VIT Lessons Learned - Calibration, Realism, Portability, and Reproducibility
Date: August 20th, 2021 (12:00 PM EST)
Speakers: Andrew Maidment (University of Pennsylvania)
Abstract: Dr. Maidment will present his experience in developing and disseminating tools for virtual imaging trials (VITs) in breast imaging. He will explain the lessons learned from this experience, with a focus on producing accurate, reproducible results. He will cover the three foundational considerations to this end – proposing tractable problems for which VIT data is likely to predict performance with human data, ensuring that the data presented to the software elicits appropriate and quantifiable results, and standardizing methods so that VIT code can become portable and reproducible for diverse groups of researchers.