Machine Learning for Healthcare
Course facts
- Payment: Free, no certificate
- Credits: Non Credit
- Learning style: Self-paced, start anytime
- Resources: Lecture Notes; Lecture Videos; Projects
MIT OpenCourseWare details
- As taught in: Spring 2019
- Department: Electrical Engineering and Computer Science, Health Sciences and Technology
| Course Details | ||||||
|---|---|---|---|---|---|---|
|
||||||
| Outcomes | ||||||
|
||||||
| Price & Terms | ||||||
|
||||||
| Technical | ||||||
|
||||||
| Specification | ||||||
|
||||||
Machine learning on clinical data: risk stratification, clinical NLP, imaging, causal inference, precision medicine, fairness, and interpretability.
Applying machine learning to medicine starts with understanding the nature of clinical data and how clinical care works. From there the course turns to risk stratification, physiological time series, clinical text, diagnosis in cardiology, pathology and mammography, causal inference, reinforcement learning, disease progression and precision medicine, with sessions on regulation, fairness, dataset shift and interpretability.
What You Will Achieve
- Describe the nature of clinical data and the clinical care it comes from
- Build risk stratification and disease progression models from clinical data
- Apply causal inference and reinforcement learning to treatment questions
- Evaluate models for fairness, robustness to dataset shift, and interpretability
Program Overview
Clinical Care and Clinical Data
Why healthcare is a different setting for machine learning.
- What makes healthcare unique
- An overview of clinical care
- Clinical data in depth
- Risk stratification and physiological time series
Language, Diagnosis and Imaging
Machine learning for clinical text and medical specialties.
- Natural language processing of clinical notes
- Translating technology into the clinic
- Machine learning for cardiology and differential diagnosis
- Machine learning for pathology and mammography
Causality, Treatment and Trust
Deciding on treatments and deploying models safely.
- Causal inference and reinforcement learning
- Disease progression, subtyping, and precision medicine
- Automating clinical workflows and regulation of ML and AI
- Fairness, robustness to dataset shift, and interpretability
Course Completion
Job Outlook
Machine learning for healthcare is used in clinical informatics, health data science, medical AI product teams, hospital analytics and biomedical research.
Good for: Learners who want to apply machine learning responsibly to clinical data and medical decisions, at a graduate level.
Keep in mind: It was taught in 2019, so check current guidance on regulating machine learning and AI in the US; the materials are lecture videos, notes and project guidelines.
Bottom line: A graduate course with many guest lecturers that treats fairness, causality and dataset shift as seriously as model accuracy.
Frequently Asked Questions
Which medical areas does the course look at?
Guest lectures and sessions cover cardiology, differential diagnosis, pathology and mammography, alongside clinical notes, physiological time series and treatment strategies.
What is the course project?
Groups write a proposal stating the problem, data, methods and evaluation plan, present a poster, and submit a final report, ideally with open-source code that lets others reproduce the results.
Does the course cover fairness and interpretability?
Yes. Closing sessions treat algorithmic fairness, robustness to dataset shift and interpretability, after sessions on regulation of ML and AI and on human subjects research.
User Reviews
Write your review
You must be logged in to post a review.
Reviews can only be posted by registered learners. Log in or create a free account, then you will return here to post.
There are no reviews yet.