Anirudh Choudhary

I am a second year PhD student in ECE at UIUC working with Professor Ravishankar Iyer. I am a part of the DEPEND group at the Coordinated Sciences Lab. My research interests lie in machine learning and probabilistic modeling for biomedical data analysis with emphasis on limited data settings. I collaborate closely with Dr. Aaron Mangold (Mayo Clinic).

Previously, I completed my Masters in Computational Science and Engineering at Georgia Tech, where I was a researcher at Biomedical Informatics Laboratory advised by Professor May Wang. My research focused on robust policy learning for clinical decision-making using electronic health records. Additionally, I also got an opportunity to work with Professor Jaydev Desai at RoboMed Lab on image-guided surgery.

Earlier, I did my MBA from IIM Calcutta, and completed my undergraduate studies at IIT Kharagpur, where I did my thesis with Prof. Ajoy Kumar Ray on texture-based oral cancer prediction using histopathology images. While at IIT Kharagpur, I pursued summer internships with Prof. Christos Davatzikos (University of Pennsylvania) and Prof. Paolo Fiorini (University of Verona). I have five years of professional experience in developing machine learning models for e-commerce and retail consumer data analysis.

Email / CV / Google Scholar / LinkedIn

Affiliations


IIT Kharagpur 2005-2010	Altair Robotics Lab (UVerona) Summer 2008	SBIA (UPenn) Summer 2009	IIM Calcutta 2011-2013	Georgia Tech 2018-2020	UIUC 2020-present

Industry Experience


Sabre Corporation 2010-2011	EXL 2013-2016	Loyalty Partner (AmEx subsidiary) 2016-2017	Mastercard 2017-2018

Recent Projects

My work focuses on leveraging weakly-supervised and probabilistic approaches to develop diagnostic/prognostic models for clinical data.

Image Processing

Weakly-supervised classification of squamous skin cancer histopathology images using attention-based multiple instance learning and graph neural networks
Unsupervised/self-supervised deep representation learning for analyzing semantic similarity in histopathology images

Reinforcement Learning/ Clinical Policy Learning

Integrating mechanistic differential equations (domain-knowledge) with reinforcement learning to predict cognition trajectories in Alzheimer's disease
Bootstrapping and adversarial learning-based frameworks to tackle model uncertainity in offline-policy learning for electronic health records (Warfarin dosing)
Incorporated meta-learning (REPTILE) with IRL-based imitation learning to develop personalized Sepsis treatment policies using MIMIC-III

Research Publications

	Reinforcement Learning based Disease Progression Model for Alzheimer's Disease K. V. Saboo, A. Choudhary, Y. Cao, G. A. Worrell, D. T. Jones, R. K. Iyer Advances in Neural Information Processing Systems, 2021 We propose a framework for modeling AD progression that combines differential equations (DE) with reinforcement learning (RL). In our model, the available DEs define the simulator and the RL agent optimizes the domain-based reward function in the simulato. rRL combined with the DEs describes the evolution of factors pertinent for modeling AD progression.
	Pose-aware C-Arm Calibration and Image Distortion Correction for Guidewire Tracking and Image Reconstruction F. Heemeyer, A. Choudhary, J. P. Desai (equal contribution) IEEE International Symposium on Medical Robotics (Georgia Tech)*, 2020 Image intensifiers, also known as C-arms, are important and low-cost tools for surgeons to guide minimally-invasive procedures. However, image intensifiers suffer from several image distortions, which can be misleading during an automated minimally-invasive surgery. We propose a combination of camera-based tracking system and calibration grid to perform accurate distortion correction.
	Learning to Evaluate Color Similarity for Histopathology Images using Triplet Networks A. Choudhary, H. Wu, L. Tong, M. D. Wang ACM International Conference on Bioinformatics, Computational Biology and Health Informatics, 2019 (Long Oral) (Invited for JBHI Special Issue) Inspired by the effectiveness of deep neural networks in evaluating perceptual similarity of natural images, in this paper, we propose a representation learning approach using triplet network for evaluating color-based perceptual similarity for whole slide images.
	Texture based segmentation of epithelial layer from oral histological images M. Krishnan, A. Choudhary, C. Chakraborty, A. K. Ray, R. Paul Micron Journal (Elsevier), 2011 A novel texture based segmentation algorithm for better delineation of the epithelial layer from histological images using Gabor Filter and multistep region-merging using watershed-generated superpixels.
	Textural characterization of histopathological images for oral sub-mucous fibrosis detection M. Krishnan, P. Shah, A. Choudhary, C. Chakraborty, R. Paul, A. K. Ray Tissue Cell Journal (Elsevier), 2011 Evaluating statistical and spectral texture features of epithelial layer in oral whole slide images for cancer classification using Support Vector Machines.
	An Entropy Based Multi-Thresholding Method for Semi-Automatic Segmentation of Liver Tumors A. Choudhary N. Moretto, F. P. Ferrarese, G. A. Zamboni Workshop on 3D Segmentation in the Clinic: Liver Tumor Segmentation Challenge, MICCAI, 2008 A semi-automatic liver tumor segmentation technique using cross-entropy minimization based thresholding and level sets.

Articles

Advancing Medical Imaging Informatics by Deep Learning-Based Domain Adaptation
A. Choudhary*, L. Tong*, Y. Zhu, M. D. Wang
IMIA Yearbook of Medical Informatics 2020

Presentations/Talks

Bootstrapping-based robust offline counterfactual learning in clinical settings (CSL Student Conference 2020)

Academic Achievements

Teachers Ranked as Excellent, Fall 2021, ECE 598 Dependable AI Systems	2021
NSF Travel Grant and Georgia Tech Student Travel Award - ACM-BCB Conference 2019	2019
Research Assistantships with full tuition support throughout graduate studies at Georgia Tech	2018-20
99.61 percentile in Common Admission Test for admission to Indian Institutes of Management	2011
Masters research scholarship awarded by Government of India at IIT Kharagpur	2010
Research assistantships during internships at University of Verona and University of Pennsylvania	2009
All India Rank 68 in IIT-JEE Prelims Examination and 507 in All India Engineering Entrance Examination	2005
CBSE Merit Certificate in Mathematics for being among top 0.01% of students at a national level in Std 12th	2004
Secured 5th rank in Regional Mathematical Olympiad and participated in Indian National Mathematics Olympiad	2002

Professional/ Extracurricular Achievements

Business Excellence Award and Quarterly Best Performer Award at Loyalty Partner, India	2017
Among top 5% performers at EXL Analytics - Rated 'Exceeds Expectations'	2015
Runners-up in Procter and Gamble’s marketing strategy case-study competition, IIM Calcutta	2013
Killer Tech Award and High Five Award for being among top 5 performers at Sabre Corporation	2010
Best Outgoing Technology Award for contributions to intra-institute technical events at IIT Kharagpur	2010
Winner at Envision, a national-level product design competition at Entrepreneurship Summit, IIT Kharagpur	2010

Courses

Computer Vision, Advanced Computer Vision
Graphical Models, Computational Data Analysis, Artificial Intelligence, Machine Learning with Limited Supervision, Computational Inference, Random Processes
Deep Learning, Reinforcement Learning
CSE Algorithms, Numerical Linear Algebra, Modeling & Simulation

Academic Projects

Simulation of HIV infection using mean-field approach and antiretroviral drug-dosing using reinforcement learning (Report) | Modeling & Simulation

Guidewire detection and 3D-Reconstruction for X-Ray image-guided surgery (Report) | Special Problem

Empirical analysis of Ordered Neurons based LSTM(ON-LSTM) on language modeling tasks and benchmarking against vanilla LSTM and AWD-LSTM on toxic comment identification task. (Report) | Deep Learning

Adversarial Domain Adaptation in Videos using spatio-temporal features for activity recognition (Video, Slides) | Machine Learning with Limited Supervision

Evaluating heuristic and local search-basedapproaches for Traveling Salesman Problem (Report, Code) | CSE Algorithms

Template Credits: Jon Barron