Lightning Talk: A Novel Domain Generalization Technique for Medical Imaging Using...
summary
A PyTorch module called ContriMix disentangles biological content and technical attributes, mixes them to generate synthetic patches, and plugs into training as an on‑the‑fly augmentation. Trained on three hospital domains and evaluated on held‑out hospitals, it achieves state‑of‑the‑art results on the Camelyon17 WILDS leaderboard. Implementation uses off‑the‑shelf PyTorch multi‑GPU and mixed‑precision support, produces realistic stain‑transferred images that preserve nuclei signals, and ships code and inference scripts.
AI-written summary · about
chapters (7)
- 0:00Introduction to pathology
- 0:49Challenges in pathology data
- 2:37Importance of AI for diagnosis
- 4:32Introducing Contrix
- 5:31Implementing with PyTorch
- 6:36Performance and evaluation
- 8:11Conclusion and key takeaways
related talks
29mSynthetic Image Generation in Rare Disease Diagnosis
39mLightning Talks
09mLightning Talk: Bridging the Gap: Engineering Compliant...
55mDeep Learning in Medical Image Diagnostics
25mDeep Learning in Biomedicine: Challenges and Opportunities
34mWhy AI in the Medical Field Both Sucks and Rocks...
29mCancer diagnostics enabled by machines: finding and classifying single tumor....
26mHistology Augmentation: Scaling Federated Learning Solutions with Kubernetes