Center for Computational Imaging and Personalized Diagnostics (CCIPD) 2015 Annual Report Director: Anant Madabhushi, PhD Professor, Department of Biomedical Engineering CCIPD Website: http://ccipd.case.edu CCIPD in 2015 Wickenden Building Dept. of Biomedical Engineering Case Western Reserve University 2071 Martin Luther King Drive Cleveland, Ohio 44106-7207 Center Space, Room 517 Faculty Offices, Room 523 Center Space, Room 525 Center Members CCIPD Members Center Director: Anant Madabhushi, PhD Research Faculty Satish Viswanath George Lee Pallavi Tiwari Research Associates Mahdi Orooji Andrew Janowczyk Rakesh Shiradkar Soumya Ghose Cheng Lu Jon Whitney Jhimli Mitra Graduate Students Sahir Ali Shoshana Ginsburg Lin Li Jacob Antunes Nathan Braman Prateek Prasanna Gregory Penzias Xiangxue Wang Niha Beig Scientific Software Engineer Ahmad Algohary Yu Zhou Visiting Scientists Angel Cruz (Colombia) Jun Xu (China) David Romo-Buchelli (Colombia) Tao Wan (China) German Corredor Prada (Colombia) Mehdi Alilou (Iran) Kan Jia (China) Gavin Hanson (Case Medical School) Undergraduate Students Patrick Leo Ania Gawlik Nikita Agrawal Jay Patel Andrew Rose Prathyush Chirra Nathan Landis Michael Volkovitsch Administrative Staff Ann Tillet Francisco Aguila Center Members Center Director Anant Madabhushi, PhD Research Faculty Pallavi Tiwari, PhD George Lee, PhD Satish Viswanath, PhD Center Members Research Scientists Mahdi Orooji, PhD Cheng Lu, PhD Jon Whitney, PhD Andrew Janowczyk, PhD Rakesh Shrikadkar, PhD Jhimli Mitra, PhD Soumya Ghose, PhD Center Members Graduate Students Niha Beig, BS Xiangxue Wang, BS Sahir Ali, MS Lin Li, BS Prateek Prasanna, MS Greg Penzias, BS Jacob Antunes, BS Shoshana Ginsburg, MS Nathan Braman, BS Center Members Visiting Scientists David RomoBucheli, PhD Jun Xu, PhD Mehdi Alilou, PhD Gavin Hanson Tao Wan, PhD Angel Cruz, PhD German Corredor Prada, Phd Kan Jia, MS Center Members Scientific Software Programmers Yu Zhou, MS Ahmad Algohary, MS Administrative Staff Ann Tillett, BS Frankie Aguila, BS Center Members Undergraduate Students Patrick Leo Ania Gawlik Jay Patel Nikita Agrawal Nathaniel Landis Prathyush Chirra Michael Volkovitsch Andrew Rose Recent Alumni George Lee, PhD, Research Professor, CCIPD Asha Singanamalli, PhD, GE Research Mirabela Rusu, PhD, GE Research Haibo Wang, PhD, Philips Research Partnerships CCIPD Computational Imaging in Health and Precision Medicine Breast Cancer Renal Cell Carcinoma Data Science at CWRU Colorectal Cancer Prostate Cancer Oropharyngeal Cancer Lung Cancer Brain Tumor Awards and Accomplishments Anant Madabhushi induction into AIMBE Crain’s 40 under 40 article on Anant Madabhushi Jacob Antunes, Ania Gawlik, and Michael Volkovitsch winners of Choose Ohio First Scholarship Awards and Accomplishments Prateek Prasanna, Trainee Stipend for ISMRM 2015 Mahdi Orooji, Linda Arena Endowed Scholarship Award Shoshana Ginsburg, 1st place Graduate Student Research Award at ShowCASE Press Coverage in 2015 “MRI-based texture analysis predicts survival in glioma patient”, AuntMinnie.com, November 15th, 2015. “CWRU researchers building digital pathology tools to predict cancer outcomes: Receive National Institutes of Health grants to lead image-analysis efforts”, Press Release (EurekAlerts), November 11th, 2015. “Find out how translational research impacts health care at Ford Distinguished Lecture Oct. 26”, Case School of Engineering, October 23rd, 2015. “Madabhushi and Viswanath awarded patent in multimodal data fusion” o The Daily, October 23rd, 2015. “Postdoctoral researcher wins scholarship for training in lung cancer prevention, treatment” o The Daily, October 9th, 2015 “Research faculty George Lee receives NIH award for big data” o The Daily, October 9th, 2015 “Several of CWRU’s own selected to Crain’s Cleveland Business “Forty Under 40” class of 2015” o http://www.crainscleveland.com/event/crains/3272035/crains-forty-under40 “Biomarker may predict who'll benefit from targeted therapy for HER2-negative breast cancer” o Press Release (EurekAlert!), September 17th, 2015 “Case-Coulter Translational Research Partnership awards $1 million in funding and support for promising biomedical engineering university technologies”, o Press Release, September 22nd, 2015. Madabhushi team awarded patent in digital pathology, biomarker quantification” o Case School of Engineering, August 24th, 2015. “5 questions with… professor of biomedical engineering Anant Madabhushi”, o The Daily, August 7th, 2015 “Faculty-student team participating in summer cohort of research commercialization program”, o The Daily, August 3rd, 2015. Press Coverage in 2015 “Ohio Department of Higher Education Announces Members of I-Corps@Ohio Summer Cohort”, Press Release, July 8th, 2015. “Madabhushi awarded NCI grant to study ductal carcinoma in situ” o Case Comprehensive Cancer Center News Letter, April 27th, 2015. “BME, CCIPD graduate student wins top honors at Research ShowCASE”, Case School of Engineering, May 30th, 2015. “Students honored with 2015 Intersections: SOURCE Symposium & Poster awards”, The Daily, May 1, 2015. Dr Madabhushi awarded patent for image based risk score”, o Case Comprehensive Cancer Center News Letter, April 13th, 2015. “CCIPD, BME Postdoctoral researcher awarded prestigious $125,000 fellowship from in prostate cancer from DOD”, o Case Comprehensive Cancer Center News Letter, March 30th, 2015. “Two Inspirata Strategic Advisory Board Members Inducted into the American Institute for Medical and Biological Engineering’s College of Fellows”, Press Release 2015. “Technical conference highlights: Medical Imaging 2015” o SPIE Newsroom, March 16th, 2015 “Anant Madabhushi named associate member of Quantitative Imaging Network, associate editor of BMC Medical Imaging” The Daily, March 13th, 2015 “Several Awards Received by Center for Computational Imaging and Personalized Diagnostics” o The Daily, April 24th, 2015. “Awards for Program Sponsored Project Grants from Case CCC Developmental Funds” o Case Comprehensive Cancer Center News Letter, March 2nd, 2015. “Medical image analysis technologies from Madabhushi team licensed to Boston-based startup” o Case School of Engineering, February 16th, 2015. “Madabhushi Named Associate Member of NCI Quantitative Imaging Network” o Case Comprehensive Cancer Center News Letter, February 16th, 2015. “BME’s Anant Madabhushi named AIMBE fellow” o The Daily, February 9th, 2015 “BME’s Anant Madabhushi to serve on editorial board of Journal of Medical Image Analysis” Conference Participation 2015 Mehdi Orooji, Mehdi Alilou, and Xiangxue Wang: I-Corps Workshop in Columbus, OH Anant Madabhushi: RSNA in Chicago, IL Ahmad Algohary: AUA in New Orleans, LA Conference Participation 2015 Mirabela Rusu & Greg Penzias: USCAP in Boston, MA Jacob Antunes: Research ShowCASE Anant Madabhushi: USCAP in Boston, MA Rakesh Shiradkar: Research ShowCASE Conference Participation 2015 Prateek Prasanna & Greg Penzias: Research ShowCASE Anant Madabhushi: SPIE in San Diego, CA Jay Patel: Research ShowCASE David Romo-Bucheli & Anant Madabhushi: SIPIAM in Cuenca, Ecuador Summary of Research Accomplishments Center Members: 34 Faculty: 4 Research Associates: 5 Graduate Students: 8 Undergraduate Students: 8 Scientific Software Engineers: 2 Administrative Assistants: 2 Visiting Scientists: 5 Theses: 0 Books: 1 Peer-Reviewed Journal Papers: 13 Peer-Reviewed Conference Papers: 12 Peer Reviewed Abstracts: 37 Awarded Grants: 9 Awarded Fellowships: 7 Ongoing Projects: 28 Issued Patents: 4 Invention Disclosures: 2 Technologies Licensed: 1 Peer Reviewed Publications for 2015 Summary 40 36 32 28 24 20 16 12 8 4 0 Books Journal Papers Conference Papers Abstracts Patents Invention Disclosures Patents in 2015 Issued Patents “Image-Based Risk Score – A Prognostic Predictor of Survival and Outcome from Digital Histopathology”, Anant Madabhushi, Ajay Basavanhally, Shridar Ganesan, United States Serial Number (USSN): 9,002,092, April 7th, 2015. "High Throughput Biomarker Segmentation Utilizing Hierarchical Normalized Cuts”, Anant Madabhushi, Andrew Janowczyk, Sharat Chandran, United States Serial Number (USSN): 9,111,179, August 19th, 2015. “Tumor Plus Adjacent Benign Signature (TABS) For Quantitative Histomorphometry” Anant Madabhushi, George Lee, Sahirzeeshan Ali, Publication number: US20150254493 A1, Sep 10, 2015. “Enhanced Multi-Protocol Analysis via Intelligent Supervised Embedding (EMPrAvISE) for Multimodal Data Fusion”, Anant Madabhushi, Satish Viswanath, United States Serial Number (USSN): 9,159,128, October 13th, 2015. New Grants Awarded in 2015 NIH 1U24CA199374-01 Pathology Image Informatics Platform for visualization, analysis and management NIH 1R21CA195152-01 Computerized morphologic-molecular predictor of progression in DCIS Case Comprehensive Cancer Center Pilot Grant Radiogenomic approach for predicting pathologic complete response in HER2+ breast cancers VeloSano Pilot Grant Genomic correlation of computer aided radiographic and histomorphometric feature extraction based models in Lung cancer and novel approach to predicting early resistance to targeted therapy. I-Corps (NSF) Decision support to identify benign infection on lung CT and reduce unnecessary surgical interventions Coulter Research Translational Partnership NeuroRadVisionTM: Image based risk score prediction of recurrent brain tumors (Phase 3) CDMRP Prostate Cancer Idea Grant Department of Defense Fused Biomarker-Based Prediction of Aggressive from Indolent Prostate Cancer Development Award in Biomedical Big Data Science (K01) NIH Big data convergence of pathology and omics for disease prognosi Research Portfolio Treatment Evaluation and Outcome Prediction Machine Learning Multimodal CoRegistration CCIPD Imaging Science Software Image Segmentation and Registration Multimodal Data Fusion Object Detection Computer Aided Diagnosis and Prognosis IMAGE REGISTRATION AND SEGMENTATION Multi-Attribute Probabilistic Prostate Elastic Registration (MAPPER)1 28R. Sparks, B. N. Bloch, E. Feleppa, D. Baratt, D. Mses, L. Ponsky, A. Madabhushi. Multiattribute probabilistic prostate elastic registration (MAPPER): Application to fusion of ultrasound and magnetic resonance imaging. Medical Physics 42(3): 1153-1163 (2015). A resolution adaptive deep hierarchical (RADHicaL) learning scheme applied to nuclear segmentation of pathology images Algorithm • • • Training images (blue paths) are resized to train M different AlexNet networks at different resolutions m. Test images (red path) are reduced to the smallest magnification, a threshold is applied to the classifier's probability mask. Only relevant pixels are mapped into next highest resolution for further analysis. Results Output Comparison of (f) our approach and (c) a naïve DL classifier on (a) original image with (b) ground truth. The blue channel, indicates the confidence in a pixel belonging to a nuclei. The red channel indicates the confidence in a particular pixel not belonging to a nuclei. Lastly, the green channel indicated the confidence that a pixel needs additional computation at the next level of magnification. At the (d) lowest resolution of 4x our algorithm is able to approximate the location of nuclei, obviating the need for the computation of the entire image at the next level as all of the red pixels can be avoided. Mapping to (e) 20x shows the blue pixels will be added to the final output, while green pixels merit additional computation at the next magnification level. Finally, when presenting (f) which is of the same magnification as Figure 6(c), the Performance metrics comparing RADHicaL versus the Naïve DL results are nearly identical, yet were produced 6x faster approach of computing all pixels. Timing, in seconds, shows a 6x Andrew Janowczyk1, Scott Doyle2, Hannah Gilmore3 , and Anant Madabhushi1 1 Case Western Reserve University, Dept. of Biomedical Engineering, Cleveland, OH, speed improvement for a typical image. 2 Pathology & Anatomical Sciences, SUNY Buffalo, NY 3 University Hospitals Case Medical Center, Cleveland OH Sparse Non-negative Matrix Factorization (SNMF) based Color Unmixing Xu J, et al. “Sparse Non-negative Matrix Factorization (SNMF) based Color Unmixing for Breast Histopathological Image Analysis”, Computerized Medical Imaging and Graphics, 2015. MACHINE LEARNING AND FEATURE ANALYSIS Classifier Error Rate Extrapolation via the Cross-Validated Inverse Power Law Method on Limited Training Data from the ADNI Dataset Abstract Citation: Agrawal, N, Basavanhally, A, Viswanath, S, Madabhushi, A, “Predicting Classifier Performance with Limited Training Data: Validation on the ADNI Dataset,” Biomedical Engineering Society (BMES), 2015. Accepted for Poster Presentation at Annual BMES Conference on Oct. 7-10, 2015 Background: • Clinical trials use machine learning classifiers to predict patient response to new therapies. • Need to predict classifier performance for large training datasets since classifier performance when trained with minimal data does not guarantee performance when exposed to larger amounts of training data. Application: • An estimated 16 million people will have Alzheimer’s Disease (AD) by 2050, signifying a need to develop and trial disease-treatment drugs. • Creation of Alzheimer's Disease Neuroimaging Initiative (ADNI) enabled development of classifiers to identify AD patients, however there is a lack of concordance in classifier methods used and generalizability of proposed classifiers is unclear. Objective: • Need to determine how associated parameters affect crossvalidated inverse power law (CVIPL) approach for extrapolating classifier error rates: number of subsampling trials, number of permutation trials, and number of cross-validation iterations • External validation of CVIPL method would be predicting performance of various classifiers on publicly available ADNI MRI data in distinguishing cognitively normal (CN) vs AD studies; compare extrapolated values against reported values from literature Results: • Increasing subsampling trials (T1) and permutation trials (T2) from 20 to 40 improved error rate extrapolation for both the Logistic Regression (LR) and Support Vector Machine (SVM) classifiers, as evident by decreased mIQR (compare plotted red and blue lines in Figures 3c). • Increasing cross-validation iterations (R) resulted in a decreased mIQR and hence improved error rate extrapolation for SVM, but an increased mIQR for the LR classifier. However, for both classifiers, more cross-validation iterations resulted in more stable mIQR values. • CVIPL methodology successfully validated for accurately extrapolating LR and SVM classifier performance against values reported in literature. Radiomics based characterization of early treatment response in renal cell carcinoma on FLT-PET/MRI Healthy Tissue Renal Cell Carcinoma Goal: Identify radiomic features and a combination thereof that best characterize early tumor response in metastatic renal cell carcinoma Results: SUV and ADC texture features ranked highest for capturing treatment response in comparison to high variability seen in T2w MRI. Antunes, J., Viswanath, S., Rusu, M., Valls L., Sher, A., Hoimes, C., Avril, N., and Madabhushi, A, “An integrated multi-parametric FLT-PET/MRI for evaluating early treatment response in renal cell carcinoma.” Presented at International Society for Magnetic Resonance in Medicine, 2015. Correlating Computer Extracted Features from Tubules on ER+ Breast Cancer Images with Oncotype DX Risk Categories Experimental Setup • 54 whole slide images (patients) were used on this study • The set was split in 39 low Oncotype DX patients (ODX score ranging 1 to 17) and 15 high scoring patients ( ODX score ranging 31 to 69) At most 50 high power fields per whole slide images were selected (Those with minimum fraction of tubule nuclei per histological sample). All the magnification fields are sampled from cancerous regions previously identified by an expert pathologist. • Results were correlated with Oncotype Dx (ODX). A genomic test commonly used to predict cancer aggressiveness Romo-Bucheli, D., Janowczyk, A., Romero E., Gilmore, H. and Madabhushi A.. “Automated Tubule Nuclei Quantification and Correlation with Oncotype DX risk categories in ER+ Breast Cancer Whole Slide Images”. Accepted SPIE Medical Imaging, 2016. Evaluating Stability of Histomorphometric Features across Scanner and Staining Variations: Predicting Biochemical Recurrence From Prostate Cancer Whole Slide Images • Tissue staining, slide preparation and scanning varies across institutions, resulting in varying feature distributions following quantitative histomorphometric analysis Accepted for presentation at SPIE Med Imag and USCAP 2016 Prostate gland lumen feature robustness Calculated a mean preparation-induced instability (PI) score. Lumen shape showed the lowest instability Features dependent on gland lumen architecture showed high volatility via PI Accepted for presentation at SPIE Med Imag 2016 and USCAP 2016 Quantifying grades of cerebral radiation necrosis 1 60 0.95 50 CoLlAGe entropy MR Signal Intensity 0.9 0.85 0.8 0.75 0.7 0. 5 40 0. 3 30 20 0. 1 10 0.65 0.6 Pure Necrosis Mixed necrosis Mixed Tumor Pure Tumor 100% CRN >80% CRN <20% CRN 0% CRN 0 Pure Necrosis Mixed necrosis Mixed Tumor Pure Tumor 100% CRN >80% CRN <20% CRN 0% CRN --> Distinguishing radiation necrosis from recurrent tumors using both pure and predominant signatures from cerebral radiation necrosis --> Dataset: 100% CRN (from head and neck cancer), mixed CRN, 0% CRN (treatment naive GBM from TCIA) Prasanna et. al. , SNO 2015 Stacked Sparse Autoencoder for Nuclei Detection Xu J, et al. “Stacked Sparse Autoencoder (SSAE) based Framework for Nuclei Patch Classification on Breast Cancer Histopathology”,ISBI2014. Xu J, et al. “Stacked Sparse Autoencoder (SSAE) for Nuclei Detection on Breast Cancer Histopathology”. IEEE Trans. on Medical Imaging, 2015 Zhang X, Dou H, Xu J, Zhang S, “Fusing Heterogeneous Features for the Image-Guided Diagnosis of Intraductal Breast Lesions”, IEEE Journal of Biomedical and Health Informatics, 2015 Lu C, Xu H, Xu J, Mandal M, and Madabhushi A, “Multiple Passes Adaptive Voting for Nuclei Detection in Histopathlogical Images“, IEEE Journal of Biomedical and Health Informatics, (Under Preparing) Automated Discrimination of Epithelial and Stroma Regions Jun Xu, et al."A Deep Convolutional Neural Networks based Feature Learning for Detecting and Classifying Epithelial and Stroma Regions in Histopathological Images" Neurocomputing, (Accepted pending changes) MULTI-MODAL DATA FUSION Multi-modal co-registration via multi-scale textural and spectral embedding representations Objective: Apply image-based textural and spectral embedding representations in coregistration of multimodality mages and compare their performance against image intensity alone. Lin Li, Mirabela Rusu, Satish Viswanath, Anant Madabhushi, accepted for poster presentation for BMES 2015 and SPIE 2016. TREATMENT EVALUATION AND OUTCOME PREDICTION Biochemical Recurrence Prediction via Tumor plus Adjacent Benign Signature • Improvement in classification of BCR following combination of features extracted from tumor-field and adjacent benign field (TABS) • Identify signatures from benign fields of view in radical prostatectomy samples for predicting biochemical recurrence Presented at USCAP and AUA 2015 Prostate Biochemical Recurrence Prediction via H&E and Feulgen • Combination shows better prediction • H&E and Feulgen features show low correlation • 30% of radical prostatectomy cases undergo biochemical recurrence • H&E describes nuclear and gland architecture and shape • Feulgen describes DNA ploidy Presented at BMES 2015 and NIH Big Data to Knowledge All Hands Meeting 2015 Radiomic features on T2w MRI to predict tumor invasiveness for pre-operative planning in colorectal cancer Goal: Identify radiomic features predictive of tumor invasiveness on a restaging MRI in colorectal cancer. Results: Gabor features were found to be most discriminatory, by capturing microarchitectural oriented gradient differences between patients with positive and negative circumferential margins (CRM) (f) (a) Projection of top 3 PCA features into 3D space for 10 patients: blue circles and red squares represent positive and negative CRM patients, respectively, clustered and separated by ellipses. (b)(d) Original T2w images with annotation for a single patient from each group (top: positive CRM; bottom: negative CRM). (c)(e) heatmaps of topranked Gabor feature overlaid on the original T2w images. Blue indicates low feature expression, whereas red indicates high feature expression. (f) Boxplots of normalized feature intensities for top 3 features between positive CRM and negative CRM patients. The difference in intensity distributions between the two classes indicates that Gabor features appear to express different information between invasive and noninvasive colorectal cancer. Jacob Antunes , Scott Steele, Conor Delaney, Joseph Willis, Justin Brady, Rajmohan Paspulati, Anant Madabhushi, Satish Viswanath. “Radiomic features on T2w MRI to predict tumor invasiveness for pre-operative planning in colorectal cancer.” Abstract pending submission for International Society for Magnetic Resonance in Medicine, 2016. Feature Importance in Nonlinear Embeddings (FINE) Identifies Histo-morphometric Features that Predict Cancer Outcome Following Treatment • FINE identifies glandular architecture as important for predicting the risk of biochemical recurrence of prostate cancer • FINE identifies nuclear area as an excellent predictor OncotypeDX risk scores for breast cancer recurrence S Ginsburg, et al. Feature Importance in Nonlinear Embeddings (FINE): Applications in Digital Pathology. IEEE TMI available online. Predicting Active Surveillance candidates from Prostate cancer biopsies • Very low risk (VLR) prostate cancer - GS 6 - low PSA - low tumor stage • Figure 2: A combined classifier consisting of statistically significant (p < 0.05) features demonstrated 0.78 AUC • Some in cohort showed • unfavorable outcomes - PSA rising - Upgrade in Gleason score Presented at USCAP and AUA 2015 Radiomic Markers Across Tumor Habitat on Treatment-Naïve MRI can Predict Survival in GBM Patients Distinguishing STS from LTS using Textural Features (a) (d) (b) (e) Necrosis Top 10 Tumor Top 10 Edema Top 10 All Top 10 (c) (f) Gd-T1 MRI for STS (a) and LTS (d). Compartmental segmentations for necrosis, tumor, edema (b, e). Qualitative per-voxel representation of Haralick Entropy feature (c, f). Tiwari et al., RSNA (2015) Histomorphometric features for predicting recurrence in early stage lung cancers AUC = 0.85 (a) (c) (e) Figure 3 Figure 2 (b) (d) (f) Figure 1 Original images of (a) recurrence and (b) non-recurrence patients, haralick, intensity correlation texture, feature image of (c) recurrence and (d) non-recurrence patients, cell cluster graph feature images of (e) recurrence and (f) non-recurrence patients. Training Validation Total Number Positive 44 51 95 Negative 40 66 106 Table1 Total Number 84 117 201 Characteristic Gender Male Female T Pathological T1/T2 T3/T4 N Pathological N0/N1 N2/N3 Major clinical and pathological features by classif result Entire Vali Set = Classi Non-Recur = Classi Recur = 117, 45, 72, N(%) N(%) N(%) P 60(51.3) 57(48.7) 27(60.0) 18(40.0) 33(45.8) 39(54.2) 0.1832 71(60.7) 46(39.3) 31(68.9) 14(31.1) 40(55.6) 32(44.4) 0.176 77(65.8) 40(34.2) 31(68.9) 14(31.1) 46(63.9) 26(36.1) 0.689 Table2 Table 1. Dataset summary; Table 2. Clinical and pathological features on entire validation set and by classification of 12 hitomorphometric features. No Features show significant correlation with classification result. Figure 2. ROC plot on Validation set Figure 3. Kaplan-Meier curve by classification results show good separation and log-rank test show statistically different Xiangxue Wang, Andrew Janowczyk, Yu Zhou, Sagar Rakshit, Vamsidhar Velcheti, Anant Madabhushi*, “Computer extracted features of nuclear morphology from digital H&E images are predictive of recurrence in stage I and stage II non-small cell lung cancer” accepted by USCAP 2016 Annual Meeting COMPUTER AIDED DIAGNOSIS AND PROGNOSIS Breast Cancer Recurrence Risk Categories and Automatic Mitotic Detection Algorithms Based on Deep Learning Experimental Setup - 54 whole slide images (patients) were used on this study - The set was split in 39 low Oncotype DX patients (ODX score ranging 1 to 17) and 15 high scoring patients ( ODX score ranging 31 to 69) - Results were correlated with Oncotype Dx (ODX). A genomic test commonly used to predict cancer aggressiveness Results The mitotic information of each case is organized on a histogram of mitotic counts. Then, each histogram is treated as a feature vector to train a SVM classifier. There are 54 feature vectors. The data points are centered at their mean, and scaled to have unit standard deviation, before training.Using a leave one out scheme, a 87% accuracy on the HH-LL dataset was obtained Low ODX and mBR score High ODX and mBR score Classified as Low Risk 36 4 Classified as High Risk 3 11 Romo-Bucheli, D., Janowczyk, A., Romero E., Gilmore, H. and Madabhushi A.. “Prediction of Breast Cancer Recurrence Risk Categories Using Automatic Mitotic Detection Algorithms Based on Deep Learning”. Accepted USCAP meeting, 2016. A Novel Computer-Assisted Approach for Prostate Cancer Diagnosis on T2W MRI • The approach well solves the crucial problem of extracting computerized image features at multiple scales from T2w MRI for prostate cancer diagnosis. • It suggests a higher accuracy and lower computational time compared to using T2w MRI intensity and exhaustive multiscale features. Haibo Wang, Satish Viswanath, Asha Singanamalli, and Anant Madabhushi, A Novel Computer-Assisted Approach for Prostate Cancer Diagnosis on T2w MRI, ISMRM 2015, Toronto, Ontario, Canada. Identifying in vivo DCE MRI Markers of Prostate Cancer Risk • • Examined correlations between DCE MRI kinetic features and pathological prognostic markers, Gleason scores and microvessel architecture. Washout gradient and enhancement ratio were found to be correlated with microvessel architectural features (max rho = -0.62, -0.52), and were predictive of low Gleason scores (AUC = 0.77, 0.78) Singanamalli, A., Rusu, M., Sparks, RE., Shih, NC., Ziober, A, Wang L, Tomaszewski, J., Rosen, M., Feldman, M., and Madabhushi, A., “Identifying in vivo DCE MRI Markers Associated with Microvessel Architecture and Gleason Grades of Prostate Cancer.” Journal of Magnetic Resonance Imaging 2015 Computer-extracted features can distinguish noncancerous confounding disease from prostatic adenocarcinoma at multiparametric MR imaging. Pathologic-MR imaging mapping procedure in two patients. (a) MR image (left), pathologic image (center), and MR image overlaid with pathologic image (right) show prostate cancer (yellow line). The large lesion has a Gleason score of 3 + 4, and the other two lesions have a Gleason score of 3 + 3. Inflammation (green line), PIN (blue line), and atrophy (orange line) are also seen. Geert Litjens, Robin Elliott, Natalie Shih, Michael Feldman, Thiele Kobus, Christina Hulsbergen – van de Kaa, Jelle Barentsz, Henkjan Huisman and Anant Madabhushi. Radiology, 2015 Feature maps of the top three selected features for atrophy, BPH, PIN, and inflammation (cf Table 6) show cancer, with low, intermediate, and high grades grouped together (red line), and the specific benign class (yellow line). The axial T2-weighted image is provided as a reference (left-most column). The selected features provide a good contrast between cancer and the specific benign class. Computer Extracted Texture Features on Multi-parametric MRI Differentiate Risk Categories of Prostate Cancer Patients on Active Surveillance PCa-positive region annotated by expert radiologist Cases identified by Exp QDA heatmap for PIRADS-positive, biopsy-positive case indicating correctly detected tumor (red outline) • 10 Radiomic features used to create probability maps; evaluated for cancer presence • Experiment 1: Radiomic features on par with PIRADS in identifying clinically significant disease. • Experiment 2: Radiomic features better than PIRADS in identifying absence of disease 1 2 Type Total PIRADS QDA maps 1,2 4,5 - + Biopsy + (≥ 3+4) 19 4 15 0 19 Biopsy + (≤ 3+3) 13 12 1 9 4 Biopsy - 10 0 10 9 1 Comparing PIRADS assessment to featurebased QDA map assessment • Radiomic features PPV = 0.76, NPV = 0.93; 52% higher than PIRADS. Ahmad Algohary, Satish Viswanath, Prateek Prasanna, Shivani Pahwa, Vikas Gulani, Daniel Moses, Ronald Shnier, Maret Böhm, Anne-Maree Haynes, Phillip Brenner, Warick Delprado, James Thompson, Marley Pulbrock, Phillip Stricker, Lee Ponsky, Anant Madabhushi. AUA 2015 SOFTWARE AutoStitcherTM: An automated program for accurate reconstruction of digitized whole histological sections from tissue fragments • Stitch using domain constraints a. b. c. Non-Overlap Region-Matching Non-Overhang Stitched High Resolution Result Low Resolution Initialization Oral Presentation, USCAP 2015, Boston MA Gregory Penzias, Andrew Janowczyk, Asha Singanamalli, Mirabela Rusu, Natalie Shih, Michael Feldman, Satish Viswanath and Anant Madabhushi HistoView A robust, user-friendly application for pathology image analysis and computer-aided annotation. • HistoView supports whole-slide images and enables high speed calculation on pyramid image. • HistoView provides a variety of automatic segmentation methods based on existing famous algorithms: color deconvolution and CCIPD’s literature: HNCuts, Deep Learning segmentation, etc. • HistoView allows the pathologist to make accurate and rapid annotation based on the segmentation result from different algorithms with classical painting tools. Windows Caffe platform • Windows Caffe platform enables the user, in an all-in-one solution, to train DL classifiers and generating segmentation result using Caffe framework. • Training includes four steps: patch extraction, cross-validation creation, database creation and training for DL classifier. • Testing provides three pre-trained classifiers: nuclear segmentation, mitosis detection and epitheliumstroma segmentation. • Windows Caffe platform is continuously maintained with new pre-trained models being added regularly to the mdoel zoo. RadCAD Ver. 3 Emerged from the inhouse developed software ProstaCAD ProstaCAD is a Tumor Detection Software for T2 Images developed at CCIPD. New MRI protocols are now supported in addition to T2w Now 4D volume images from Dynamic Contrast Enhancement is supported. New Tumor classification Radiomic features were added (e.g. Coliage) Peer Reviewed Publications for 2015 Books Gurcan, M, Madabhushi, A, Medical Imaging 2015: Digital Pathology (Proceedings Volume), Proceedings of International Society of Optics and Photonics (SPIE) Volume: 942001; doi:10.1117/12.2193831 Journal Papers Bhargava, R and Madabhushi, A, “Emerging Themes in Image Informatics and Molecular Analysis for Digital Pathology”, Annual Reviews of Biomedical Engineering 2016, Accepted. Tiwari, P, Danish, S, Madabhushi, A, “Computerized texture features on MRI capture early treatment response following laser ablation for neuropathic cancer pain: Preliminary findings”, Journal of Medical Imaging, In Press. Varadan, V, Kamalakaran, S, Gilmore, H, Banerjee, N, Janevski, A, Miskimen, K, Williams, N, Basavanhally, A, Madabhushi, A, Lezon-Geyda, K, Bossuyt, V, Lannin, D, Abu-Khalaf, M, Sikov, W, Dimitrova, N, Harris, L, “Brief-exposure to preoperative bevacizumab reveals a TGF-β signature predictive of response in HER2-negative breast cancers”, Int J Cancer. 2015 Aug 18. doi: 10.1002/ijc.29808. [Epub ahead of print] (PMID: 26284485). Xu, J, Xiang, L, Liu, Q, Gilmore, H, Wu, J, Tang, J, Madabhushi, A, “Stacked Sparse Autoencoder (SSAE) for Nuclei Detection on Breast Cancer Histopathology images”, IEEE Transactions on Medical Imaging, 2015 Jul 20. [Epub ahead of print] (PMID: 26208307). Rusu, M, Wang, H, Golden, Thea, Gow, A, Madabhushi, A, “Framework for 3D histologic reconstruction and fusion with in vivo MRI: Preliminary results of characterizing pulmonary inflammation in a mouse model”, Medical Physics, 42(8):4822. doi: 10.1118/1.4923161., (PMID: 26233209). Peer Reviewed Publications for 2015 Journal Papers (Contd.) Ginsburg, S, Lee, G, Ali, S, Madabhushi, A, Feature Importance in Nonlinear Embeddings (FINE): Applications in Digital Pathology, IEEE Transactions on Medical Imaging, 2015 Jul 14. [Epub ahead of print] PMID: 26186772). Litjens, G, Huisman, H, Elliot, R, Shih, N, Feldman, M, Viswanath, S, Bomers, J, Madabhushi, A, “Computer-extracted features can distinguish benign confounding disease from prostatic adenocarcinoma on multi-parametric MRI”, Radiology, 2015 Jul 17:142856. [Epub ahead of print] (PMID: 26192734). Shridar, A, Doyle, S, Madabhushi, A, “Content-Based Image Retrieval of Digitized Histopathology in Boosted Spectrally Embedded Spaces”, Journal of Pathology Informatics, Jun 29;6:41. doi: 10.4103/2153-3539.159441. eCollection 2015 (PMID: 26167385). Singanamalli, A, Sparks, R, Rusu, M, Shih, N, Ziober, A, Tomaszewski, J, Rosen, M, Feldman, M, Madabhushi, A, “Identifying in vivo DCE MRI markers associated with Microvessel Architecture and Gleason Grades of Prostate Cancer—Preliminary Findings”, Journal of Magnetic Resonance Imaging, doi: 10.1002/jmri.24975. [Epub ahead of print] (PMID:26110513). Basavanhally, A, Viswanath, S, Madabhushi, A, “Predicting Classifier Performance with Limited Training Data: Applications to Computer-Aided Diagnosis in Breast and Prostate Cancer”, PLOS One, 2015 May 18;10(5):e0117900. doi: 10.1371/journal.pone.0117900. eCollection 2015. (PMID:25993029). Xu, J, Xiang, L, Wang, G, Ganesan, S, Feldman, M, Shih, N, Gilmore, H, Madabhushi, A, “Sparse Non-negative Matrix Factorization (SNMF) based Color Unmixing for Breast Histopathological Image Analysis”, Computerized Medical Imaging and Graphics (CMIG), pii: S0895-6111(15)00077-4. doi: 10.1016/j.compmedimag.2015.04.002. [Epub ahead of print] (PMID: 25958195). Peer Reviewed Publications for 2015 Journal Papers (Contd.) Sparks, R, Bloch, N, Moses, D, Ponsky, L, Barratt, D, Feleppa, E, Madabhushi, A, “Multi-Attribute Probabilistic Prostate Elastic Registration (MAPPER): Application to Fusion of Ultrasound and Magnetic Resonance Imaging”, Medical Physics, Mar;42(3):1153-63. doi: 10.1118/1.4905104. (PMID:25735270) Hoimes, C, Madabhushi A, “Editorial Comment on Biparametric Prostate MRI in Prostate Cancer Detection”, Journal of Urology, In Press Peer Reviewed Publications for 2015 Peer-reviewed Conference Papers Cruz-Roa, A, Arevalo, J, Judkins, A, Madabhushi, A, Gonzalez, F, “A method for Medulloblastoma Tumor Differentiation based on Convolutional Neural Networks and Transfer Learning, Proc. SPIE, 11th International Symposium on Medical Information Processing and Analysis, Accepted. Janowczyk, A, Doyle, S, Gilmore, H, Madabhushi, A, “A resolution adaptive hierarchical deep learning scheme applied to nuclear segmentation in histology images”, Workshop on Deep Learning, Medical Image Computing and Computer-Assisted Intervention (MICCAI) 2015, Accepted. Otálora, S, Cruz-Roa, A, Arevalo, J, Atzori, M, Madabhushi, A, Judkins, AR, González, F, Müller, H, Depeursinge, A. “Anaplastic Medulloblastoma tumor differentiation by combining Unsupervised Feature Learning and Riesz Wavelets for histopathology image representation”, Medical Image Computing and Computer-Assisted Intervention (MICCAI) 2015, pp. 581-588. Cruz Roa, A, Xu, J, Madabhushi, A,“ A note on the stability and discriminability of graph-based features for classification problems in digital pathology”, Proc. SPIE 9287, 10th International Symposium on Medical Information Processing and Analysis, 928703 (January 28, 2015); doi:10.1117/12.2085141. Cruz Roa, A, Osorio, FA, Madabhushi, A, Gonzalez, F, “A comparative evaluation of supervised and unsupervised representation learning approaches for anaplastic medulloblastoma differentiation”, In 10th International Symposium on Medical Information Processing and Analysis, Accepted. Li, L, Rusu, M, Viswanath, S, Madabhushi, A, Multi-modal co-registration via multi-scale textural and spectral embedding representations, The International Society for Optics and Photonics (SPIE) Medical Imaging, 2015. Romo-Bucheli, DE, Janowczyk, A, Romero, E, Madabhushi, A, Automated tubule nuclei quantification on ER+ breast cancer images: Comparison with Oncotype DX risk categories, The International Society for Optics and Photonics (SPIE) Medical Imaging, 2015. Peer Reviewed Publications for 2015 Peer-reviewed Conference Papers (Contd.) Leo, P, Lee, G, Madabhushi, A, Evaluating stability of histomorphometric features across scanner and staining variations: predicting biochemical recurrence from prostate cancer whole slide images, The International Society for Optics and Photonics (SPIE) Medical Imaging, 2015. Singanamalli, A, Pisipati, S., Ali, A., Wang, V., Tang, C.T., Taouli, B., Tewari, A., and Madabhushi, A, "Correlating gland orientation patterns on ex vivo 7 Tesla MRI with corresponding histology for prostate cancer diagnosis: Preliminary results," The International Society for Optics and Photonics (SPIE) Medical Imaging, 2015. Romo-Bucheli, DE, Janowczyk, A, Romero, E, Madabhushi, A, Automated tubule nuclei quantification on ER+ breast cancer images: Comparison with Oncotype DX risk categories, The International Society for Optics and Photonics (SPIE) Medical Imaging, 2015. Leo, P, Lee, G, Madabhushi, A, Evaluating stability of histomorphometric features across scanner and staining variations: predicting biochemical recurrence from prostate cancer whole slide images, The International Society for Optics and Photonics (SPIE) Medical Imaging, 2015. Singanamalli, A, Pisipati, S., Ali, A., Wang, V., Tang, C.T., Taouli, B., Tewari, A., and Madabhushi, A, "Correlating gland orientation patterns on ex vivo 7 Tesla MRI with corresponding histology for prostate cancer diagnosis: Preliminary results," The International Society for Optics and Photonics (SPIE) Medical Imaging, 2015. Peer Reviewed Publications for 2015 Peer-reviewed Abstracts Wang, X, Janowczyk, A, Sagar, R, Velcheti, V, Madabhushi, A, “Computer Extracted Features of Nuclear Morphology from Digital H&E Images Are Predictive of Recurrence in Stage I and II NonSmall Cell Lung Cancer”, United States and Canadian Academy of Pathology's 105th Annual Meeting, Salt Lake City, UT, March 12-18, 2016. Romo, D, Janowczyk, A, Gilmore, H, Romero, E, Madabhushi, A, “Prediction of Breast Cancer Recurrence Risk Categories Using Automatic Mitotic Detection Algorithms Based on Deep Learning”, United States and Canadian Academy of Pathology's 105th Annual Meeting, Salt Lake City, UT, March 12-18, 2016. Janowczyk, A, Basavanhally, A, Madabhushi, A, “Stain Normalization in Digital Pathology Images Using Deep Learning”, United States and Canadian Academy of Pathology's 105th Annual Meeting, Salt Lake City, UT, March 12-18, 2016. Janowczyk, A, Doyle, S, Gilmore, H, Madabhushi, A, “Fully Automated, Accurate, and Efficient Segmentation of Cancer Nuclei in Breast Pathology Images”, United States and Canadian Academy of Pathology's 105th Annual Meeting, Salt Lake City, UT, March 12-18, 2016. Leo, P, Lee, G, Madabhushi, A, Evaluating Reproducibility of Computer Extracted Histologic Image Features for Predicting Biochemical Recurrence in Prostate Cancer: A Multi-Site, Multi-Scanner Study, United States and Canadian Academy of Pathology's 105th Annual Meeting, Salt Lake City, UT, March 12-18, 2016. Ali, S, Rimm, D, Ganesan, D, Madabhushi, A, “Local nuclear architecture features from H&E images predict early versus distant recurrence in lymph node negative, ER+ breast cancers”, 2015 San Antonio Breast Cancer Symposium, December 8-12, 2015 in San Antonio, Texas, (Poster). Tiwari, P, Prasanna, P, Patel, J, Madabhushi, A, “Computer extracted texture descriptors from different tissue compartments within the tumor habitat on treatment-naïve MRI predict clinical survival in glioblastoma patients”, Society of Neurooncology, 2015, Accepted. Peer Reviewed Publications for 2015 Peer-reviewed Abstracts (Contd.) Prasanna, P, Siddalingappa, A, Wolansky, L, Rogers, L, Tai-Chung Lam, V, Madabhushi, A, Tiwari, P, “Morphologic heterogeneity at a pixel-level captured via entropy of gradient orientations on T1-post contrast MRI enables discrimination of tumor recurrence from cerebral radiation necrosis”, Society of Neurooncology, 2015, Accepted. Tiwari, P, Patel, J, Partovi, S, Prasanna, P, Madabhushi, A, “Computer extracted texture descriptors from different tissue compartments within the tumor habitat on treatment-naïve MRI predict clinical survival in glioblastoma patients”, Proceedings of the Radiologic Society of North America, 2015, Accepted. Rusu, M, Orooji, M, Gilkeson, R, Prabhakar, R, Yang, M, Jacono, F, Linden, P, Madabhushi, A, “Accurate co-registration of ex vivo histology and in vivo CT in ground glass nodules enables the identification of computer extracted textural features to predict extent of invasion”, Proceedings of the Radiologic Society of North America, 2015, Accepted. Li, L, Rusu, M, Madabhushi, A, “Texture based similarity measure for multi-modal co-registration”, Biomedical Engineering Society (BMES), 2015. Agrawal, N, Basavanhally, A, Viswanath, S, Madabhushi, A, “Predicting Classifier Performance with Limited Training Data: Validation on the ADNI Dataset”, Biomedical Engineering Society (BMES), 2015. Gawlik, A, Lee, G, Whitney, J, Epstein, J, Veltri, R, Madabhushi, A, “Computer Extracted Nuclear Features from Feulgen and H&E Images Predict Prostate Cancer Outcomes”, Biomedical Engineering Society (BMES), 2015. Golden, T, Massa CB, Rusu M, Wang H, Madabhushi, A, Gow AJ, Structural and Functional Modeling Of Chronic Lung Inflammation: Loss of Function Mechanisms, Experimental Biology 2015 Viswanath S, Crawshaw, B, Willis J, Delaney C, Paspulati R and Madabhushi, A “Computational Radiology-Pathology Fusion for Improved Treatment Planning and Patient Selection in Rectal Cancer” 3rd World Rectal Conference on Organ preserving Perspectives, Montreal, Canada, 2015. Peer Reviewed Publications for 2015 Peer-reviewed Abstracts (Contd.) Wang, H, Viswanath, S, Singanamalli, A, Madabhushi, A, A Novel Computer-Assisted Approach for Prostate Cancer Diagnosis on T2w MRI, Annual Meeting of the International Society for Magnetic Resonance in Medicine (ISMRM), Toronto, Canada, 2015. Prasanna, P, Tiwari, P, Siddalingappa, A, Wolansky, L, Rogers, L, Lam, TC, To, V, and Madabhushi, A, Study of contrast-enhanced T1-w MRI markers of cerebral radiation necrosis manifested in head-andneck cancers, primary, and metastatic brain tumors: Preliminary findings, Annual Meeting of the International Society for Magnetic Resonance in Medicine (ISMRM), Toronto, Canada, 2015. Antunes, J, Viswanath, S, Rusu, M, Vallis, L, Avril, N, Hoimes, C, Madabhushi, A, Quantitative Analysis of Multi-parametric FLT-PET/MRI in Evaluating Early Treatment Response in Renal Cell Carcinoma, Annual Meeting of the International Society for Magnetic Resonance in Medicine (ISMRM), Toronto, Canada, 2015. Viswanath, S, Yim, CC, Bloch, NB, Rosen, M, Kurhanewicz, J, Madabhushi, A, A multi-site study to develop a new pseudo-quantitative T2w MRI map for prostate cancer characterization: Preliminary findings, Annual Meeting of the International Society for Magnetic Resonance in Medicine (ISMRM), Toronto, Canada, 2015. Algohary, A, Viswanath, S, Prasanna, P, Pahwa, S, Gulani, V, Ponsky, L, Stricker, P, Moses, D, Shnier, R, Madabhushi, A, Quantitative assessment of T2-weighted MRI to better identify patients with prostate cancer in a screening population, American Urologic Association (AUA), 2015. Lee, G, Veltri, R, Ali, S, Epstein, J, Christudass, C, Madabhushi, A, Prostate cancer recurrence can be predicted by measuring nuclear organization and shape parameters in adjacent benign regions on radical prostatectomy specimens, American Urologic Association (AUA), 2015. Lee, G, Veltri, R, Zhu, G, Carter, B, Landis, P, Epstein, J, Madabhushi, A, Quantitative Histomorphometric Analysis of Prostate Biopsy Images Predict Favorable Outcome in Active Surveillance Patients, American Urologic Association (AUA), 2015. Peer Reviewed Publications for 2015 Peer-reviewed Abstracts (Contd.) Zhu, G, Lee, G, Davis, C, Kagohara, LT, Epstein, JI, Landis, P, Carter, BH, Madabhushi, A, Veltri, R. Prediction of favorable and unfavorable biopsy pathology results of active surveillance patients using nuclear morphometry and molecular biomarkers, American Association for Cancer Research, 2015. Veltri, R, Ali, S, Lin, W-C, Zhu, G, Epstein, JI, Li, C-C, Madabhushi, A, Cancer Histologic and Cell Nucleus Architecture differentiate Prostate Cancer Gleason Patterns, American Association for Cancer Research, 2015. Ali S, Basavanhally, A, Madabhushi, A, “Histogram of Hosoya Indices for Assessing Similarity Across Subgraph Populations: Breast Cancer Prognosis Prediction From Digital Pathology” United States and Canadian Academy of Pathology's 104th Annual Meeting, Boston, MA, March 21-27, 2015. Ali S, Lewis J, and Madabhushi, A “A Quantitative Histomorphometric Classifier Identifies Role of Stromal and Epithelial Features in Prediction of Disease Recurrence in p16+ Oropharyngeal Squamous Cell Carcinoma” United States and Canadian Academy of Pathology's 104th Annual Meeting, Boston, MA, March 21-27, 2015. Cruz-Roa A, Basavanhally A, Gonzalez F, Feldman M, Ganesan S, Shih N, Tomaszewsky J, Gilmore H, Madabhushi, A “A Feature Learning Framework for Reproducible Invasive Tumor Detection of Breast Cancer in Whole-Slide Images” United States and Canadian Academy of Pathology's 104th Annual Meeting, Boston, MA, March 21-27, 2015. Lee G, Veltri, R, Zhu G, Epstein J, and Madabhushi, A “Computerized Nuclear Shape Analysis of Prostate Biopsy Images Predict Favorable Outcome in Active Surveillance Patients” United States and Canadian Academy of Pathology's 104th Annual Meeting, Boston, MA, March 21-27, 2015. Lee G, Veltri R, Ali S, Epstein J, Christudass C, and Madabhushi, A “Prostate cancer recurrence can be predicted by measuring cell graph and nuclear shape parameters in the benign cancer-adjacent field of surgical specimens” United States and Canadian Academy of Pathology's 104th Annual Meeting, Boston, MA, March 21-27, 2015. Peer Reviewed Publications for 2015 Peer-reviewed Abstracts (Contd.) Penzias G, Janowczyk A, Singanamalli A, Rusu M, Shih N, Feldman M, Viswanath S and Madabhushi, A “AutoStitcher©: An Automated Program for Accurate Reconstruction of Digitized Whole Histological Sections From Tissue Fragments” United States and Canadian Academy of Pathology's 104th Annual Meeting, Boston, MA, March 21-27, 2015. Rusu M, Yang M, Rajiah P, Jacono F, Gilkeson R, Linden P, and Madabhushi, A “Histology – CT Fusion Facilitates the Characterization of Suspicious Lung Lesions with No, Minimal, and Significant Invasion on CT” United States and Canadian Academy of Pathology's 104th Annual Meeting, Boston, MA, March 21-27, 2015. Viswanath S, Paspulati R, Delaney C, Willis J, and Madabhushi, A “A Novel Pathology-Radiology Fusion Workflow for Predicting Treatment Response and Patient Outcome in Rectal Cancers” United States and Canadian Academy of Pathology's 104th Annual Meeting, Boston, MA, March 21-27, 2015. Non-Peer Reviewed Publications for 2015 Non-Peer Reviewed Abstracts Prasanna P, Tiwari P, Madabhushi A. Co-occurrence of Local Anisotropic Gradient Orientations (CoLIAGe): Novel radiomic feature descriptors for brain and breast cancer characterization on MRI, Data and Life Science Symposium, Case Western Reserve University, 2015 Romo-Bucheli, D., Janowczyk, A., Romero, E., Gilmore, H., Madabhushi, A. Correlating Computer Extracted Features from Tubules on ER+ Breast Cancer Images with Oncotype DX Risk Categories, Data and Life Science Symposium, Case Western Reserve University, 2015 Orooji M, Alilou M, Rajiah P, Velcheti V, Yang M, Jacono F, Gilkeson R, Linden P, Madabhushi, A. A Combination of Shape and Texture Features Enables Discrimination of Benign Fungal Infection from Non Small Cell Lung Cancer on Chest CT, Data and Life Science Symposium, Case Western Reserve University, 2015 Wang X, Janowczyk A., Velcheti V., Madabhushi, A. Computer extracted Nuclear features predict recurrence in stage I, stage II non-small cell lung cancer, Data and Life Science Symposium, Case Western Reserve University, 2015 Madabhushi, A, Viswanath, S, Tiwari, P, Rusu, M, Singanamalli, A, Orooji, M, Algohary, A, Lee, G, Janowyzk, A, Ali, S, Ginsberg, S, Prasanna, P, Penzias, G, "Center for Computational Imaging and Personalized Diagnostics", Research ShowCASE, April, 2015. INTERESTED IN JOINING CCIPD? We are always looking for enthusiastic and motivated graduate, undergraduate students, postdoctoral and research scientists. If you think you would be a good fit for CCIPD, send over your complete CV and 3 representative publications to “anantm” @ “case.edu” Follow us on Twitter: @CCIPD_Case