Translating Human-Centred Artificial Intelligence for Clinical Decision Support Systems into Practice: A Medical Retina Case Study.

Artificial intelligence has potential to enhance healthcare outcomes through applications such as early diagnosis and enhanced treatment planning but can only deliver such potential if it is integrated into clinical practice. Using eyecare as a case study, my research explores optometrists' requirements for an AI clinical decision support system (AI-CDSS) and addresses the 'gap' between AI design motivated by research success vs clinical application. For example, findings from an interview study with 20 optometrists highlighted that clinicians' risk-adverse tendencies can significantly affect their interpretation of AI outputs when making clinical decisions. The way in which outputs are presented should neither encourage suboptimal risk-averse behaviours nor convey misleading information. In the latter stages of my Ph.D., I aim to further investigate how the level of risk associated with clinical decisions can affect interpretations of AI support, as well as testing available methods for improving the interpretability of AI outputs.