AI-Powered Pathology Education
A Proposal for the UCSF Health Professions Education Pathway
Charlie Herndon, MD
PGY-1, Department of Pathology
The Problem: Static & Unsustainable Learning
- Information Overload: The scope of pathology is too vast for any individual to master with current tools.
- Inefficient Search: Finding specific, high-quality visual examples is a laborious, manual process for trainees.
- Static Resources: Textbooks and lectures are one-size-fits-all and quickly become outdated.
- Redundant Content Creation: Educators continually recreate similar materials, a time-consuming and unsustainable effort.
The Vision: A Dynamic Learning Ecosystem
To transition pathology education from a library of static resources to a dynamic, AI-driven engine that generates personalized learning experiences on demand.
Instead of searching for content, learners ask for what they need, and the system creates it instantly—tailored to their style and knowledge gaps.
Core Innovation: Structured Image-Text Pairs
The foundation is a machine-readable dataset linking visual data to semantic meaning. We create this by extracting figure-caption pairs from high-quality sources.
Curated Textbooks
AI-Powered Extraction
Image-Text Pairs
Structured Database
The Application: Generative Learning Engine
This structured data fuels an AI that can build bespoke educational materials. A user can ask:
"Generate a 10-question quiz on glial neoplasms with classic images," or "Create flashcards for my weak points in GI pathology."
- Personalized Outputs: Quizzes, flashcards, slide decks, and summaries.
- Adaptive Learning: Content is tailored to user's knowledge level and format preference.
- Live Demo: See a prototype quiz generator in action here.
Challenge: Navigating Intellectual Property
The most significant barrier to academic legitimacy and scaling is copyright. My current proof-of-concept uses watermarked images, which is not a long-term solution.
A core goal of this research track is to solve this challenge by exploring partnerships, open-access sources, or institutional slide archives.
The Ask: A Home in the HPE Pathway
This project's goals align perfectly with the mission of the Health Professions Education (HPE) Pathway.
- Curriculum Development: It's a new model for creating and delivering medical curricula.
- Implementation Science: Provides a perfect testbed for studying how new educational tools are adopted and impact learning.
- Mentorship: I am seeking guidance from education experts to rigorously develop, validate, and publish this work.
Proposed Research Plan (within HPE)
- Phase 1 (MVP): Build the extraction pipeline for one subspecialty (e.g., Neuropathology) and deploy a prototype quiz generator for pilot testing among residents.
- Phase 2 (Validation): Conduct formal A/B testing to measure learning outcomes (engagement, retention, performance) against traditional methods.
- Phase 3 (Publish & Scale): Publish findings in medical education journals and develop a strategy for scaling the dataset by solving the IP challenge.
Next Steps & Discussion
I am seeking guidance and institutional support to formalize this project.
- How can this project be best structured within the HPE Pathway or a formal degree (MS/PhD)?
- What are the opportunities to connect with mentors in the Academy of Medical Educators?
- What is the best approach for securing pilot data and exploring funding opportunities within UCSF?