My research is centered on the early detection and classification of neurodegenerative disorders, particularly Parkinson’s Disease (PD) and Alzheimer’s Disease (AD), leveraging advanced machine learning (ML) and deep learning (DL) techniques on neuroimaging data. Motivated by the growing global burden of these disorders and the critical need for timely diagnosis, my work focuses on designing robust, interpretable, and computationally efficient models to extract and analyze discriminative features from 3D structural brain data.
In my paper titled "Early Detection and Classification of Alzheimer’s through Integration of 3D Local Binary Patterns and SVM classifier" (IEEE ISACC 2025), I developed a volumetric feature extraction framework that integrates 3D Local Binary Patterns (3D-LBP) with a Support Vector Machine (SVM) classifier. The methodology utilizes structural MRI data and emphasizes capturing spatial dependencies across 3D voxel neighborhoods—an enhancement over traditional 2D feature extraction methods. Experimental validation on the ADNI1 dataset demonstrated the model’s outstanding classification accuracy of 99.01% in distinguishing Alzheimer's patients from cognitively normal controls, confirming the effectiveness of volumetric texture descriptors in early AD detection.
Complementing this, my co-authored review article "Review on computational methods for the detection and classification of Parkinson’s Disease" (Computers in Biology and Medicine, 2025) provides a comprehensive survey of biomarkers, imaging modalities, and ML/DL algorithms for PD diagnosis. The paper underscores the limitations in current approaches, particularly the need for high-quality, multi-modal datasets, better feature interpretability, and domain-specific feature engineering. It also highlights the importance of explainable AI tools (e.g., LIME, SHAP) in mitigating the black-box nature of DL models and improving clinical trust.
Further extending this work, the Springer book chapter "An Ensemble Learning Framework for Classification of Alzheimer’s Disease using MRI Data" explores ensemble-based strategies combining multiple classifiers to enhance robustness in AD diagnosis. The ensemble approach builds upon base classifiers trained on distinct feature sets extracted from MRI scans, demonstrating improved generalization over single-model systems. This line of research aligns with my long-term interest in model fusion and multimodal learning for healthcare applications.
Together, these projects establish a unified theme in my research: designing interpretable, efficient, and generalizable computational frameworks for the automated classification of neurodegenerative diseases using medical imaging. My future work will focus on three key directions:
- Integrating multi-modal imaging biomarkers (e.g., PET, fMRI) with genetic and clinical data for holistic disease modeling.
- Expanding current models to handle longitudinal data for progression prediction.
- Enhancing model interpretability and clinical applicability through explainable AI and user-centric visualization tools.
Ultimately, my goal is to contribute tools that not only advance scientific understanding but also support clinicians in real-world decision-making for early diagnosis and personalized treatment planning.
