Synthetic Image Generation in Rare Disease Diagnosis
summary
Claims that synthetic chest x‑rays improve rare‑disease classifiers, demonstrated by two demos: an image‑to‑image pipeline using a latent diffusion model (autoencoder + diffusion in latent space) that outperforms baselines on validation MSE, and a text‑to‑image pipeline using stable diffusion via Hugging Face to synthesize pneumonia examples. Code uses MONAI and PyTorch, requires 512×512 inputs, can generate thousands of images, but requires IRB approvals, standardization, clinical QC, and expert validation before clinical use.
AI-written summary · about
chapters (9)
- 0:00AI in rare disease diagnosis
- 1:12Medical imaging challenges
- 4:29Synthetic image data benefits
- 6:15Image-to-image vs text-to-image
- 9:042D chest X-ray generation demo
- 17:58Text-to-image for pneumonia detection
- 24:44Image-to-image vs text-to-image pros/cons
- 26:12Clinical synthetic data steps
- 28:20Conclusion and future
related talks
30m🤯 No data? No problem! Synthetic data to the rescue
55mDeep Learning in Medical Image Diagnostics
27mAI for social good
28mMachine Learning & Healthcare
41mBuilding Generative Image & Video models at Scale
09mLightning Talk: A Novel Domain Generalization Technique for Medical Imaging Using...
36mResponsible AI
18mMedical AI Beyond Prediction: From Normative Learning to Agentic Healthcare AI