A team from the Department of Electronics, Information and Bioengineering – Politecnico di Milano, led by researcher Andrea Moglia, has developed the first online application that helps identify which Artificial Intelligence model is most suitable for generating 3D images of each individual organ. In this way, patient care becomes more precise and reliable.
The tool stems from a study published in the prestigious journal Information Fusion, which examined both general-purpose and organ-specific AI models. It is designed for healthcare professionals, technicians who generate images of organs, lesions, or fractures, and physicians who interpret those images for surgery or therapy planning.
The free online application can be browsed starting from specific organs or anatomical regions such as the chest, neck, or abdomen. Once the area of interest is selected, the tool lists all existing AI models that have been tested on available image datasets. The models can be ranked by dataset performance, from the most to the least effective. The organ selection is extremely detailed — for instance, users can choose individual vertebrae or heart ventricles. Another key feature is the ability to sort models based on their performance in generating images of tumors and lesions, including those caused by strokes or ischemias.
For some time now, doctors and technicians have been using AI models to produce images of organs or lesions — a process technically known as segmentation. This involves delineating the contour of a given object in a 2D image to reconstruct it in 3D. In medicine, this means combining multiple X-ray or CT images and highlighting the organ or lesion of interest with a colored outline. Using AI models makes this process faster and reduces human bias and error.
The project also involved Prof. Luca Mainardi and PhD candidate Matteo Leccardi, both from the same department, as well as Prof. Pietro Cerveri from the EssilorLuxottica Smart Eyewear Lab, affiliated with Politecnico di Milano. The work was partly funded through the National Recovery and Resilience Plan (PNRR) by the Future Artificial Intelligence Research Foundation (FAIR).