I am Shreyas Fadnavis, currently serving as a postdoctoral research scientist at Johnson & Johnson within the Computer Vision Team. My work is primarily focused on developing Foundational Models for Transthoracic Echocardiography (TTE), Endoscopy, and exploring interpretability methods for deep learning models. Among my significant contributions is the "Patch2Self" framework, a novel approach that leverages self-supervised learning for the denoising of diffusion MRI data, and "Patch2Self-Sketch," which introduces matrix sketching for more efficient data processing. Holding a PhD from Indiana University Bloomington and having completed a postdoctoral fellowship at Harvard University/Harvard Medical School, my journey has been dedicated to advancing the fields of biomedical engineering, machine learning, and neuroimaging, aiming to solve complex challenges in medical diagnostics and treatments with innovative mathematical and computational techniques.
Multimodal ML, Time Series Modeling, Interpretable ML, Inverse
Problems in Neuroimaging (MRI), Tissue Microstructure Analyses, Harmonization
and Causal Inference.
August 2017 - June 2022 (Bloomington, Indiana)
Indiana University Bloomington, GPA 3.9/4.0
PhD in Engineering - Intelligent Systems
Medical Imaging, Machine Learning, Computational Topology,
Random Matrix Algorithms,Deep Learning, Statistical Modeling, Software
Development, Neuroscience.
August 2017 - May 2019 (Bloomington, Indiana)
Indiana University Bloomington
MS in Engineering - Intelligent Systems, GPA 3.9/4.0
Spotlight Presentation at Neural Information Processing
Systems
(NeurIPS) top 2.2% (top 208 of 9454 submissions)
Oral at International Society for Magnetic Resonance in
Medicine
(ISMRM) top 15% (top 878 of 5000+ submissions)
1st place in IEEE ISBI Competition for Biophysical Modeling
of
Tissue Microstructure (MEMENTO)
Graduate Fellowship for PhD, Indiana University
PUBLICATIONS (Citations: 1476+)
Published:
Fadnavis, S.*, Schilling, K.*, et. al. “Denoising of diffusion MRI in
the
cervical spinal cord–effects of denoising strategy and
acquisition on intra-cord contrast, signal modeling, and feature
conspicuity”,
NeuroImage, 2022
Fadnavis, S., et. al., “PainPoints: A Framework for Language-based
Detection of Chronic Pain and Expert-Collaborative
Text-Summarization”, Machine Learning for Health, 2022 [extended abstract],
under journal submission: Nature Scientific Reports
[arxiv |poster]
Fadnavis, S. et. al., “NUQ: A noise metric for diffusion MRI via
uncertainty discrepancy quantification”, Machine Learning for Health,
2022 [extended abstract], under journal submission: NeuroImage [arxiv |poster |video]
Fadnavis, S. el. al., “Effect of Denoising on Retrospective
Harmonization
of Diffusion Magnetic Resonance Images”, Neural
Information Processing Systems (MedNeurIPS), 2022 [poster]
Fadnavis, S.et. al. “Patch2Self: Denoising Diffusion MRI with
Self-Supervised Learning”, Thirty-fourth Conference on Neural
Information Processing Systems, (NeurIPS), 2020 [Spotlight Presentation]
[slides |video |poster]
Cieslak, M. et. al., “QSIPrep: An integrative platform for preprocessing and
reconstructing diffusion MRI”, Nature Methods, 2021
[software package link]
Fadnavis, S. et. al., “Bifurcated Topological Optimization of IVIM”,
Frontiers in Neuroscience, Brain Imaging Methods, 2021 [video]
Fadnavis, S. et. al., “MVD-Fuse: Detection of White Matter
Degeneration
via Multi-View Learning of Diffusion Microstructure ”,
Thirty-fourth Conference on Neural Information Processing Systems,
(NeurIPS),
Medical Imaging Workshop, 2020 [video]
De Luca, et. al., “On the generalizability of diffusion MRI signal
representations across acquisition parameters, sequences and tissue
types”, NeuroImage, 2021 [video]
Hu, et.al. “Segmentation of the Brain using Direction-averaged Signal of DWI
Images”, Magnetic Resonance Imaging, 2020
Hansen et. al., “Pandora: 4-D white matter bundle population-based atlases
derived from diffusion MRI fiber tractography”,
Neuroinformatics, 2020, DOI: 10.1007/s12021-020-09497-1
Fadnavis, S.. et. al., “NoiseFactors: Generative modeling for 3D
image
denoising via Randomized Factor Models”, International Society
for Magnetic Resonance in Medicine, 2020
Fadnavis, S.. et. al., “Variable Projection Optimization for
Intravoxel
Incoherent Motion (IVIM) MRI estimation”, Thirty-third
Conference on Neural Information Processing Systems(NeurIPS), Medical
Imaging
Workshop, 2019
Fadnavis, S.. et. al., “MicroLearn: Framework for machine learning,
reconstruction, optimization and microstructure modeling”,
International Society for Magnetic Resonance in Medicine, 2019
Fadnavis, S. “Maximizing Information of Severe Weather Events
Retrieved
from Satellite Images”, International Journal of
Engineering Research and Applications, ISSN : 2248-9622, 2017
Fadnavis, S.“Optimal Partitioning Methods for Image Segmentation”,
IET -
Journal of Engineering ISSN : 2051-3305, 2015
Garyfallidis, E., et. al. "Dipy, a library for the analysis of diffusion MRI
data." Frontiers in neuroinformatics 8, 2014
Fadnavis, S., “Image Interpolation Techniques in Digital Image
Processing”, International Journal of Engineering Research and
Applications, ISSN : 2248-9622, 2014
Fadnavis, S. “Roundoff Error Propagation in Simulation of RC
Circuit”,
Advances in Computing and Information Technology, ISBN :
978-981-07-8859-9, 2014
Garyfallidis, E., et. al. "Dipy, a library for the analysis of diffusion MRI
data." Frontiers in neuroinformatics 8, 2014
Pre-prints / Under Review:
Fadnavis, S.*, et. al., “ViewFormer: a View-Independent Transformer
Model
for Disease Detection from Transthoracic
Echocardiograms”, CVPR, 2024 (Under Review)
Park, J., Fadnavis, S., et. al. “EVAC+: Multi-scale V-net with Deep
Feature CRF Layers for Brain Extraction”, submitted to Medical
Image Analysis. ArXiv [code] - ISMRM’23 Oral.
Fadnavis, S. et. al. “Interpretable Specificity of Psychoactive Drug
Effects in Informal Dialogue Using Large Language Models.”
Biological Psychiatry, 2023 (under review)
Fadnavis, S. et. al. “PrompType: Prompting Large Language Models with
Prototypes”, 2023 (under submission)
Fadnavis, S., el. al., “Unsupervised Ensemble Learning of Risk
Predictors
for Clinically High Risk Psychosis”, JAMA Psychiatry, 2023
(Under submission)
Fadnavis, S.,. et. al. “Patch2Self-Sketch: Self-supervised Denoising
on
Coresets via Matrix Sketching”, submittedto Nature
Communications, 2023
Endres, S, et. al., “A Linear Complexity Simplicial Complex Data Structure
for
Fast Homology and Boundary Computation”, submitting
to TMLR, 2023
Garyfallidis, E, Fadnavis, S, et.al. “ThetA -- fast and robust clustering
via a
distance parameter”, ArXiv
Large Language Models, SpaCy, NLTK, Scikit-image,
Flask, Plotly Dash
RESEARCH HIGHLIGHTS
Patch2Self: Denoising Diffusion MRI with Self-Supervised
Learning”, Thirty-fourth Conference on Neural Information Processing Systems, (NeurIPS), 2020
Raw Data
Patch2Self
Noise Removed
“ViewFormer: a View-Independent Transformer Model for Disease Detection from
Transthoracic Echocardiograms”, under review, 2024
"QSIPrep: An integrative platform for preprocessing and reconstructing diffusion
MRI”, Nature Methods, 2021
“Bifurcated Topological Optimization of IVIM”, Frontiers in Neuroscience, Brain
Imaging Methods, 2021
“Patch2Self-Sketch: Self-supervised Denoising on Coresets via Matrix Sketching”,
under submission, 2023
Self supervised Learning and Imaging Denoising
“Denoising of diffusion MRI in the cervical spinal cord– effects of denoising
strategy and acquisition on intra-cord contrast, signal modeling, and feature conspicuity”,
NeuroImage, 2022
“NUQ: A noise metric for diffusion MRI via uncertainty discrepancy
quantification”, Machine Learning for Health, 2022
“Effect of Denoising on Retrospective Harmonization of Diffusion Magnetic
Resonance Images”, Neural Information Processing Systems (MedNeurIPS), 2022
“Patch2Self: Denoising Diffusion MRI with Self-Supervised Learning”, Thirty-fourth
Conference on Neural Information Processing Systems, (NeurIPS), 2020
"QSIPrep: An integrative platform for preprocessing and reconstructing diffusion
MRI”, Nature Methods, 2021
"Dipy, a library for the analysis of diffusion MRI data." Frontiers in
neuroinformatics 8, 2014
“Patch2Self-Sketch: Self-supervised Denoising on Coresets via Matrix Sketching”,
under submission, 2023
NLP, LLMs and Computational Psychiatry
“PainPoints: A Framework for Language-based Detection of Chronic Pain and
Expert-Collaborative Text-Summarization”, Machine Learning for Health, 2022
“Interpretable Specificity of Psychoactive Drug Effects in Informal Dialogue Using
Large Language Models.” Biological Psychiatry, 2023
“PrompType: Prompting Large Language Models with Prototypes”, under review, 2023
Multi-view Learning, Combinatorial Topology and other
Methods
“MVD-Fuse: Detection of White Matter Degeneration via Multi-View Learning of
Diffusion Microstructure”, NeurIPS, 2020
“Bifurcated Topological Optimization of IVIM”, Frontiers in Neuroscience, Brain
Imaging Methods, 2021