Abstract
Healthcare is, traditionally, a knowledge-driven enterprise with an enormous amount of data - both structured and unstructured. These data can impact positively on the development of data-driven health care including precision medicine and precision public health. In recent years, large scale medical/clinical datasets, such as “omics” data and radiology reports are increasingly available. We have also witnessed an increasing number of successful AI/ML applications using such datasets to address problems such as drug repurposing and preliminary screening of radiology reports. To facilitate the adoption of such AI/ML in practice, we have simultaneously witnessed an increasing adoption/innovation of using explainability methods to analyze/present AI for Health. In this deep learning era, What is the current status of AI/ML applications in healthcare? What are the standard methods of explaining such AI models for healthcare? What are the roles of causality in AI/ML practices? What are the state-of-the-art developments in causal AI in health and health care domains? What are the limitations and how are the different facets of trust and explanations (see figure 1 below) being addressed in practice? Can knowledge-backed AI lead to more robust and interpretable models? How do data scientists and physicians apply this knowledge in collaboration and via human-centered AI approaches to further the field and improve healthcare? How are regulatory requirements for transparency and trustworthiness of models and data privacy being defined and how can they be fulfilled? After witnessing so many great achievements from deep learning lately, we propose to invite world-leading experts from both data science and healthcare to discuss and debate the path forward for practical applications of AI/ML in healthcare, including demos, early work, and critiques on the current state and the path forward for explainability and trustworthiness in AI. More specifically, we plan to attract high-quality original research from emerging areas with significant implications in healthcare and invite open discussions on controversial yet crucial topics regarding healthcare transformation.