Challenge
Identifying the underlying cause or subtype of a stroke from microscopic tissue images is a complex research challenge, requiring rigorous image curation, standardized preprocessing, and specialized deep-learning development.
Solution
The team developed an end-to-end AI system combining a curated image repository with HistomicsTK-based preprocessing and segmentation, followed by training and validation of a specialized deep-learning model.
Results
- Achieved a Silver Medal benchmark: 46th of 896 teams
- Delivered independently validated technical performance
- Created reusable assets to accelerate clinical research
- Enabled continued experimentation beyond the benchmark
Technology
Python · PyTorch · MONAI · HistomicsTK
Powered by DAC Digital