Area of Interest
Artificial Intelligence and Machine learning on
1) Computational neuroscience
2) Biomedical signal processing
3) Healthcare informatics
Expertise
Core Research Domains
Brain Functional Connectivity (BFC) Analysis with emphasis on neurodevelopmental disorders (ADHD, ASD).
Cardio-renal risk prediction integrating CVD and CKD datasets with statistical learning.
Investigation of the interrelationship between neurodevelopmental disorders and cardio-renal risk factors using data-driven methods.
Development of geodesic brain connectivity frameworks for realistic modeling of nonlinear neural interactions.
Advanced Methodological Contributions
Proposal and implementation of Composite Graph Spectral Analysis (CGSA) combining:
Traditional Graph Fourier Transform
Riemannian Graph Spectral methods
Data-driven ROI extraction using Grouped ICA + Dictionary Learning for robust feature representation.
Mutual Information (MI)–based dynamic Functional Connectivity (dFC) construction with optimized sliding window strategies.
Threshold-based connectivity modeling and energy-based window optimization.
Machine Learning & Deep Learning
Design and evaluation of:
1D-CNN architectures for brain connectivity classification
Ensemble learning frameworks for improved diagnostic accuracy
Comparative analysis of ML vs DL performance across thresholds and spectral powers.
Feature subset evaluation using correlation-guided R² scoring.
Hyperparameter tuning (e.g., Pegasos, regularization grids).
Implementation of custom k-means (without built-ins) and other algorithmic ML methods.
Statistical & Mathematical Expertise
Strong foundation in:
Linear algebra (eigen decomposition, SVD, covariance structure)
Optimization (gradients, Hessians, LMS, convex loss)
Monte Carlo integration with region-wise probability estimation
Multivariate statistical analysis for risk factor significance testing.
Continuous vs categorical modeling (regression variants beyond logistic).
Signal Processing & Graph Theory
Graph-based modeling of brain networks with spectral domain feature extraction.
Handling high-dimensional ROI connectivity matrices.
Connectivity strength visualization and maximum-edge analysis.
Integration of topological and spectral features for classification.
CALL US
