Pradeep Redddy Raamana Research Scientist at Simon Fraser University pkr1@sfu.ca Summary Highlights 8 years of experience in machine learning, data mining and statistical modelling. 6 years of experience in applying machine learning in large-scale HPC environment. 6 years of experience in medical image processing, segmentation and analysis. 2 years of experience in object detection, activity recognition and computer vision. 5 years of leadership, mentoring, management and collaboration experience. Publications: 4 first-author peer-reviewed journal publications from PhD alone, in venues such as Neurobiology of Aging, Neuroimage Clinical and Frontiers in Neurology. Total journals: 6. Areas of expertise: medical image analysis, machine learning, pattern recognition, predictive analytics, statistical analysis, computer-aided diagnosis techniques, anatomic structure segmentation, development of imaging biomarkers, shape representation, and automation of large-scale processing. Thesis focus: Development of computer-aided diagnostic and prognostic tools for various neurodegenerative diseases using machine learning techniques and medical image analysis. Experience Research Scientist at Simon Fraser University September 2014 - Present (6 months) Independent investigation, development and analysis of 1) medical image processing and analysis methods 2) accurate and robust predictive models of neurodegeneration 3) automated methods for big-data analysis Project management: supervision and advising of graduate students. Grant writing: developing independent grants as well as assisting my supervisor with grant-writing. Linux Admin at Medical Image Analysis Laboratory, Simon Fraser University May 2011 - Present (3 years 10 months) - First person of contact for Linux-related problem-solving. - Troubleshooting, administering and helping users for issues with Linux and file systems. - Proficient in Linux, bash scripting and automating various simple tasks. Ph. D Student at Simon Fraser University May 2008 - August 2014 (6 years 4 months) Biomedical Engineering at School of Engineering Science Research Assistant at Simon Fraser University April 2008 - August 2014 (6 years 5 months) Page1
Developed machine learning techniques for early detection of Alzheimer s disease, to deliver state-ofthe-art accuracy of 93% using a single magnetic resonance imaging (MRI) scan. Designed shape features and feature extraction techniques for computer-assisted diagnostic systems. Optimized multi-atlas fusion techniques for accurate segmentation of subcortical structures in the brain from T1 MR images. Conceived classifier fusion, and multiple kernel learning techniques delivering improved accuracy in classification. Successfully obtained $61,050 in funding for my independent research proposal on classifier fusion. Collaborated with clinicians, statisticians and radiologists from multiple universities in Australia, Canada and USA. Main Project Coordinator for a multi-site project on Early Detection of Alzheimer s disease funded by Pacific Alzheimer Research Foundation, from 2008 2012 (4 years). Duties included: Coordinating meetings, tracking and communicating progress on a regular basis. Responsible for consolidating, securing and synchronizing data across the sites. Collaboration across 4 universities in Australia, Canada and USA, including Johns Hopkins and Northwestern Universities. President, Engineering Science Graduate Student Association at Simon Fraser University September 2009 - September 2010 (1 year 1 month) President, Engineering Science Graduate Student Association Research Assistant at Aalborg University, Denmark February 2006 - October 2007 (1 year 9 months) Implemented computer vision techniques (particle filtering) for human activity recognition in video as part of multi-site project on building humanoids with artificial intelligence (PACO-Plus). Optimized the performance of hidden markov models in recog- nizing table-top activities. Compiled a large motioncapture dataset (infrared sensor and video) of simple human actions. Project Associate at Indian Institute of Technology, Madras May 2005 - January 2006 (9 months) Development of Nondestructive Evaluation techniques for the detection, characterization and staging of subsurface defects using Infrared Thermography. Education Simon Fraser University Ph. D, Biomedical Engineering, 2008-2014 Activities and Societies: Engineering science Graduate Student Association, Graduate Student Society Indian Institute of Technology, Madras M.S, Physics, 2003-2005 Sri Venkateswara University B.S, Mathematics, Physics, Computer Science, 2000-2003 Honors and Awards Alzheimer Society Canada Research Scholarship Page2
Alzheimer Society Canada April 2011 Alzheimer Society Canada Research Scholarship, valued at $61,590, for my research proposal on developing novel computer-aided diagnosis algorithms for early detection of Alzheimer's disease, as well as differential diagnosis of neurodegenerative disorders such as frontotemporal disease etc. Three Minute Thesis Contest Winner Simon Fraser University March 2014 People's Choice award based on voting by 150+ graduate students and staff at SFU. More details at: http:// www.sfu.ca/dean-gradstudies/events/three-minute-thesis.html Graduate Fellowship Simon Fraser University May 2010 Valued: $6,250 Westak International Research Award Westak International January 2010 Travel fellowship Alzheimer Association May 2013 to present at the 2013 Alzheimer Association International conference (complete with air fare, accommodation and registration). Publications The sub-classification of amnestic MCI using Novel ThickNet features Neuroimage Clinical August 4, 2014 Authors: Pradeep Redddy Raamana, Wei Wen, Nicole A. Kochan, Henry P. Brodaty, Perminder Sachdev, Lei Wang, Mirza Faisal Beg BACKGROUND: Amnestic mild cognitive impairment (amci) is considered to be a transitional stage between healthy aging and Alzheimer s disease (AD), and consists of two subtypes: single-domain amci (sd-amci) and multi-domain amci (md-amci). Individuals with md-amci are found to exhibit higher risk of conversion to AD. Accurate discrimination among amci subtypes (sd- or md-amci) and controls could assist in predicting future decline. METHODS: We apply our novel thickness network (thicknet) features to discriminate md-amci from healthy controls (NC). Thicknet features are extracted form the properties of a graph constructed from inter-regional co- variation of cortical thickness. We fuse these thicknet features using multiple kernel learning to form a composite classifier. We apply the proposed thicknet classifier to discriminate between md-amci and NC, sd-amci and NC and; and also between sd-amci and md-amci, using baseline T1 MR scans from the Sydney Memory and Ageing Study. RESULTS: Thicknet classifier achieved an area under curve (AUC) of 0.74, with 70% sensitivity and 69% specificity in discriminating md- Page3
amci from healthy controls. The same classifier resulted in AUC=0.67 and 0.67 for sd-amci/nc and sdamci/md-amci classification experiments respectively. CONCLUSIONS: The proposed thicknet classifier demonstrated potential for discriminating md-amci from controls, and in discriminating sd-amci from mdamci, using cortical features from baseline MRI scan alone. Use of the proposed novel thicknet features demonstrates significant improvements over previous experiments using cortical thickness alone. This result may offer the possibility of early detection of Alzheimer s disease via improved discrimination of amci subtypes. Thickness Network (ThickNet) features for Prognostic Applications in Dementia Neurobiology of Aging (in press). March 15, 2014 Authors: Pradeep Redddy Raamana, Michael W. Weiner, Lei Wang, Mirza Faisal Beg Regional analysis of cortical thickness has been studied extensively in building imaging biomarkers for early detection of Alzheimer s disease (AD), but not its inter-regional covariation of thickness. We present novel features based on the inter-regional co-variation of cortical thickness. Initially the cortical labels of each subject is partitioned into small patches (graph nodes) by spatial k-means clustering. A graph is then constructed by establishing a link between two nodes if the difference in thickness between the nodes is below a certain threshold. From this binary graph, a thickness network (ThickNet) is computed using nodal degree, betweenness and clustering coefficient measures. Fusing them with multiple kernel learning, it is observed that ThickNet features discriminate mild cognitive impairment (MCI) converters from controls (CN) with an area under curve (AUC) of 0.83, 74% sensitivity and 76% specificity on a large subset obtained from the ADNI dataset. A comparison of predictive utility in AD/CN classification (AUC of 0.92, 80% sensitivity and 90% specificity), in discriminating CN from MCI (converters and non-converters combined; AUC of 0.75, SENS/SPEC of 64%/73%) and in discriminating between MCI non-converters and MCI converters (AUC of 0.68, SENS/SPEC of 65%/64%) is also presented. Thicknet features as defined here are novel, can be derived from a single MRI scan and demonstrate the potential for the computer-aided prognostic applications. Three class differential diagnosis among Alzheimer's Disease, Frontotemporal Disease and Healthy Controls Frontiers in Neurology 5(71) April 28, 2014 Authors: Pradeep Redddy Raamana, Howard Rosen, Bruce Miller, Michael Weiner, Lei Wang, Mirza Faisal Beg Biomarkers derived from brain magnetic resonance imaging have promise in being able to assist in the clinical diagnosis of brain pathologies. These have been used in many studies in which the goal has been to distinguish between pathologies such as Alzheimer's disease and healthy aging. However, other dementias, in particular, Fronto-temporal dementia, also present overlapping pathological brain morphometry patterns. Hence, a classifier that can discriminate morphometric features from a brain MRI from the three classes of normal aging, Alzheimer s disease (AD) and Frontotemporal dementia (FTD) would offer considerable utility in aiding in correct group identification. Compared to the conventional use of multiple pair-wise binary classifiers that learn to discriminate between two classes at each stage, we propose a single threeway classification system that can discriminate between three classes at the same time. We present a novel Page4
classifier that is able to perform a three-class discrimination test for discriminating among AD, FTD and normal controls using volumes, shape invariants and local displacements of hippocampi and lateral ventricles obtained from brain MR images. In order to quantify its utility in correct discrimination, we optimize the three-class classifier on a training set and evaluate its performance using a separate test set. This is a novel, first-of-its-kind study in a three-class setting using multiple morphometric features. Our results demonstrate that local atrophy features in lateral ventricles offer the potential to be a biomarker in discriminating among Alzheimer s disease, frontotemporal dementia and normal controls in a 3-class setting for individual patient classification. The sub-classification of amnestic MCI using MRI-based cortical thickness measures Frontiers in Neurology 5(76) March 1, 2014 Authors: Pradeep Redddy Raamana, Wei Wen, Nicole A. Kochan, Henry P. Brodaty, Perminder S. Sachdev, Lei Wang, Mirza Faisal Beg Background: Amnestic Mild cognitive impairment (amci) is considered to be the transitional stage between healthy aging and Alzheimer s disease (AD). Moreover, amci individuals with impairment in one or more non-memory cognitive domains are at higher risk of conversion to AD. Hence accurate identification of the subtypes of amci would enable earlier detection of individuals progressing to AD. Methods: We examine the group differences in cortical thickness between single-domain and multiple-domain subtypes of amci, and as well as with respect to age-matched controls in a well- balanced cohort from the Sydney Memory and Ageing Study. In addition, We assess the diagnostic value of cortical thickness in the sub-classification of amci as well as from normal controls using support vector machine (SVM) classifier, using a novel crossvalidation technique averaging the results from 250 repetitions. Results: The best performance of classifier for the pairs 1) single domain amci and normal controls, 2) multiple domain amci and normal controls and 3) single and multiple domain amci is AUC=0.52, 0.66 and 0.54 respectively. The accuracy of the classifier for the three pairs was just over 50% exhibiting low specificity (44-60%) and similar sensitivity (53-68%). Conclusions: The results show that discrimination among single, multiple domain subtypes of amci and normal controls is limited using baseline cortical thickness measures. Comparison of Four Shape Features for the detection of Hippocampal Changes in Early Alzheimer s Journal of Statistical Methods in Medical Research May 30, 2012 Authors: Pradeep Redddy Raamana, Mirza Faisal Beg We compare four methods for generating shape-based features from 3D binary images of the hippocampus for use in group discrimination and classification. The first method we investigate is based on decomposing the hippocampal binary segmentation onto an orthonormal basis of spherical harmonics, followed by computation of shape invariants by tensor contraction using the Clebsch Gordan coefficients. The second method we investigate is based on the classical 3D moment invariants; these are a special case of the spherical harmonics-based tensor invariants. The third method is based on solving the Helmholtz equation on the geometry of the binary hippocampal segmentation, and construction of shape-descriptive features from the eigenvalues of the Fourier-like modes of the geometry represented by the Laplacian eigenfunctions. The fourth method investigates the use of initial momentum obtained from the large-deformation diffeomorphic metric mapping method as a shape feature. Each of these shape features is tested for group differences Page5
in the control (Clinical Dementia Rating Scale CDR 0) and the early (very mild) Alzheimer's (CDR 0.5) population. Classification of individual shapes is performed via a linear support vector machine based classifer with leave-one-out cross validation to test for overall performance. These experiments show that all of these feature computation approaches gave stable and reasonable classification results on the same database, and with the same classifier. The best performance was achieved with the shape-features constructed from large-deformation diffeomorphic metric mapping-based initial momentum. Thickness Network (ThickNET) measures for the detection of prodromal Alzheimer Disease Accepted in MICCAI (Medical Image Computing and Computer Aissisted Interevention), Machine Learning in Medical Imaging workshop March 1, 2013 Authors: Pradeep Redddy Raamana, Lei Wang, Mirza Faisal Beg Regional analysis of cortical thickness has been studied extensively in building imaging biomarkers for early detection of Alzheimer s disease (AD), but not its inter-regional covariation of thickness. We present novel features based on the inter-regional co-variation of cortical thickness. Initially the cortical labels of each subject is partitioned into small patches (graph nodes) by spatial k-means clustering. A graph is then constructed by establishing a link between two nodes if the difference in thickness between the nodes is below a certain threshold. From this binary graph, a thickness network (ThickNet) is computed using nodal degree, betweenness and clustering coefficient measures. Fusing them with multiple kernel learning, it is observed that ThickNet features discriminate mild cognitive impairment (MCI) converters from controls (CN) with an area under curve (AUC) of 0.83, 74% sensitivity and 76% specificity on a large subset obtained from the ADNI dataset. A comparison of predictive utility in AD/CN classification (AUC of 0.92, 80% sensitivity and 90% specificity), in discriminating CN from MCI (converters and non-converters combined; AUC of 0.75, SENS/SPEC of 64%/73%) and in discriminating between MCI non-converters and MCI converters (AUC of 0.68, SENS/SPEC of 65%/64%) is also presented. Thicknet features as defined here are novel, can be derived from a single MRI scan and demonstrate the potential for the computer-aided prognostic applications. Evidence of possible matter shift in human brain due to neurodegeneration Organization of Human Brain Mapping 2012 Authors: Pradeep Redddy Raamana, Lei Wang, Mirza Faisal Beg Introduction: Magnetic resonance imaging made it possible to precisely image deep subcortical structures in the human brain. Deformation caused by neurodegeneration in structures such as hippocampus, or pattern of such deformation on its surface, is a an early stage biomarker of Alzheimer's disease. Proper analysis of subcortical atrophy requires accurate spatial positioning of hippocampus in the atlas head space. Methods: Research studies so far relied on atlas-based approaches to quantify the pattern of atrophy on hippocampus surface. Such studies typically perform a rigid alignment of hippocampi (see Figure 1) - prior to a highdimensional registration, which allows precise computation of deformation on the hippocampal surface [2]. Rigid registration methods based on least-squares optimization have the tendency to drive the two surfaces being registered to align the centers of mass. We hypothesize this process, not only ignores, but also the masks gross shift of matter occurring in the brain due to neurodegeneration in various parts of the brain. In order to account for that we propose that the initial registration be done via registering the skull of the Page6
patient to that of the atlas (see Figure 2), as neuro-degeneration doesn't affect the size of skull [1]. We also present an atlas-based method to segment the skull in a given MR image accurately. Results: We analyze the deformation features obtained from both the methods and present results which seem to suggest a gross matter shift happening in the medical temporal lobe. Conclusions: The results suggest that there might be a gross matter shift occurring in the brain even in the early course of Alzheimer disease due to the neurodegeneration. Automated detection of amnestic mild cognitive impairment in community-dwelling elderly adults: a combined spatial atrophy and white matter alteration approach NeuroImage November 1, 2011 Authors: Pradeep Redddy Raamana, Yue Cui Amnestic mild cognitive impairment (amci) is a syndrome widely considered to be prodromal Alzheimer's disease. Accurate diagnosis of amci would enable earlier treatment, and could thus help minimize the prevalence of Alzheimer's disease. The aim of the present study was to evaluate a magnetic resonance imaging-based automated classification schema for identifying amci. This was carried out in a sample of community-dwelling adults aged 70-90 years old: 79 with a clinical diagnosis of amci and 204 who were cognitively normal. Our schema was novel in using measures of both spatial atrophy, derived from T1- weighted images, and white matter alterations, assessed with diffusion tensor imaging (DTI) tract-based spatial statistics (TBSS). Subcortical volumetric features were extracted using a FreeSurfer-initialized Large Deformation Diffeomorphic Metric Mapping (FS+LDDMM) segmentation approach, and fractional anisotropy (FA) values obtained for white matter regions of interest. Features were ranked by their ability to discriminate between amci and normal cognition, and a support vector machine (SVM) selected an optimal feature subset that was used to train SVM classifiers. As evaluated via 10-fold cross-validation, the classification performance characteristics achieved by our schema were: accuracy, 71.09%; sensitivity, 51.96%; specificity, 78.40%; and area under the curve, 0.7003. Additionally, we identified numerous sociodemographic, lifestyle, health and other factors potentially implicated in the misclassification of individuals by our schema and those previously used by others. Given its high level of performance, our classification schema could facilitate the early detection of amci in community-dwelling elderly adults. Human Action Recognition in Table-top Scenarios: An HMM-based Analysis to Optimize the Performance 12th International Conference on Computer Analysis of Images and Patterns 2007 Authors: Pradeep Redddy Raamana, Daniel Grest, Volker Krueger Hidden Markov models have been extensively and successfully used for the recognition of human actions. Though there exist well-established algorithms to optimize the transition and output probabilities, the type of features to use and specifically the number of states and Gaussian have to be chosen manually. Here we present a quantitative study on selecting the optimal feature set for recognition of simple object manipulation actions pointing, rotating and grasping in a table-top scenario. This study has resulted in recognition rate higher than 90%. Also three different parameters, namely the number of states and Gaussian for HMM and the number of training iterations, are considered for optimization of the recognition rate with 5 different feature sets on our motion capture data set from 10 persons. Page7
Courses Ph. D, Biomedical Engineering Simon Fraser University Biomedical Image Computing Machine Learning Mathematic Image Processing Statistical Signal Processing CMPT829 CMPT726 APMA990 ENSC810 Skills & Expertise Machine Learning Pattern Recognition Image Processing Medical Imaging Biomedical Engineering Matlab LaTeX R Signal Processing Programming Feature Selection C++ C Java Python Statistical Modeling Data Analysis Microsoft Office Mathematical Modeling Computer Vision Statistics Computer Science Medical Image Analysis Image Analysis SQL Numerical Analysis Statistical Learning Alzheimer's disease Computer Aided Diagnosis Differential Diagnosis Feature Extraction Digital Signal Processing magnetic resonance imaging Page8
Data Visualization Linux Scientific Computing 3D visualization Digital Image Processing High Performance Computing Image Segmentation programm Algorithm Development Linear Algebra Data Mining Optimization Parallel Programming feature extra Algorithms Classifiers Microsoft Excel Honors and Awards First Prize in Three Minute Thesis presentation contest (SFU, Faculty of Applied Sciences) Second prize in Poster competition (among Ph.D's and PostDoc's) in MITACS-Fields Conference of Mathematics in Medical Imaging, 2011. Best Project Award during Summer School at Institute for Plasma Research, India. University Topper in B.S(Mathematics, Physics and Computing Science) from among thousands of students at Sri Venkateswara University, India. Topped several admission examinations for M.S in Physics to several universities in India - including IIT Madras. Interests * Medical Image Processing and Analysis * Early Detection of Alzheimer's Dementia * Differential diagnosis of neurodegenerative disorders * Building novel imaging biomarkers * Building novel feature reduction, selection methods, as well as customized classifiers * Computer Aided Diagnosis * Statistical Learning * Pattern Recognition & Prediction Volunteer Experience Fundraising at BC Cancer Agency June 2011 - Present (3 years 9 months) Raised money (CAD$ 1500) to support research for cancers below the belt. Languages Hindi Telugu Page9
Pradeep Redddy Raamana Research Scientist at Simon Fraser University pkr1@sfu.ca Contact Pradeep Redddy on LinkedIn Page10