Research

Research Overview

My research focuses on advancing the application of graph neural networks and deep learning techniques to solve complex problems in medical imaging. I am particularly interested in addressing challenges related to homophily in graph-based learning and developing more effective methods for feature extraction and classification in medical image analysis.

Research Interests

1. Graph Neural Networks for Image Classification

Addressing homophily challenges in graph neural networks when applied to image classification tasks. Developing novel approaches to improve GNN performance on non-homophilic graph structures, particularly for medical image classification.

2. Medical Image Analysis with Graph Neural Networks

Leveraging graph neural networks and deep learning models for medical image analysis:

• Medical image classification using graph-based approaches
• Ensemble methods for improved diagnostic accuracy
• Spectral feature analysis in medical imaging
• Edge convolution techniques for medical image processing
• Dynamic filter applications in graph convolutional networks

3. Subgraph Clustering and Atom Learning

Investigating subgraph clustering techniques combined with atom learning approaches to enhance image classification performance, particularly in medical imaging contexts.

4. Graph-based Feature Extraction

Developing compact and capable models using graph neural networks with edge convolution techniques for medical image classification, focusing on improved spectral feature analysis and class discrimination.

Research Highlights

Improving GNNs for Image Classification: Addressing Homophily Challenges

Addressing the fundamental challenge of homophily in graph neural networks when applied to image classification tasks. This work explores novel approaches to improve GNN performance on non-homophilic graph structures.

Graph Neural Networks for Medical Image Classification

Developing ensemble methods and classification techniques using graph neural networks specifically tailored for medical imaging applications. This research focuses on improving diagnostic accuracy through advanced graph-based feature extraction.

Subgraph Clustering and Atom Learning

Investigating subgraph clustering techniques combined with atom learning approaches to enhance image classification performance, particularly in medical imaging contexts.

Compact & Capable: Edge Convolution for Medical Imaging

Exploring the use of graph neural networks with edge convolution techniques to create more compact yet capable models for medical image classification.

Dynamic Filter Application in Graph Convolutional Networks

Developing dynamic filter mechanisms for graph convolutional networks to enhance spectral feature analysis and improve class discrimination in medical imaging applications.

Research Tools & Technologies

• PyTorch, TensorFlow
• Graph Neural Network Libraries (PyTorch Geometric, DGL)
• Medical Imaging Libraries (ITK, SimpleITK, MONAI)
• Cloud Computing (Microsoft Azure, Azure ML)
• Data Processing (Apache Kafka, Apache Storm)

For the most up-to-date list of publications and citations, please visit my Google Scholar profile.