The ARAMIS project (Emergence, Cancéropôle Ile-de-France) aims to develop an artificial intelligence algorithm capable of automatically analyzing FDG PET scans in oncology. The goal is to better understand and predict the cancer progression by extracting relevant information from the images, such as the features from tumor lesions and non-tumor tissues. Using an unsupervised learning algorithm, ARAMIS will explore a latent space that can structure and interpret the data from PET scans.
The project relies on a retrospective database of over 4,000 PET scans from patients with various types of cancer. The analysis of this latent space will allow the development of a classifier capable of automatically identifying the organ of origin of the cancer and addressing other complex tasks, such as predicting treatment response or patient survival.