AI Deployment Journey in Healthcare: Governance, Design, and Adoption

This project, awarded by the ICON grant of Tilburg University, considers AI deployment in clinical workflow as a journey with critical components of governance, design, and adoption. The aim of the project is to address the challenges associated with successful implementation of AI applications in healthcare. Part of this project will address the fundamental issue of transparency of AI in healthcare. Common AI approaches apply deep learning mechanisms to tackle a predefined problem. An inherent characteristic of deep learning is its opaque nature due to its self-learning mechanism. In other words, the AI system adjusts its parameters to fit the data. Why and how these parameters are changed is not necessarily known to the user nor to the developer. In healthcare, this kind of black box models is challenging for making medical related decisions as patient safety might be at stake due to the unknown algorithmic process.

In this ICON AI Deployment project, we aim to tackle these issues from governance, design, and adoption perspectives. Specifically, we are motivated to develop an integral system of practices, processes, and technological tools to support healthcare organizations’ deployment of AI technologies. The proposed three content work packages will look at i) AI governance and explainability to establish novel frameworks, policies, and standards that ensure effective deployment of AI in healthcare; ii) AI design and deployment to develop innovative AI algorithm, interfaces and reporting structures tailored to address the unique challenges in engaging clinicians to give feedback to an AI model; iii) the role of trust in AI adoption in order to identify and implement trust-building mechanisms that increase acceptance and deployment of AI systems among healthcare professionals, patients, and other stakeholders.