Carry the Two Emerging Technologies Episode 2: Computation Medicine
4 snips
Aug 14, 2025 Yixiang Deng, an assistant professor specializing in machine learning for diabetes management, and Fides Schwartz, a radiologist focusing on advanced CT imaging, share their insights on computational medicine. They discuss innovative glucose monitoring using machine learning, aiming to create an artificial pancreas. The duo also explores the significance of patient involvement in AI-driven health decisions and the balance of human expertise in radiology amid rising AI technology. Their engaging banter illustrates the fascinating intersection of mathematics and healthcare.
AI Snips
Chapters
Transcript
Episode notes
Personalized Glucose Prediction Works
- Diabetes causes unstable glucose fluctuations that require personalized management.
- Yixiang Deng's team uses time-series machine learning to predict glucose and improve care.
Fine-Tune Models Per Patient
- Fine-tune general deep learning models on each patient's data to improve predictions.
- Use RNNs, CNNs, and transformers specialized for sequential glucose data.
Public Data Enables Better Models
- Public datasets like Ohio T1DM enable reproducible research and benchmarking.
- Data imbalance (few low-glucose events) requires augmentation to avoid biased models.