New guideline improves reporting of AI diagnostic accuracy studies

A new guideline, STARD-AI, has been developed to improve the reporting of diagnostic accuracy studies involving artificial intelligence. The STARD-AI statement provides a minimum set of criteria for comprehensive reporting of AI-centered diagnostic studies, addressing unique considerations like dataset practices, algorithmic bias, and fairness. It was developed through a multistage process involving over 240 international stakeholders. This guideline aims to enhance transparency and allow stakeholders to better evaluate the biases, applicability, and generalizability of AI diagnostic study findings.