for image classification in presence of limited training data”, In Proc of Medical Image Computing and Computer Assisted Interventions (MICCAI), pp. 33-40, 2011 (PMID: 22003681). Tchikindas, L, Sparks, R, Baccon, J, Ellison, D, Judkins, A, Madabhushi, A, “Segmentation of Nodular Medulloblastoma Using Random Walker and Hierarchical Normalized Cuts”, IEEE Northeast Bioengineering Conference (NEBC), pp. 1-2, 2011. Lai, Y, Viswanath, S, Baccon, J, Ellison, D, Judkins, A, Madabhushi, A, “A Texture-based Classifier to Discriminate Anaplastic from Non-Anaplastic Medulloblastoma”, IEEE Northeast Bioengineering Conference (NEBC), pp. 1-2, 2011. Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 PEER REVIEWED PUBLICATIONS PEER-REVIEWED CONFERENCE PROCEEDINGS (CONTD) Basavanhally, A, Ganesan, S, Shih, Natalie, Mies, C, Feldman, M, Tomaszewski, J, Madabhushi, A, “A Boosted Classifier for Integrating Multiple Fields of View: Breast Cancer Grading in Histopathology”, IEEE International Symposium on Biomedical Imaging, Chicago, IL, pp. 125-128, 2011. Sparks, R, Madabhushi, A, “Out-of-Sample Extrapolation using Semi-Supervised Manifold Learning (OSE-SSL): Content-based Image Retrieval for Prostate Histology Grading”, IEEE International Symposium on Biomedical Imaging, Chicago, IL, pp. 734-737, 2011. Shridar, A, Doyle, S, Madabhushi, A, “Boosted Spectral Embedding (BoSE): Applications to ContentBased Image Retrieval of Histopathology”, IEEE International Symposium on Biomedical Imaging, Chicago, IL, pp. 1897-1900, 2011. Tiwari, P, Viswanath, S, Lee, G, Madabhushi, A, “Multi-modal data fusion schemes for integrated classification of imaging and non-imaging data”, IEEE International Symposium on Biomedical Imaging, Chicago, IL, pp. 165-168, 2011. Toth, R, Sparks, R, Madabhushi, A, “Medial Axis based Statistical Shape Model (MASSM): Applications to 3D Prostate Segmentation on MRI”, IEEE International Symposium on Biomedical Imaging, Chicago, IL, pp. 1463-466, 2011. Doyle, S, Feldman, M, Tomaszewski, J, Shih, N, Madabhushi, A, “Cascaded Multi-class Pair-wise Classifier (CASCAMPA) for normal, cancerous, and cancer confounder classes in prostate histology”, IEEE International Symposium on Biomedical Imaging, Chicago, IL, pp. 715-718, 2011. Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 PEER REVIEWED PUBLICATIONS PEER-REVIEWED CONFERENCE PROCEEDINGS (CONTD) Ali, S, Madabhushi, A, ”Active Contour for Overlap Resolution using Watershed based Initialization (ACOReW): Applications to Histopathology”, IEEE International Symposium on Biomedical Imaging, Chicago, IL, pp. 614-617, 2011. Viswanath, S, Tiwari, P, Chappelow, J, Toth, R, Kurhanewicz, J, Madabhushi, A, “CadOnc©: An Integrated Toolkit for Evaluating Radiation Therapy Related Changes in the Prostate Using MultiParametric MRI”, IEEE International Symposium on Biomedical Imaging, Chicago, IL, pp. 2095-98, 2011. Toth, R, Chappelow, J, Kutter, O, Vetter, C, Russ, C, Feldman, M, Tomaszewski, J, Shih, N, Madabhushi, A, Incorporating the whole-mount prostate histology reconstruction program Histostitcher© into the extensible imaging platform (XIP) framework, SPIE Medical Imaging, 2012, Accepted. Sparks, R, Madabhushi, A, Gleason grading of prostate histology utilizing statistical shape model of manifolds (SSMM), SPIE Medical Imaging, 2012, Accepted. Agner, S, Xu, J, Rosen, M, Karthigeyan, S, Englander, S, Madabhushi, A, Spectral embedding based active contour (SEAC): application to breast lesion segmentation on DCE-MRI. SPIE Medical Imaging, vol. 7963, pp. 796305-1 – 12, 2011. Ali, S, Madabhushi, A, Segmenting multiple overlapping objects via an integrated region and boundary based active contour incorporating shape priors: applications to histopathology, SPIE Medical Imaging, 2011, Accepted. Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 PEER REVIEWED PUBLICATIONS PEER-REVIEWED CONFERENCE PROCEEDINGS (CONTD) Sparks, R, Madabhushi, A, Content-based image retrieval utilizing shape modeling and manifold learning, SPIE Medical Imaging, 2011, Accepted. Chowdhury, N, Chappelow, J, Toth, R, Madabhushi, A, Novel approach for building linked statistical shape models for multimodal prostate radiotherapy planning, SPIE Medical Imaging, 2011, Accepted. Janowczyk, A, Chandran, S, Feldman, M, Madabhushi, A, Local morphologic scale: application to segmenting tumor infiltrating lymphocytes in ovarian cancer TMAs, SPIE Medical Imaging, 2011, Accepted. Basavanhally, A, Yu, E, Ganesan, S, Feldman, M, Tomaszewski, J, Madabhushi, A, Incorporating domain knowledge for tubule detection in breast histopathology using O'Callaghan neighborhoods, SPIE Medical Imaging, 2011, Accepted. Toth, R, Bulman, J, Patel, A, Bloch, N, Genega, E, Rofsky, N, Lenkinski, R, Madabhushi, A, Integrating an Adaptive Region Based Appearance Model with a Landmark Free Statistical Shape Model: Application to Prostate MRI Segmentation, SPIE Medical Imaging, 2011, vol. 7962, In Press. Viswanath, S, Chappelow, J, Patel, P, A, Bloch, N, Genega, E, Rofsky, N, Lenkinski, R, Madabhushi, A, Enhanced multiprotocol analysis via intelligent supervised embedding (EMPrAvISE) for prostate cancer detection on MRI, SPIE Medical Imaging, 2011, In Press. Viswanath, S, Palumbo, D, Chappelow, J, Patel, P, A, Bloch, N, Genega, E, Rofsky, N, Lenkinski, R, Madabhushi, A, Empirical evaluation of bias field correction algorithms for computer-aided detection of prostate cancer on T2w MRI, SPIE Medical Imaging, 2011, In Press. Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 PEER REVIEWED ABSTRACTS P Raess, J Monaco, R Chawla, A Bagg, M Weiss, J Choi, and A Madabhushi “Alpha-Hemoglobin Stabilizing Protein Specifically Identifies Nucleated Erythroid Precursors and Enables Identification of Architectural Distortion in Myelodysplastic Syndromes by Computerized Image Analysis", United States and Canadian Academy of Pathology's 101st Annual Meeting, March 17-23, 2012 Accepted. Hipp, J, Smith, S, Cheng, J, Tomlins, S, Monaco, J, Madabhushi, A, Kunju, P, Balis, U, “Optimization of detection of complex cancer morphology using the SIVQ pattern recognition algorithm”, First Congress of the International Academy of Digital Pathology, pp. 28, 2011 (Best Poster Award). Tiwari, P, Kurhanewicz, J, Madabhushi, A, “Computerized quantitative data integration of multiprotocol MRI for identification of high grade prostate cancer in vivo”, in Proc. ISMRM, pp. 2640, 2011. Sparks, R, Madabhushi, A, “Computerized Classification of Benign and Malignant Breast Lesions on DCE-MRI Utilizing Novel Shape Descriptors“ in Proc. ISMRM, pp. 2626, 2011. Agner, S, Xu, J, Karthigeyan, S, Madabhushi, A, “Computerized Lesion Segmentation on DCE MRI Using Active Contours and Spectral Embedding”, in Proc. ISMRM, pp. 136, 2011. Toth, R., Bloch, B.N., Genega, E.M., Rofsky, N.M., Lenkinski, R.E., Rosen, M, Madabhushi, A, "Accurate Prostate Volume Determination from T2-w MRI using Statistical Shape Models," in Proc. ISMRM, pp. 2643, 2011. Xiao, G, Bloch, B. N., Chappelow, J, Genega, E, Rofsky, N, Lenkinski, R, Tomaszewski, J, Feldman, M, Rosen, M, Kalyanpur, A, and Madabhushi, A, “Determining histology-MRI slice correspondences for mapping prostate cancer extent in vivo”, in Proc. ISMRM, pp. 2645, 2011. Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 PEER REVIEWED ABSTRACTS (CONTD) Viswanath, S, Chappelow, J, Tiwari, P, Kurhanewicz, J, Madabhushi, A, "CADOnc: A Computerized Decision Support System for Quantifying Radiation Therapy Changes in the Prostate via MultiParametric MRI", in Proc. ISMRM, pp. 2647, 2011. Viswanath, S, Bloch, B. N., Chappelow, J, Patel, P, Rofsky, N, Lenkinski, R, Genega, E, and Madabhushi, A, "EMPrAvISE: A Computerized Decision Support System for Automated Prostate Cancer Detection from Multi-Protocol MRI", in Proc. ISMRM, pp. 2642, 2011. Hipp, J, Cheng, J, Hansen, J, Hewitt, S, Monaco, J, Madabhushi, A, Han, S, Yan, W, Rodriguez-canales, W, Hipp, J, Tangrea, M, Emmert-Buck, M, Balis, U, “Ring Vector Image analysis (SIVQ): A highthroughput discovery tool for surgical pathologists”, United States and Canadian Academy of Pathology's 100th Annual Meeting, San Antonio, TX, February 26 - March 4, 2011, Accepted. Janowczyk, A, Chandran, S, Feldman, M, Madabhushi, A, “Quantifying Tumor Infiltrating Lymphocytes in Ovarian Cancer TMAs”, United States and Canadian Academy of Pathology's 100th Annual Meeting, San Antonio, TX, February 26 - March 4, 2011, Accepted. Xu, J, Sparks, R, Janowczyk, A, Tomaszewski, J, Feldman, M, Madabhushi, A, “High-Throughput Prostate Cancer Gland Segmentation and Classification from Digitized Needle Core Biopsies”, United States and Canadian Academy of Pathology's 100th Annual Meeting, San Antonio, TX, February 26 - March 4, 2011, Accepted. Sparks, R, Madabhushi, A, “Quantifying Gland Morphology for Computerized Prostate Cancer Detection and Gleason Grading”, United States and Canadian Academy of Pathology's 100th Annual Meeting, San Antonio, TX, February 26 - March 4, 2011, Accepted. Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 PEER REVIEWED ABSTRACTS (CONTD) Monaco, J, Tomaszewski, J, Madabhushi, A, “Automated Tumor Volume Estimation Using Digitized Prostatectomy Specimens”, United States and Canadian Academy of Pathology's 100th Annual Meeting, San Antonio, TX, February 26 - March 4, 2011, Accepted. Lee, G, Doyle, S, Monaco, J, Feldman, M, Tomaszewski, J, Master, S, Madabhushi, A, “Fusion of proteomic and histologic image features for predicting prostate cancer recurrence after radical prostatectomy”, United States and Canadian Academy of Pathology's 100th Annual Meeting, San Antonio, TX, February 26 - March 4, 2011, Accepted. Basavanhally, A, Ganesan, S, Feldman, M, Mies, C, Tomaszewski, J, Madabhushi, A, “Histologic Imagebased Classifier for Predicting Outcome of ER+ Breast Cancers”, Laboratory Investigation, vol. 91[1], pp. 27A-27A, 2011. Sridhar, A, Doyle, S, Tomaszewski, J, Feldman, M, Madabhushi, A, “A Content-Based Image Retrieval System for Digitized Prostate Histopathology”, United States and Canadian Academy of Pathology's 100th Annual Meeting, San Antonio, TX, February 26 - March 4, 2011, Accepted. Tiwari, P, Kurhanewicz, J, Madabhushi, A, “Multimodal Integration of Magnetic Resonance Imaging and Spectroscopy for Detection of Aggressive Prostate Cancer”, Innovative Minds in Prostate Cancer Today (IMPaCT), 2011, Accepted. Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 LCIB WORKSHOPS 2011 1. Histopathology Image Analysis, Clinical Challenges and Quantitative Image Analysis Solutions, MICCAI 2011 http://lcib.rutgers.edu/hima2011/ 2. Prostate Cancer Imaging, Computer Aided Diagnosis, Prognosis, and Intervention, MICCAI 2011 http://lcib.rutgers.edu/pci2011/ 3. Tutorial on Manifold Learning With Medical Images, MICCAI 2011 http://campar.in.tum.de/personal/mateus/2011MiccaiManifoldTutorial/html 32 Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 SPECIAL ISSUE JOURNAL OF PATHOLOGY INFORMATICS In conjunction with the Histopathology Image Analysis (HIMA) workshop at MICCAI 2011 Editors: Anant Madabhushi, Nasir Rajpoot, Metin Gurcan, Michael D. Feldman http://www.jpathinformatics.org/ 33 Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 ACTIVE RESEARCH METHODS MULTIMODAL DATA INTEGRATION COMPUTER AIDED DIAGNOSIS IMAGE SEGMENTATION MACHINE LEARNING IMAGE REGISTRATION RADIOLOGY HISTOPATHOLOGY BIOINFORMATICS PROSTATE CANCER OVARIAN CANCER BREAST CANCER MEDULLOBLASTOMA MYELODYSPLASTIC SYNDROME 34 Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 METHODS CLASSIFICATION FEATURE SELECTION ACTIVE LEARNING & CONTENT BASED IMAGE RETRIEVAL IMAGE SEGMENTATION MULTI-MODAL REGISTRATION NON-LINEAR DIMENSIONALITY REDUCTION KNOWLEDGE REPRESENTATION AND MULTI-MODAL DATA FUSION Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 IMAGE REGISTRATION 36 Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 Incorporating Histostitcher© into the Extensible Imaging Platform (XIP) Framework • • Histostitcher algorithm incorporated into professional XIP software framework. GPU rendering, “Google Maps” – like zooming and scrolling Previous version of Histostitcher© graphical user interface usable prototype developed using Matlab. New version of Histostitcher© graphical user interface developed using XIP. 37 Toth, R., Chappelow, J., Kutter, O., Vetter, C., Russ, C., Feldman, M., Tomaszewski, J., Shih, N., Madabhushi, A., “Incorporating the Whole-Mount Prostate Histology Reconstruction Program Histostitcher© into the Extensible Imaging Platform (XIP) Framework,” SPIE Medical Imaging, 2012, Accepted. Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 MULTIMODAL IMAGE REGISTRATION 38 Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 Histology-MRI Fusion Step 1: Multiprotocol MRI Registration T2+DCE DCE T2 Step 3: Histology + MRI Fusion • Extent of CaP is difficult to find on multi protocol MRI Step 2: Stitching sections using Histostitcher © f f 1st Stitch 2nd Stitch 3rd Stitch f • CaP ground truth exists on high resolution histology fragments 39 Patel, P, Chappelow J, Tomaszewski JE, Feldman MD, Rosen MA, Shih N, Madabhushi A. 2011. Spatially weighted mutual information (SWMI) for registration of digitally reconstructed ex vivo whole mount histology and in vivo prostate MRI. IEEE International Conference of Engineering in Medicine and Biology Society (EMBS). :6269-627 Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 Determination of histology-MRI slice correspondences group-wise comparison Final histology-MRI slice correspondences 2D affine histology-MRI registration 40 itted to Computerized Medical Imaging and Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 KNOWLEDGE REPRESENTATION AND DATA FUSION Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 Weighted Multi-Kernel Learning (WMKL) Classification Accuracy A multi-channel framework for merging imaging and non-imaging modalities for stratifying at-risk prostate cancer patients WMKL (Fusion) Proteomics Architectural Morphological 35 45 55 65 75 85 95 Lee, G, Feldman MD, Master SR, Tomaszewski, JE, Madabhushi, A. Weighted Multi-Kernel Learning (WMKL): Predicting Biochemical Recurrence by Combining Histologic Imaging and Proteomic Signatures (Submitted to Bioinformatics 2011) 42 Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 Supervised Regularized Canonical Correlation Analysis Objective: Integrate Histology and Proteomic data to make prognostic decisions Histology Images Novel Variation: Uses a Feature Selection method to determine optimal embedding Proteomic data Class 1 Score 1 Score 2 : Score n T-test WRST WLT Class 2 Optimal Score Supervised Variation SRCCA produces higher classification accuracies Optimal SRCCA Embedding 43 Golugula, A, Lee G, Master SR, Feldman MD, Tomaszewski JE, Madabhushi A. 2011. Supervised Regularized Canonical Correlation Analysis: Integrating Histologic and Proteomic Measurements for Predicting Biochemical Recurrence Following Prostate Surgery. Accepted for publication in BMC Bioiformatics 2011 Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 IMAGE SEGMENTATION Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 ASM Initialization MRS signals inside the prostate are distinct from signals outside the prostate The cluster with the most signals, in all experiments, shown to be background spectra (green below) Therefore, largest cluster is removed 45 Toth, R., Tiwari, P., Rosen, M., Reed, G., Kurhanewicz, J., Kalyanpur, A., Pungavkar, S., Madabhushi, A. “A Magnetic Resonance Spectroscopy Driven Initialization Scheme for Active Shape Model Based Prostate Segmentation.” Medical Image Analysis 15 (2), Apr 2011. pp. 214-225. Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 Prostate Volume Estimation Red = Ground Truth Green = ASM Result R2 = .82 Each voxel: = 0.22cm x 0.027cm x 0.027cm = 1.6 x 10-4 mL R2 = .70 Robert Toth, B. Nicolas Bloch, Elizabeth M. Genega, Neil M. Rofsky, Robert E. Lenkinski, Mark A. Rosen, Arjun Kalyanpur, Sona Pungavkar, Anant Madabhushi., “Accurate Prostate Volume Estimation Using Multi-Feature Active Shape Models on T2-Weighted MRI,” Academic Radiology, 2011, 18 (2), Jun 2011, pp. 745-754. Bulman, J.C., Toth, R., Patel, A.D., Bloch, N.B., MacMahon C.J., Ngo L., Madabhushi, A., Rofsky, N.M., "Automated Computer-Derived Prostate Volumes from MRI Data: Comparison to Radiologist-Derived MRI Volumes and Pathology Specimen Volumes," Radiology 2011. Accepted pending revisions. 46 Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 Prostate Volume Estimation (b) Ellipsoid (yellow) is generated from the axes, and volume of ellipsoid is calculated. Ellipsoid (yellow) may be significantly different from ground truth volume (green). Aggregate of slice segmentations (pink) are used to estimate the prostate volume (green). 47 Toth, R., Bloch, B.N., Genega, E.M., Rofsky, N.M., Lenkinski, R.E., Rosen, M.A.,Madabhushi, A., "Accurate Prostate Volume Determination from T2-w MRI using Statistical Shape Models," ISMRM 2011. Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 Prostate Volume Estimation • R2 = 0.865 160 140 Path Volume (cc) • Comparison of automated segmentation prostate volume (“MFA”) to clinical pathology volume. y = 0.888x + 12.554 R² = 0.865 120 100 80 60 40 20 0 0 50 100 150 MFA Volume (cc) 48 Bulman, J.C., Toth, R., Patel, A.D., Bloch, N.B., MacMahon C.J., Ngo L., Madabhushi, A., Rofsky, N.M., "Automated Computer-Derived Prostate Volumes from MRI Data: Comparison to Radiologist-Derived MRI Volumes and Pathology Specimen Volumes," Radiology 2011. Accepted pending revisions. Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 Medial Axis Statistical Shape Model • Markers (“atoms”) along the medial axis of an object can capture underlying shape variations. • Much less reconstruction error (dark red on the right) than traditional landmark based shape models. • Accurate shape variations capture in application to prostate segmentation (bottom right) 49 Robert Toth, Rachel Sparks, Anant Madabhushi., “Medial Axis Based Statistical Shape Model (Massm): Applications To 3d Prostate Segmentation On MRI,” ISBI 2011, Under Review. Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 Landmark Free Statistical Shape Models and Region Based Statistical Appearance Models • Levelset function used to create statistical shape model. • Smooth and accurate • Bayesian based appearance model adapts to current image, and is intelligently combined with statistical shape model. • Highly accurate segmentation results for prostate MRI. 50 Robert Toth, Julie Bulman, Amish Patel, B. Nicholas Bloch, Elizabeth M. Genega, Neil M. Rofsky, Robert E. Lenkinski, Anant Madabhushi. “Integrating an Adaptive Region Based Appearance Model with a Landmark Free Statistical Shape Model: Application to Prostate MRI Segmentation.” SPIE Medical Imaging 2011, Accepted. Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 Adaptive Energy Selective Active Contour with Shape Priors for Nuclear Segmentation and Gleason Grading of Prostate Cancer - Able to predict Gleason grade 3 vs 4 with 84.1% accuracy - Using 7 nuclear features extracted from individual nuclei - Developed an efficient segmentation scheme that selectively invoked shape prior in to active contour. Evolve Initial Scene Segmentation from Watershed Transformation Detect Overlap – Concavity Detection Selectively invoke Shape Prior appropriate level sets Figure PCA representation of nuclear morphologic features reveals separation of (a) primary grade 3 and grade 4 and (b) Gleason patterns 6, 7, 8 and 9. How many studies used here – specify clearly. 51 Ali, S, Veltri R, Epstein JI, Christudass C, Madabhushi A. 2011. Adaptive Energy Selective Active Contour with Shape Priors for Nuclear Segmentation and Gleason Grading of Prostate Cancer. International Conference on Medical Image Computing and Computer-Assisted Intervention (MICCAI) . 6891:661–669 Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 GPU Implementation of Integrated Shape Based Active Contour: Application to Digital Pathology A GPU accelerated framework, implemented using the NVIDIAs (CUDA), aimed for the segmentation of nuclei in (H&E) images based on an integrated active contour model. Parallelized implementation of the Ali et al. scheme with multiple level sets operating in parallel. Exploit parallelism and efficient memory management in the GPU to achieve massive speed up of 19x compared to CPU. Ali, S, Madabhushi, A. ”Segmenting multiple overlapping objects via an integrated region and boundary based active contour incorporating shape priors: applications to histopathology”, SPIE Medical Imaging, 79622W (2011). Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 Segmentation of Nodular Medulloblastoma A novel fusion of Hierarchical Normalized Cuts (HN-Cuts) and Random Walker (RW) to segment medulloblastoma nodules (a) Original Image (b) HN-Cuts Segmentation (c) RW Probability Map (d) Final Segmentation 53 Tchikindas, L, Sparks R, Baccon J, Ellison D, Judkins AR, Madabhushi A. 2011. Segmentation of nodular medulloblastoma using Random Walker and Hierarchical Normalized Cuts. IEEE North-East Bioengineering Conference (NEBEC). :1-2 Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 High-throughput Gland Segmentation and Classification as Benign and Malignant from Digitized Needle Core Biopsies Detection: a hierarchical frequency weighted mean shift normalized cut (HNCut) for initial detection of glands HNCut-CGAC model for gland segmentation Segmentation: Color gradient based geodesic active contour model (CGAC) for gland segmentation Classification: a diffeomorphic based similarity (DBS) feature extraction for classification of glands as benign or cancerous. SVM for gland classification Morphological feature extraction 54 Xu, J, Janowczyk A, Chandran S, Madabhushi A. 2011. A high-throughput active contour scheme for segmentation of histopathological imagery. Medical Image Analysis Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 NON-LINEAR DIMENSIONALITY REDUCTION AND MANIFOLD LEARNING Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 Out-of-Sample Extrapolation Utilizing Semi-Supervised Learning (OSE-SSL) Semi-supervised learning Learn embedding with class labels known for some samples Out of Sample Extrapolation Estimate embeddings for new samples OSE-SSL Integrating the use of known label information (SSL) and the ability to estimate new embeddings (OSE). Resulting in an manifold learning technique that is able to extrapolate new cases and take semantic information into account. 56 Sparks, RE, Madabhushi, A. Out-of-Sample Extrapolation for Semi-Supervised Manifold Learning (OSE-SSL): Content Based Image Retrieval of Prostate Histology. IEEE International Society of Biomedical Imaging (ISBI) 2011, pp 734-738. Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 Statistical Shape Model of Manifolds (SSMM) Changes in dataset affect low dimensional embedding obtained from manifold learning (b) (b) (a) (c) (c) (a) (a) Figure 1. Two manifolds for a dataset of 888 prostate histopathology glands, where each manifold is trained with only 800 glands. (1) From a set of manifolds, (2) SSMM constrains the manifold for the full dataset, giving a robust representation we construct a SSMM. (b) (b) μ+3σ (c) μ (a) μ-3σ (c) 57 Sparks, RE, Madabhushi, A. Gleason grading of prostate histology utilizing statistical shape model of manifolds (SSMM.) In Proc. SPIE 2012, in press. Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 Semi-Supervised Graph Embedding with Active Learning New methodology leverages Active Learning with Semi-Supervised Dimensionality Reduction (DR) for improved data representations New Data Representation Original Data Representation SVM Decision Boundary New Active Learning Candidates Active Learning Candidates improve data representation Initial Training Labels F Class 1 Class 2 Query Unlabeled Lee, G., Madabhushi A., "Semi-Supervised Graph Embedding Scheme with Active Learning (SSGEAL): Classifying High Dimensional Biomedical Data", in Pattern Recognition in Bioinformatics (PRIB), LNCS 6282, pp. 207-218, 2010. Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 CLASSIFICATION Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 Aggregated Distance Metric Learning (ADM) for Image Classification in Presence of Limited Training Data BDM Aggregated distance t … … … … BD M t Final classification via NN search Training cohort of BDM on each different classes: Training image sub-sets ( : image, : label) to get distance Distance of test image metric to each class t based on all learned Xiao, G., Madabhushi, A.: Aggregated distance metric learning (ADM) for image classification in presence of limited training data, MICCAI, Part III, LNCS 6893, 33-40 (2011) 60 Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 ACTIVE LEARNING AND CONTENT BASED IMAGE RETRIEVAL Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 Boosted Spectral Embedding (BoSE): Applications to Content-Based Image Retrieval Query, Annotated Database … Feature Extraction … … Feature Weighting … Reduce dimensionality of data using BoSE: it selectively weights the features that are best able to discriminate between two classes. BoSE PR Curves Perform image retrieval in lowerdimensional space created by BoSE and compare it to image retrieval in lower-dimensional space created by spectral embedding (SE). The greater the area under the Precision-Recall curve (AUC), the better the performance of the CBIR system. 62 Sridhar, A,. Doyle S., Madabhushi, A. Content-Based Image Retrieval of Digitized Histopathology in Boosted Spectrally Embedded Spaces . IEEE Transactions on Information Technology in Biomedicine, November 2011. (submitted) Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 Content Based Image Retrieval (CBIR) with Explicit Shape Descriptors a. Database Image Feature Extraction OSE-SSL Results b. Learn Embeddings c. Query Image Feature Extraction d. Extrapolate Query e. Image Retrieval …. Precision –Recall Curves. OSESSL has better performance rates compared to manifold learning or OSE. Evaluated on a prostate histopathology dataset containing 888 glands. 63 Sparks, RE, Madabhushi, A.. Content-based image retrieval utilizing Explicit Shape Descriptors: applications to breast MRI and prostate histopathology. Proc. SPIE 2011, Vol. 7692 pp. 79621I--13. Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 Consensus Active Learning Active Learning Example Random Learning: All Samples Eligible Active Learning: Intermediate Samples Eligible Consensus Learning: Intermediate Samples from Multiple Algorithms + 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 High Certainty Intermediate Certainty High Certainty = Eligible for Uninformative Consensus Learning Labeling Points Points Active Learning (left) identifies samples whose class is uncertain, and these samples are assigned to an expert for labeling. Compared with Random Learning (top right), Active Learning reduces the number of eligible samples for labeling. Multiple Active Learning algorithms can be combined (bottom right) to generate a consensus learning set, which yields an accurate classifier with fewer samples eligible for labeling. Doyle, S, Monaco JP, Feldman MD, Tomaszewski JE, Madabhushi A. 2011. An Active Learning Based Classification Strategy for the Minority Class Problem: Application to Histopathology Annotation. BMC bioinformatics. 12(1):424 64 Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 BIOINFORMATICS APPLICATIONS FEATURE SELECTION/EVALUATION Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 Variable Ranking with PCA A method for finding biomarkers applied to identify multiparametric MR imaging markers for prostate cancer diagnosis and grading Identify imaging markers Perform PCA Compute PCA-VIP 66 Ginsburg, S, Tiwari, P, Kurhanewicz, J, Madabhushi, A. 2011. Variable ranking with PCA: Finding multiparametric MR imaging markers for prostate cancer diagnosis and grading. Workshop on Prostate Cancer Imaging: Computer-Aided Diagnosis, Prognosis, and Intervention (in conjunction with MICCAI). 6963:146-157. Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 Robustness of a Feature Selection (FS) Scheme Problem: Classification Accuracy, de facto measure to evaluate FS, is not always sufficient Objective: To determine a new measure to evaluate FS in the context of small sample size data With subsampling to create a small sample size, few features appear more than 10 times and results are not consistent High Dimensional Proteomic Feature vector Robustness: higher is more optimal 67 Golugula, A, Lee, G and Madabhushi, A, Evaluating Feature Selection Strategies for High Dimensional, Small Sample Size Datasets, IEEE Engineering in Medicine and Biology Conference, 2011 Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 COMPUTER AIDED DIAGNOSIS (CAD) Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 CAD FOR PROSTATE MRI Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 Evaluation of bias field correction (BFC) schemes for prostate cancer detection on T2w MRI Compared 3 different algorithms, N3 found to offer best BFC. Choice of BFC scheme should be based on overarching application, current evaluation measures not dependent on application. 70 Viswanath, SE, Palumbo D, Chappelow J, Patel P, Bloch BN, Rofsky NM, Lenkinski RE, Genega EM, Madabhushi A. 2011. Empirical evaluation of bias field correction algorithms for computer-aided detection of prostate cancer on T2w MRI. SPIE Medical Imaging. 7963:79630V. Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 Interplay between bias field correction, intensity standardization, and noise filtering for T2w MRI Quantitatively evaluated 7 different combinations of 3 pre-processing operations, determined best for optimal prostate cancer detection Original After Bias field correction After Noise correction After Intensity standardization Histograms from 4 different studies showing the effect on the intensity distributions after each operation. Last set show relatively noisefree, well-aligned distributions, shown to be most optimal for classification. 71 Palumbo, D, Yee B, O'Dea P, Leedy S, Viswanath SE, Madabhushi A. 2011. Interplay between bias field correction, intensity standardization, and noise filtering for T2-weighted MRI. IEEE International Conference of Engineering in Medicine and Biology Society (EMBS). :5080-5083. Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 Determining Quantitative Imaging Signatures (QIS) for Central Gland and Peripheral Zone Prostate Tumors on T2-w MRI Determined that unique combinations of texture features characterize each of CG and PZ CaP Classification via QDA yielded 0.86 AUC for CG CaP, 0.74 AUC for PZ CaP Central Gland (CG) Prostate Cancer Original MRI + CaP extent Top texture feature in QIS CaP detection result based off QIS Peripheral Zone (PZ) Prostate Cancer 72 S. Viswanath, BN Bloch, JC Chappelow, R Toth, NM Rofsky, EM Genega, RE Lenkinski, A. Madabhushi, Central Gland and Peripheral Zone Prostate Tumors have Significantly Different Quantitative Imaging Signatures on 3 Tesla Endorectal, In Vivo T2-Weighted Magnetic Resonance Imagery, JMRI, (Provisionally accepted), 2011 Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 Multi-modal Weighted Embedding Representation for data Combination (MaWERiC) for prostate cancer diagnosis Feature Extraction Wavelet Feature Extraction- MRI Wavelet Feature Extraction- MRS Data Representation 0.6 DR on high D Gabor features 0.4 0.2 0 -0.2 DR on high D wavelet MRS features 0.6 0.4 0.2 0 -0.4 0.6 -0.2 0.4 0.3 0.4 0.2 0.2 0.1 0 -0.2 0.4 0.2 0 -0.1 0.2 0 -0.2 0 -0.2 -0.2 -0.4 -0.4 -0.6 Data Integration (MRI+MRS) Data classification Tiwari, P, et al.. Multimodal wavelet embedding representation for data combination (MaWERiC): integrating magnetic resonance imaging and spectroscopy for prostate cancer detection. NMR in Biomedicine, 2011. Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 Weighted Combination of Multi-parametric MR markers for evaluating radiation therapy related changes in the prostate Heat map for treatment evaluation Expert delineated ground truth Metabolic map T2w MRI MP-MRI map ADC map High treatment change CaP recurrence CaP pre-RT New CaP foci Low treatment change 74 Tiwari, P, et al. . Weighted Combination of Multi-Parametric MR Imaging Markers for Evaluating Radiation Therapy Related Changes in the Prostate. Workshop on Prostate Cancer Imaging: Computer-Aided Diagnosis, Prognosis, and Intervention (in conjunction with MICCAI). 6963:80-91 Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 Concurrent segmentation of the prostate on MRI and CT via linked statistical shape models for radiotherapy planning 75 Chowdhury, N, ,Toth R, Chappelow J, Kim S, Motwani, S, Punekar, S, Hahn S, Vapiwala N, Lin H, Both S, Madabhushi A. Concurrent segmentation of the prostate on MRI and CT via linked statistical shape models for radiotherapy planning (accepted, pending changes to Medical Physics) Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 Multimodal integration of magnetic resonance imaging and spectroscopy for detection of high grade prostate cancer Clinical Goal: To quantitatively combine imaging (T2-w MRI) and nonimaging (MRS) data to distinguish high (aggressive) vs. low-grade (indolent) prostate cancer (CaP) Step 1. Integrated low dimensional representation of T2w MRI/MRS 1 0.9 0.8 Receiver operating characteristic (ROC) curve for high vs. low grade CaP classified on a per voxel basis via a RF classifier. 0.7 0.6 0.5 0.4 0.3 T2-w MRI MRS T2-w MRI + MRS 0.2 0.1 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 76 Tiwari, P, Kurhanewicz J, Rosen MA, Madabhushi A. 2010. Semi Supervised Multi Kernel (SeSMiK) Graph Embedding: Identifying Aggressive Prostate Cancer via Magnetic Resonance Imaging and Spectroscopy. International Conference on Medical Image Computing and Computer-Assisted Intervention (MICCAI). 6363:666-673. Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 CAD FOR PROSTATE WHOLE MOUNT HISTOLOGY Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 Detection of Prostate Cancer on Histopathology using Color Fractals and Probabilistic Pairwise Markov Models Novel per pixel color fractal dimension algorithm for whole-mount prostate histopathology Incorporated novel Markov Prior (PPMM) for spatial dependencies Provide means for tumor volume estimation; facilitate initial step of Gleason grading; reduce inter-expert variability Something Ground Truth CFD Alone CFD + PPMM 78 Yu E, Monaco JP, Tomaszewski JE, Shih N, Feldman MD, Madabhushi A. 2011. Detection of prostate cancer on histopathology using color fractals and Probabilistic Pairwise Markov models. IEEE International Conference of Engineering in Medicine and Biology Society (EMBS): 3427-3430. Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 Two Novel Means for Varying the Performance of MRF-based Classification Systems Prostate histology images First method allows variation of performance of MRFs classified using MAP estimation Second method allows variation of performance of MRFs classified using MPM estimation 79 J Monaco and A Madabhushi, Class-Specific Weighting for Markov Random Field Estimation: Application to Medical Image Segmentation, submitted to Medical Image Analysis. Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 Weighted Maximum Posterior Marginals for Random Fields (a) H&E stained whole-mount prostate histology section; black ink mark indicates “ground-truth” of CaP extent as delineated by a pathologist. (b) Result of automated gland segmentation. (c) Magnified view of white box in (b). (d) Green dots indicate the centroids of those glands labeled as malignant. 80 Monaco, JP, Madabhushi A. 2011. Weighted Maximum Posterior Marginals for Random Fields using an Ensemble of Conditional Densities from Multiple Markov Chain Monte Carlo Simulations.. IEEE Transactions on Medical Imaging Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 Cascaded Multi-Category Classification Dataset Cancer Grade 3+4 Grade 3 Non-Cancer Grade 5 Grade 4 • Distinguish classes sequentially, determined via domain knowledge • Advantages: Binary decision boundaries, simple classifiers, optimized feature sets, high class separability, minimal class heterogeneity Doyle, S, Feldman MD, Tomaszewski JE, Shih N, Madabhushi A. 2011. Cascaded Multi-Class Pairwise Classifier (CascaMPa) For Normal, Cancerous, And Cancer Confounder Classes In Prostate Histology. IEEE International Symposium on Biomedical Imaging (ISBI). :715-718 Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 CAD FOR BREAST DCE MRI Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 Spectral Embedding based Active Contour (SEAC) Spectral Embedding Active contour segmentation Agner, S, Xu J, Rosen MA, Karthigeyan S, Englander S, Madabhushi A. 2011. Spectral embedding based active contour (SEAC): application to breast lesion segmentation on DCE-MRI. SPIE Medical Imaging. 7963 Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 Spectral Embedding based Registration Registration of Breast Dynamic Contrast Enhanced Magnetic Resonance Imaging Unregistered SI Registered SERg 84 Karthigeyan S, Agner, S, et al. Submitted for Aresty Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 Explicit Shape Descriptors Quantifying Breast DCEMRI Morphology Dissimilar High Value Similar Low Value 85 Sparks, RE, Madabhushi, A. Computerized Classification of Benign and Malignant Breast Lesions on DCE-MRI Utilizing Novel Shape Descriptors. ISMRM Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 CAD ON BREAST CANCER HISTOPATHOLOGY Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 Tubule Detection in Breast Cancer Histopathology Identifying tubules by quantifying the spatial arrangement of nuclei and lumen objects 87 Basavanhally, A, Yu E, Xu J, Ganesan S, Feldman MD, Tomaszewski JE, Madabhushi A. 2011. Incorporating domain knowledge for tubule detection in breast histopathology using O'Callaghan neighborhoods. SPIE Medical Imaging. 7963:796310. Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 Boosted Classifier For Integrating Multiple Fields Of View: Breast Cancer Grading In Histopathology Combining FOVs at various sizes to yield improved grading for ER+ breast cancer patients Multi-FOV Framework Results FOV size 4000 x 4000 2000 x 2000 1000 x 1000 500 x 500 250 x 250 multi-FOV AUC 0.573 0.702 0.787 0.819 0.605 0.816 Basavanhally, A, Ganesan S, Shih N, Mies C, Feldman MD, Tomaszewski JE, Madabhushi A. 2011. A Boosted Classifier For Integrating Multiple Fields Of View: Breast Cancer Grading In Histopathology. IEEE International Symposium on Biomedical Imaging (ISBI). :125-128 Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 Multi-Parametric, Multi-FOV ER+ Breast Cancer Prognosis: Combination of H & E and IHC Stained Histopathology Independent multi-FOV classifiers for tissue architecture (H & E) and CD34 expression (IHC) are combined to yield improved prognostic prediction Tissue architecture Patient Prediction of patient outcome CD34 expression Basavanhally, A, Feldman MD, Shih N, Mies C, Tomaszewski JE, Ganesan S, Madabhushi A., Multi-field-of-view strategy for image-based outcome prediction of multi-parametric estrogen receptor-positive breast cancer histopathology: Comparison to Oncotype DX, Journal of Pathology Informatics, 2011 (in press). Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 CAD ON OVARIAN CANCER TISSUE MICROARRAYS 90 Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 Local Morphologic Scale (LMS): Classifying Tumor Infiltrating Lymphocytes in Ovarian Cancer Tissue MicroArrays TILs Non-TILs 1. Identify lymphocytes using HNcut 2. Quantify local morphology using LMS 3. Generate features from LMS signature 4. Use supervised classifier to identify Tumor Infiltrating Lymphocytes (TILs), a prognostic indicator for Ovarian cancer A. Janowczyk, S. Chandran, M. Feldman, A. Madabhushi Local Morphologic Scale: Application to Segmenting Tumor Infiltrating Lymphocytes in Ovarian Cancer Tissue MicroArrays SPIE Medical Imaging 2011. Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 CAD FOR THE DETECTION OF MYELODYSPLASTIC SYNDROME 92 Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 Automated Detection of Myelodysplastic Syndrome Goal: differentiate MDS from normal tissue via the analysis of islands erythroid precursors Procedure: Step1) Segment islands, Step2) Extract features, Step 3) Classify images MDS tissue specimen (erythroid precursors in brown) Map of adipose (black) and everything else (white) 85th Percentile of Area (cyan) Adipose Area (green) Both (black) 93 Segmentations or precursor islands Classification Results: Receiver Operator Characteristic Curve P Raess, J Monaco, R Chawla, A Bagg, M Weiss, J Choi, and A Madabhushi, Image Segmentation with Implicit Color Standardization Using Cascaded EM: Detection of Myelodysplastic Syndromes, United States and Canadian Academy of Pathology's 101st Annual Meeting, March 17-23, 2012, Accepted Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 CLASSIFICATION OF ANAPLASTIC MEDULLOBLASTOMA 94 Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 An Integrated Texton and Bag of Words Classifier for Identifying Anaplastic Medulloblastomas Determine the aggressiveness of medulloblastoma to help construct personalized treatment plans for patients, a novel application of the Texton and Bag of Words classification scheme Currently achieving an average of .76 classification accuracy, a .17 increase over initial parameters Original Image Feature Generation Texton Application Texton Map Texton Model for Classifying Bag of Words Application Galaro, J, Judkins AR, Ellison D, Baccon J, Madabhushi A. 2011. An integrated texton and bag of words classifier for identifying anaplastic medulloblastomas. IEEE International Conference of Engineering in Medicine and Biology Society (EMBS). :3443-3446. Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 A Texture-based Classifier to Discriminate Anaplastic from Non-Anaplastic Medulloblastoma Haar Haralick Laws Texture features in conjunction with ensemble classification found to allow differentiation between anaplastic and nonanaplastic medulloblastoma 96 Lai, Y, Viswanath SE, Baccon J, Ellison D, Judkins AR, Madabhushi A. 2011. A texture-based classifier to discriminate anaplastic from non-anaplastic medulloblastoma. IEEE North-East Bioengineering Conference (NEBEC). :1-2. Laboratory for Computational Imaging and Bioinformatics (LCIB) Annual Report 2011 Funding Agencies (Acknowledgements) • • • • • • • • National Cancer Institute (NIH) Department of Defense (DOD) New Jersey Commission on Cancer Research (NJCCR) Society for Imaging Informatics in Medicine (SIIM) Aresty Foundation Rutgers University Cancer Institute of New Jersey Burroughs Wellcome Fund