MEDI-SLATE: medical imaging slide-lecture aligned teaching ensemble

Area of Science:

• Biomedical Engineering
• Medical Imaging Education
• Educational Technology

Background:

• Undergraduate medical imaging education relies heavily on slide-based lectures.
• Existing resources lack openly available, aligned slide-narration datasets for research.
• There is a need for high-quality, structured data to advance medical imaging pedagogy.

Purpose of the Study:

• To introduce MEDI-SLATE, a novel dataset of medical imaging lecture slides paired with refined narration.
• To provide a comprehensive resource for research in medical imaging education and multimodal learning.
• To facilitate the development of AI-assisted instructional tools for medical imaging.

Main Methods:

• Constructed MEDI-SLATE from a complete undergraduate biomedical engineering medical imaging course.
• Collected 1117 high-resolution slides and paired them with narration refined via automatic speech recognition and manual cleanup.
• Included lecture-level difficulty tags, key ideas, common student misunderstandings, and practice questions.

Main Results:

• MEDI-SLATE contains 1117 slide-narration pairs covering diverse medical imaging topics.
• The dataset includes metadata such as difficulty tags and common student misconceptions.
• A reproducible preprocessing pipeline for data extraction, refinement, and alignment is provided.

Conclusions:

• MEDI-SLATE is a high-fidelity, openly available resource for medical imaging education.
• The dataset supports curriculum development, multimodal learning research, and AI tool creation.
• All data and code are released for transparent use and future extensions.