Building AI That Learns and Adapts: A Case Study in MRI Diagnostics
summary
Case study builds an MRI pipeline to detect focal cortical dysplasias using a 201 MRI series dataset (192 slices per series) with T1 and FLAIR inputs, brain tissue segmentation and morphometric channels, feeding a U-Net segmentation model. Baseline model hits ~70% sensitivity and 54% precision and finds 7 of 9 hard clinical cases versus neurologists’ 3 of 9. Switching to elastic weight consolidation continual learning raises specificity, reduces artifact-driven false positives, and yields a more robust deployable screening tool.
AI-written summary · about
chapters (13)
- 0:00AI in medical diagnostics
- 3:39Focal cortical dysplasia in epilepsy
- 8:57MRI dataset for FCD detection
- 13:12Neural network for FCD segmentation
- 19:18Medical AI deployment challenges
- 20:01Traditional ML training limits
- 23:17Continual learning importance
- 27:08Regularization in continual learning
- 28:54Memory-based AI adaptation
- 30:21Architecture for continual learning
- 31:47Continual learning for FCD detection
- 37:01Continual learning and LLMs
- 40:43Future of AI with continual learning
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
43mBuilding AI That Learns and Adapts: A Case Study in MRI Diagnostics
44mBuilding Adaptive AI for MRI Diagnostics: A Continual Learning Case Study
15mBrain Tumor Detection and Classification Using Deep Learning Models
44mAI in the Clinic: Prediction is Better Than Cure
34mArtificial intelligence to accelerate MRI scans - Philips & LUMC winners in the...
34mWhy AI in the Medical Field Both Sucks and Rocks...