The Field Guide Mindset: You will be asked to use AI tools that you did not design, do not fully understand, and do not have time to rigorously evaluate. This course teaches you to build those tools, evaluate them with discipline, and write the field guide someone else could use.
Welcome to MPHY 6120: Artificial Intelligence for Medicine! This course is designed for a broad audience including medical physicists, MD/PhDs, bioengineers, computer scientists, and translational researchers. Throughout the semester, you will learn to:
The goal is to produce graduates who can both build models and write the field guide that tells a real clinic how to use them safely.
Course Logistics - Spring 2026
We encourage in-person attendance for lab sessions. Contact the instructor if you cannot attend a specific day.
Prerequisites: Basic Python programming and familiarity with linear algebra and statistics. If you need a refresher, we recommend Penn’s free Introduction to Python Programming on Coursera.
The fields of data science and artificial intelligence are transforming medicine. This course provides a pragmatic field guide for researchers and clinicians who want to meaningfully contribute to clinical AI applications.
We cover:
Models vs. Systems — Models are the easy part. Deployment, workflow, governance, and monitoring are the hard part.
Metrics vs. Readiness — A good ROC curve or Dice score is not the same as a tool ready for the clinic. Local validation, workflow fit, and oversight rules matter.
Governance as Quantitative Discipline — Governance is not bureaucracy. It is constraints (what must always be true), experiments (acceptance tests), and logs (traceability).
Use Cases at Different Maturity Levels — Students learn to place AI use cases on a maturity spectrum, from research to clinical deployment.
Field Guide Mindset — Always ask: “If I were a busy community clinician, what would I need to know to use this tool?”
We meet twice weekly for 90-minute sessions:
Primary Text:
About the Little Book: This is a living document being developed alongside the course. You’ll receive draft chapters as we progress through modules. Your feedback and questions will shape the final version. Think of it as our collective field guide to the field—and a model for the field guides you’ll write for your own projects.
Code & Resources:
Optional References:
| Component | Weight | Description |
|---|---|---|
| Homework Assignments | 50% | Coding tasks and short writing assignments |
| Midterm Project | 20% | Medical imaging case study + mini field guide |
| Final Project | 20% | Technical artifact + full field guide + presentation to Review Board |
| Participation | 10% | In-class discussions, peer feedback |
| Score | Grade |
|---|---|
| ≥ 97 | A+ |
| 93–97 | A |
| 90–93 | A− |
| 87–90 | B+ |
| 83–87 | B |
| 80–83 | B− |
| 75–80 | C+ |
| 70–75 | C |
| 65–70 | C− |
| 50–65 | D |
| < 50 | F |
Projects have two parts:
A document for non-experts that could actually be handed to a busy clinician. This is graded separately from your technical work.
| Section | What to Include | Points |
|---|---|---|
| 1. Tool Summary | Plain-language description a non-technical clinician could understand in 2 minutes | 10 |
| 2. Intended Use | Specific clinical context, patient population, workflow placement | 10 |
| 3. Performance Summary | Key metrics, what they mean clinically, comparison to alternatives | 15 |
| 4. Limitations & Failure Modes | When it breaks, edge cases, known biases | 15 |
| 5. Human Oversight Rules | When to review, when to override, red flags to watch for | 15 |
| 6. Local Validation Plan | How you would test this at a new site (cohort, metrics, thresholds) | 15 |
| 7. Monitoring Plan | What to track, how often, who reviews, escalation criteria | 10 |
| 8. Patient Explanation | One paragraph for informed consent / patient questions | 10 |
Total: 100 points (Field guide graded separately from technical artifact)
Final project presentations simulate a real clinical AI deployment review. A Review Board of senior clinical and technical leaders will attend—including clinical informaticists, department leadership, and industry experts. They will ask questions as if you were proposing to deploy your tool at their institution.
This is intentional: bridging technical depth with clinical context and leadership communication is the whole point. You’re not just building a model; you’re learning to advocate for its responsible deployment.
Students choose one of three tracks based on their background and interests. You select a track for your midterm project and can switch or continue for the final project.
Track A: Medical Imaging
Track B: Clinical NLP
Track C: Structured Clinical Data
All code assignments are submitted via GitHub Classroom:
We use GitHub (not Gradescope) because version control is an essential skill for medical AI development. Your commit history shows your problem-solving process and distinguishes your work from AI-generated code. Grades are recorded in Canvas.
We encourage the use of AI coding assistants (GitHub Copilot, ChatGPT, Claude) to enhance your learning. However:
Instructor: TBD (will be announced first week of class)
TA: TBD
Office hours are drop-in; no appointment needed. For complex questions, email ahead so we can prepare.
Thank you for visiting the course page. We look forward to a dynamic semester exploring the latest developments in AI for medicine!