Designing a human-centered AI solution with radiologists to improve prostate cancer diagnosis

Abstract Prostate cancer is a significant health concern with rising incidence rates in Germany. While recent research highlights the potential of artificial intelligence (AI) to support radiologists by increasing diagnostic accuracy and efficiency, the successful integration of such technology depends on carefully designed, practice-oriented approaches. This paper presents a design case study (DCS) on a human-centered artificial intelligence (HCAI) solution to assist radiologists in diagnosing prostate cancer. In our contribution, we emphasize that by combining human-centered design (HCD) and practice-centered computing, we can contribute towards a seamless and usable integration of AI into healthcare. Through interdisciplinary collaboration among radiologists, user experience (UX) designers, AI developers, and human-computer interaction (HCI) researchers, we employed Design Thinking (DT) to iteratively develop low-, mid-, and high-fidelity prototypes of an AI solution for improving diagnostic workflow, which were evaluated by radiologists. Our findings advance the state-of-the-art in computing support for prostate cancer diagnosis by introducing an artifact designed with both HCD and practice-centered computing in mind.