GMENet: Generative Mixture of Experts Network for Multi-Center Glioma Diagnosis with Incomplete Imaging Sequences
Researchers have developed GMENet, a novel Generative Mixture of Experts Network designed to improve glioma diagnosis from incomplete MRI sequences. This network synthesizes missing imaging data using cross-attention and dynamic gating, allowing for the utilization of more clinical data. GMENet also employs a dynamically weighted experts fusion module for multi-task prediction. Evaluations on a large, multi-center cohort demonstrated that GMENet can expand usable training data by 97% and outperforms existing state-of-the-art methods, showing increased robustness across different clinical centers. AI
IMPACT Improves diagnostic accuracy in medical imaging by enabling the use of incomplete datasets, potentially leading to earlier and more effective treatment.