Building Adaptive AI for MRI Diagnostics: A Continual Learning Case Study
summary
A UNet encoder–decoder with a variational decoder and T1 and FLAIR plus junction/extension morphometrics processes MRIs to localize Focal Cortical Dysplasias using a dataset of >200 FCD series (≈200 slices each) plus normals and other pathologies; initial results hit ≈70% case-level accuracy and outperformed neuro-radiologists (7 of 9 detections vs 3 of 9). Applying Elastic Weight Consolidation continual learning on smaller batches (30 FCD /150 normal/120 other) preserves sensitivity (8–9/10) while markedly increasing specificity and robustness.
AI-written summary · about
chapters (7)
- 0:00Continual learning in AI
- 10:23Focal Cortical Dysplasia detection
- 19:32Why traditional machine learning falls short
- 23:26Catastrophic forgetting
- 28:00Continual learning methods
- 30:35FCD detection with continual learning
- 38:56Summary and future
related talks
50mBuilding AI That Learns and Adapts: A Case Study in MRI Diagnostics
49mBuilding AI That Learns and Adapts: A Case Study in MRI Diagnostics
42mBuilding AI That Learns and Adapts: A Case Study in MRI Diagnostics
43mBuilding AI That Learns and Adapts: A Case Study in MRI Diagnostics
15mBrain Tumor Detection and Classification Using Deep Learning Models
34mWhy AI in the Medical Field Both Sucks and Rocks...
55mDeep Learning in Medical Image Diagnostics
47mThe whole truth about using deep learning in medical imaging