The course will be held by prof. Stefano Diciotti and is part of the training provided for IBES PhD Students (a.y.2025/2026) within the Bioengineering (BIO) curriculum.
Date:
Event location: LIB-M - UOS - Via dell’Università, 50 - Cesena
Type: IBES Course
ABSTRACT
This short course provides the essential tools to design, evaluate, and report AI models in Medicine with a focus on reproducibility and the prevention of data leakage. Participants will learn how common methodological pitfalls—such as using slice-level instead of subject-level data splits in brain MRI—can lead to artificially inflated performance. The course demonstrates how to implement leakage-safe validation strategies and properly handle data harmonization steps (e.g., avoiding the misuse of ComBat before train-test splitting) using scikit-learn pipelines to remove optimistic bias. Finally, the course emphasizes transparent reporting practices, including the use of structured documentation such as Model Info Sheets.
This course is part of the IBES training activities for the a.y. 2025/2026.
Further information about the schedule, the minimum attendance rate and the exam format is outlined HERE and detailed by the instructor during the course.
This form is meant to help instructors estimate the number of participants and organize the activity accordingly. It is not binding, but we kindly ask you to complete it responsibly.