Artificial Intelligence in Medicine II

Harvard - BMIF 203 and BMI 702, Spring 2025

Advances in AI will have a broad and profound impact on science and medicine, offering new approaches to transform medical research and practice. This course provides a comprehensive overview of cutting-edge AI paradigms, including self-supervised learning, generative models, and multimodal techniques that integrate diverse data types. Beyond foundational methods, the course dives into a range of real-world applications in natural language processing, medical image analysis, relational and structure understanding, and longitudinal patient data.

Faculty Instructor

Marinka Zitnik

marinka@hms.harvard.edu

Office Hours: Mon, 12pm - 1pm, Countway 309

Curriculum Fellow

Grey Kuling

Grey_Kuling@hms.harvard.edu

Week 1

Course overview, Introduction to NLP, NLP in a clinical setting, Medical terminology challenges, Concept extraction from clinical notes, Note summarization, Clinical trial matching

Jan 28

LectureNLP I: Slides, Reading List

Jan 29

QuizWeek 2 pre-class quiz (due Feb 4): Canvas

Week 2

Embeddings and their role in NLP, Transformers and BERT, Hugging Face library for NLP applications, Clinical BERT and RNNs, Stack-encoder and Stack-decoder architectures, De-identification methods, LLM-based medical question-answering. Lecture by Dr. Carlos Morato

Feb 4

LectureNLP II: Embeddings & Transformers: Slides, Reading List

Feb 5

QuizWeek 3 pre-class quiz (due Feb 11): Canvas

Week 3

Variational Autoencoders (VAEs), Generative Adversarial Networks (GANs), GenAI Fundamentals (training, optimization and RLHF), LLMs and Multimodal LLMs, Grounding and RAG, Healthcare applications of generative AI, Synthetic data generation, Data privacy concerns. Lecture by Dr. Carlos Morato

Feb 11

LectureGenerative AI: Slides, Reading List

Feb 11

Course ProjectProject Proposal Due for Peer Review: Canvas

Feb 12

QuizWeek 4 pre-class quiz (due Feb 18): Canvas

Week 4

Designing LLM-driven agents to answer complex clinical queries with evidence-backed reasoning, Strategies to evaluate accuracy, robustness, and interpretability in high-stakes medical contexts, Case studies of LLM-based agents in clinical decision-making, drug discovery, and patient triage, Emerging trends, such as real-time conversational agents, collaborative multi-agent systems, and autonomous discovery

Feb 18

LectureAgentic AI: Slides, Reading List

Feb 18

Course ProjectProject Proposal Due: Canvas

Feb 19

QuizWeek 5 pre-class quiz (due Feb 25): Canvas

Week 5

Understand the various types of medical imaging (radiology, oncology, pathology, and other imaging modalities), Learn the basic tasks in AI for medical imaging: classification, regression, and segmentation, Explore how AI is applied in different medical imaging contexts, Understand convolutional neural networks (CNNs) and their role in medical imaging, Learn segmentation techniques, focusing on the U-Net architecture, Apply CNNs to biomedical image segmentation tasks, including preprocessing and evaluation. Lecture by Dr. Grey Kuling

Feb 25

LectureMedical Imaging I: Slides, Reading List

Feb 26

QuizWeek 6 pre-class quiz (due March 4): Canvas

Week 6

Explore advanced applications of AI in medical imaging, with a focus on generalist medical AI models, Understand the development and validation of medical imaging interpretation models, Discuss best practices for evaluating medical imaging AI models, with emphasis on robustness and performance across diverse populations. Lecture by Prof. Pranav Rajpurkar

Mar 4

LectureMedical Imaging II: Slides, Reading List

Mar 5

QuizWeek 7 pre-class quiz (due Mar 11): Canvas

Week 7

Explainability and interpretability in medical AI, Feature importance and Shapley values, Interpreting CNNs with heatmaps and other methods, Discussion: Is explainability critical or overrated?

Mar 11

LectureTrustworthy AI: Slides, Reading List

Mar 11

Course ProjectMid-Term Student Presentations: Canvas

Mar 12

QuizWeek 8 pre-class quiz (due Mar 25): Canvas

Week 8

Structured and relational datasets, Introduction to Graph Neural Networks (GNNs), Graph transformers, Combining multiple data modalities with GNNs

Mar 25

LectureNetworks I: Slides, Reading List

Mar 26

QuizWeek 9 pre-class quiz (due Apr 1): Canvas

Week 9

Knowledge graphs, Building multimodal knowledge graphs, Structure-inducing pre-training, Knowledge-based foundation models

Apr 1

LectureNetworks II: Slides, Reading List

Apr 2

QuizWeek 10 pre-class quiz (due Apr 8): Canvas

Week 10

AI for protein structure prediction, Drug discovery and therapeutic science, Structure- and sequence-based co-design, Biological foundation models

Apr 8

LectureMolecular AI: Slides, Reading List

Apr 9

QuizWeek 11 pre-class quiz (due Apr 15): Canvas

Week 11

Addressing label scarcity in medical data, Semi-supervised and self-supervised learning, Combining image and text modalities in AI (ConVIRT, CLIP), Exploring models like CheXzero and DALL-E

Apr 15

LectureMultimodal AI: Slides, Reading List

Apr 16

QuizWeek 12 pre-class quiz (due Apr 22): Canvas

Week 12

Regulation of AI algorithms and devices in healthcare, FDA oversight and liability concerns, Prospective clinical trials for AI systems and AI-augmented devices. Lecture by Prof. Sara Gerke

April 22

LectureEthical and Legal Considerations: Slides, Reading List

Apr 23

QuizWeek 13 pre-class quiz (due Apr 29): Canvas

Week 13

Digital biomarkers and disease progression tracking, Patient/disease progression modeling using transformers, In-home health and disease monitoring systems, Intelligent and accessible AI systems for healthcare delivery

Apr 29

LectureTime Series & Sensors: Slides, Reading List

Apr 29

Course ProjectFinal Student Presentations: Canvas