Towards Trustworthy Machine Learning in Healthcare: Addressing Challenges in Explainability, Fairness, and Privacy through Interdisciplinary Collaboration

Mashrin Srivastava

Abstract: Machine learning (ML) has shown great potential in various healthcare tasks, with some models surpassing human performance. However, its application in real-world healthcare scenarios is limited due to the lack of trustworthiness in ML models. This paper investigates the challenges of explainability, fairness, privacy, and generalization to out-of-distribution samples in healthcare ML. We examine the current state-of-the-art methods to address these challenges and propose potential solutions through interdisciplinary collaboration between ML researchers, clinical practitioners, and medical imaging experts. By integrating expertise from diverse backgrounds, we aim to advance the development of trustworthy ML models in healthcare and facilitate their translation into clinical practice.