Severity Prediction in Spinal Cord Injury using Multimodal MRI Paper
Developed deep-learning models combining T2 and DTI MRI features to classify ASIA impairment levels (A–D). Utilized 3D CNNs and transformer encoders for multimodal fusion and achieved >94 % accuracy.

Magnetization Transfer (MT) MRI Biomarkers in Pediatric Spinal Cord Paper
Investigated quantitative MT metrics to assess myelin integrity and their relationship with diffusion and structural measures across vertebral levels in typically developing children.

Whitening-Aided Network for Radar-Based Human Activity Recognition Paper
Developed a Whitening-Aided Network for radar micro-Doppler activity recognition to enhance domain generalization and feature decorrelation across sensors and environments. Integrated whitening normalization and attention mechanisms to improve robustness and reduce overfitting.

Semi-Supervised Attention-Augmented Autoencoder Paper
Proposed a semi-supervised autoencoder with multi-head attention layers for learning from large unlabeled radar datasets. Achieved >95 % accuracy and >20 % improvement over baseline CNNs.

Age-Clustering Framework for Pediatric Spinal Cord Morphometry
Implemented decision-tree and clustering approaches using structural MRI metrics (CSA, AP, RL widths) to identify biologically consistent pediatric subgroups. Provides new rules for age/height-normalized analyses.

Adult Spinal Cord Injury Severity Assessment using Deep Learning
Built a multimodal 3D deep-learning framework integrating T2 and DTI data for AIS classification in acute SCI. Evaluated generalization to chronic follow-ups.

Modular MRI Analysis Pipeline (GUI-Based)
Designed a fully modular Python pipeline supporting processing for T2 and HYDI MRI. Includes segmentation, QC, feature extraction, and report generation.

Classification of Hand Movements from EEG Records report
Used CNN and LSTM models to classify hand movements from time-series EEG recordings, achieving 87.9 % accuracy.

RatTracker: Computer-Vision Platform for Behavioral CPP Analysis report
Developed a CV system to track and analyze rodents in conditioned place-preference tests with 3 ms temporal precision.

Predicting User Preferences in Social Networks Using Clustering Algorithms report Designed a clustering-based model for predicting users’ post-like preferences across 90,000 WordPress blogs.

News Title Text Mining and Classification report
Analyzed two decades of newspaper headlines (1996–2013) and created a classification model achieving 80 % accuracy.