BMI 702 | Biomedical Artificial Intelligence

Harvard - Foundations of Biomedical Informatics II, Spring 2023

Artificial intelligence is poised to enable breakthroughs in science and reshape medicine. This introductory course provides a survey of artificial intelligence for biomedical informatics, covering methods for key data modalities: clinical data, networks, language, and images. It introduces machine learning problems from a practical perspective, focusing on tasks that drive the adoption of machine learning in biology and medicine.

It outlines foundational algorithms and emphasizes the subtleties of working with biomedical data and ways to evaluate and transition machine learning methods into biomedical and clinical implementation. An important consideration in this course is the broader impact of artificial intelligence, particularly topics of bias and fairness, interpretability, and ethical and legal considerations when dealing with artificial intelligence.

Marinka Zitnik

marinka@hms.harvard.edu

Office Hours: Mo, 3pm - 4pm

Week 1

Course overview and introduction to biomedical AI

Jan 23

LectureWhat makes biomedical problems unique: Slides, Reading List

Jan 24

QuizWeek 2 pre-class quiz (due Jan 29): Canvas

Week 2

Clinical research using EHR data, subtype discovery, disease diagnosis and prognosis prediction

Jan 30

Module 1LectureClinical AI Part I: Slides, Reading List

Jan 31

QuizWeek 3 pre-class quiz (due Feb 5): Canvas

Week 3

Multi-institutional EHR systems, transfer learning, federated learning, clinical workflows

Feb 6

Module 1LectureClinical AI Part II: Slides, Reading List

Feb 7

QuizWeek 4 pre-class quiz (due Feb 12): Canvas

Week 4

Interpretability and explainability in biomedical AI

Feb 13

Module 2LectureTrustworthy AI Part I: Slides, Reading List

Feb 13

QuizWeek 5 pre-class quiz (due Feb 26): Canvas

Week 5

Bias and fairness in biomedical AI

Feb 27

Module 2LectureTrustworthy AI Part II: Slides, Reading List

Feb 28

QuizWeek 6 pre-class quiz (due Mar 5): Canvas

Mar 1

PSet releasedPSet 1: Bias, explainability, and fairness: Canvas

Week 6

Foundations of geometric deep learning, graph representation learning, link prediction, node classification, graph clustering, graph classification, semi-supervised learning, label propagation, network medicine, disease modules and endotypes

Mar 6

Module 3LectureBiomedical graph learning Part I: Slides, Reading List

Mar 7

QuizWeek 7 pre-class quiz (due Mar 12): Canvas

Week 7

Machine learning with heterogeneous graphs, multimodal learning, graph neural networks, knowledge graph embeddings, reasoning over knowledge graphs

Mar 20

Module 3LectureBiomedical graph learning Part II: Slides - Part 1, Slides - Part 2, Reading List

Mar 21

QuizWeek 8 pre-class quiz (due Mar 26): Canvas

Mar 22

PSet duePSet 1: Bias, explainability, and fairness: Canvas

Week 8

Foundations of natural language processing and understanding

Mar 27

Module 4LectureMedical language modeling Part I: Slides, Reading List

Mar 28

QuizWeek 9 pre-class quiz (due Apr 2): Canvas

Mar 29

PSet releasedPSet 2: Biomedical networks and graph embeddings: Canvas

Week 9

Clinical trial site identification, patient trial matching, clinical trial recruitment

Apr 3

Module 4LectureMedical language modeling Part II: Slides, Reading List

Apr 4

QuizWeek 10 pre-class quiz (due Apr 9): Canvas

Week 10

Foundations of biomedical imaging, self-supervised learning, analysis of radiology images

Apr 10

Module 5LectureBiomedical imaging Part I: Slides, Reading List

Apr 11

QuizWeek 11 pre-class quiz (due Apr 16): Canvas

Apr 12

PSet duePSet 2: Biomedical networks and graph embeddings: Canvas

Week 11

Multimodal learning, analysis of histopathology slides, quantitative pathology in cancer diagnosis and prognosis

Apr 17

Module 5LectureBiomedical imaging Part II: Slides, Reading List

Apr 18

QuizWeek 12 pre-class quiz (due Apr 23): Canvas

Apr 19

PSet releasedPSet 3: Biomedical imaging methods and applications: Canvas

Week 12

Overview of drug discovery and development, AI-guided drug design, small-molecule generation, molecule optimization, identification and characterization of therapeutic targets, high-throughput chemical and genetic perturbations

Apr 24