• Lagged functional connectivity analysis for gender classification: Worked on the design, validation and analysis of a graph neural network (GNN) to incorporate lagged correlation analysis of blood-oxygen-level-dependent (BOLD) signals to extract richer features. The add-on network can be cascaded with different state-of-the-art neural network architectures to boost their classification accuracy and explainability.
  • Utilizing transformer neural network architecture for fMRI signal analysis: Design of a transformer encoder with a novel multi-head-self-attention (MHSA) layer to achieve blood-oxygen-level-dependent (BOLD) signal based gender, disease and task prediction.
  • Graph Neural Network based functional connectivity analysis for Parkinson’s disease (PD): Worked with Parkinson’s Progression Markers Initiative (PPMI) database to design and train graph neural networks to predict PD patients from fMRI scans.
  • Design of a Wavelet Based Decomposition for a Hierarchical Representation of 3D Scalar Wave Fields Related to Diffraction: Studying wavelet transform based optical signal processing.