PET imaging has become an essential tool in the management of patients, particularly in oncology. The limited number of nuclear medicine physicians will become a limiting factor in a few years, given the significant increase in the number of examinations prescribed. Automated assistance in sorting relevant examinations before human analysis will therefore be relevant to overcome this imbalance, and allow imaging experts to focus their interpretation time on complex examinations.
The objective of CLARITI is to develop an automated solution for the selective sorting of FDG PET/CT scans (normal vs. abnormal) by taking advantage of recent advances in AI / supervised deep learning. The expected results are automatic classification performances close to human classifications of FDG PET scans.