Overview
Medical AI: Foundations and Frontiers examines the foundations and emerging directions of artificial intelligence in medicine. The course brings together medical foundation models, causal reasoning, multimodal language models, and medical image analysis, with an emphasis on understanding methodological assumptions, evaluating empirical evidence, and critically reading contemporary research.
Logistics
- Schedule: Tuesdays & Fridays, 14:00–16:35
- Dates: November 10, 2026 – January 1, 2027
- Location: Huaxi Campus, Weixin Building, Room 1-1（华西卫新大楼1-1）
Note: The January 1, 2027 session falls on New Year’s Day and is tentative. Any rescheduling will be announced after confirmation from the university.
Content
What is this course about?
Medical AI brings together machine learning, biomedical data, and medical research. Understanding its methods requires both a foundation in how models learn and an appreciation of the data, assumptions, and evaluation challenges that arise in medicine.
This course connects fundamental ideas with emerging research across five modules: Frontiers in Medical AI, Foundations of Modern AI, Causal Foundation Models, Multimodal LLM, and Medical Image AI. We will examine how these approaches represent medical information, how their performance is assessed, and how to read the research critically.
Prerequisites
The following background is recommended:
- Python programming. Comfort with basic Python and array-based computation. Familiarity with NumPy and PyTorch is helpful.
- Mathematics and statistics. Basic linear algebra, calculus, probability, and statistics.
- Machine learning. Familiarity with supervised learning, neural networks, model training, and evaluation.
- Medical context. Some familiarity with medical terminology, biomedical data, or medical images is helpful.
Syllabus
Five modules are followed by a concluding discussion. Remaining session topics: TBA.
- Frontiers in Medical AI: 2 sessions (6 teaching hours).
- Foundations of Modern AI: 4 sessions (12 teaching hours).
- Causal Foundation Models: 3 sessions (9 teaching hours).
- Multimodal LLM: 3 sessions (9 teaching hours).
- Medical Image AI: 3 sessions (9 teaching hours).
Concluding session: Summary and Discussion (3 teaching hours).
Schedule
Tuesdays & Fridays, 14:00–16:35. Each session is 3 teaching hours. Remaining topics and materials: TBA.
Scroll horizontally to view the full table.
| # | Teaching week | Date | Description | Course materials |
|---|---|---|---|---|
| 1 | 11 | Tue, Nov 10, 2026 | Frontiers in Medical AIMedical Foundation Models and Imaging AI | |
| 2 | 11 | Fri, Nov 13, 2026 | Frontiers in Medical AIClinical AI: From Data to Practice | |
| 3 | 12 | Tue, Nov 17, 2026 | Foundations of Modern AITransformers I: From Attention to ArchitectureCore topics: Attention; QKV; Multi-head Attention; Position Encoding; Transformer Blocks | SlidesNotesReadingsCode |
| 4 | 12 | Fri, Nov 20, 2026 | Foundations of Modern AITransformers II: From Pretraining to Foundation ModelsCore topics: Self-supervised Learning; Pretraining; ViT; LLM; Transfer Learning | SlidesNotesReadingsCode |
| 5 | 13 | Tue, Nov 24, 2026 | Foundations of Modern AIConvolutional Neural NetworksCore topics: Convolution; Inductive Bias; ResNet; U-Net; Visual Representation | SlidesNotesReadingsCode |
| 6 | 13 | Fri, Nov 27, 2026 | Foundations of Modern AIModern Generative Modeling: Diffusion and Flow MatchingCore topics: Score Matching; Diffusion; Flow Matching; Rectified Flow; DiT | SlidesNotesReadingsCode |
| 7 | 14 | Tue, Dec 1, 2026 | Causal Foundation ModelsTBA | SlidesNotesReadingsCode |
| 8 | 14 | Fri, Dec 4, 2026 | Causal Foundation ModelsTBA | SlidesNotesReadingsCode |
| 9 | 15 | Tue, Dec 8, 2026 | Causal Foundation ModelsTBA | SlidesNotesReadingsCode |
| 10 | 15 | Fri, Dec 11, 2026 | Multimodal LLMTBA | SlidesNotesReadingsCode |
| 11 | 16 | Tue, Dec 15, 2026 | Multimodal LLMTBA | SlidesNotesReadingsCode |
| 12 | 16 | Fri, Dec 18, 2026 | Multimodal LLMTBA | SlidesNotesReadingsCode |
| 13 | 17 | Tue, Dec 22, 2026 | Medical Image AITBA | SlidesNotesReadingsCode |
| 14 | 17 | Fri, Dec 25, 2026 | Medical Image AITBA | SlidesNotesReadingsCode |
| 15 | 18 | Tue, Dec 29, 2026 | Medical Image AITBA | SlidesNotesReadingsCode |
| 16 | 18 | Fri, Jan 1, 2027*Tentative | Summary and Discussion | SlidesNotesReadingsCode |
* January 1, 2027 (New Year’s Day): Tentative; awaiting confirmation from the university on possible rescheduling.
Slides, Notes, Readings, and Code links will be added when available.
Readings
Required readings are listed by session. Additional readings will be added as the course develops.
Session 1 — Tue, Nov 10, 2026
Medical Foundation Models and Imaging AI
- An open vision-language model for diverse medical applicationsSellergren, A., Kazemzadeh, S., Mahvar, F. et al. Nature Medicine (2026).Download PDF (3.3 MB)
- Large-scale esophageal cancer screening through noncontrast computed tomography and artificial intelligenceZhou, J., Guo, G., Yao, J. et al. Nature Medicine (2026).Download PDF (13.9 MB)
Session 2 — Fri, Nov 13, 2026
Clinical AI: From Data to Practice
- Computable longitudinal patient journeys from structured and unstructured EHR dataKim, E., Foty, R., Thakor, A. S. et al. Nature Medicine (2026).Download PDF (12.4 MB)
- Prospective evaluation of a large language model clinical decision support system in the emergency departmentLeibovitch, L., Ahituv, A., Gorenshtein, A. et al. Nature Medicine (2026).Download PDF (9.2 MB)