Researchers at Ben-Gurion University of the Negev have developed DISCOVER, a computational tool that "reverse engineers" AI decisions in medical imaging by breaking down images into clinically meaningful components. This innovation helps explain the reasoning behind AI models, essential in fields like medicine where understanding decisions is critical.

In collaboration with Israeli startup AIVF, the method was tested on IVF embryo images, MRI scans of Alzheimer's patients, and standard photos to interpret AI decision-making. DISCOVER uses generative AI to create synthetic images that highlight specific features crucial to the AI's conclusions, providing insights for clinical applications and improving diagnostic accuracy.

The technology demonstrates how AI can identify key medical features that may go unnoticed by human experts, offering potential advancements in various fields, including fertility treatments and disease diagnosis. The research, led by doctoral student Oded Rotem under Prof. Assaf Zaritsky, was recently published in Nature Communications and supported by the Rosetrees Trust and the Israeli Council for Higher Education.