Breast Cancer Detection via Multimodal Fusion with GCN

Jan 1, 2022 · 1 min read
Multimodal mammogram analysis with GCN for mass detection

Breast Cancer Detection via Multimodal Fusion with GCN builds Graph Convolutional Neural Network models to reason over bilateral and ipsilateral mammogram views for improved malignancy detection and semantic segmentation.

Key Contributions

  • Multimodal Fusion: Developed GCN models with Mask-RCNN backend to fuse information across multiple mammogram views, capturing bilateral and ipsilateral cues for architectural distortion analysis.

  • Architectural Distortion Augmentation: Implemented augmentation strategies to improve sensitivity to temporal and structural breast tissue changes over multiple clinical visits.

  • Mass Detection & Segmentation: Models perform both malignancy detection and semantic segmentation of suspicious regions, supporting radiologist decision-making.

  • Course Context: Initial development in CS 791 (Mass Detection in Mammograms) at UNR; extends expertise in multi-view fusion and spatial–temporal pattern learning from retinal imaging to mammography.

Medical Imaging Deep Learning Graph Neural Networks PyTorch
Md Abu Sayed
Authors
Md Abu Sayed
Ph.D. Candidate | Gen AI | Naval Security | Simulation | Robotic Vision