Multi-task learning for medical foundation models

Abstract

To address the challenge of pretraining foundational models with large datasets, a multi-task approach is proposed, thus helping to overcome the data scarcity problem in biomedical imaging.