Development of an organ-specific 18-FDG PET normative atlas with clinical application to abnormal uptake detection

In clinical practice, 18-FDG PET/CT scans are widely used to detect tumoral or inflammatory pathologies through visual assessment and semi-quantitative metrics based on SUVs. These quantification methods are quick to obtain, but they require clinicians to manually draw regions of interest. Moreover, they focus only on the most intense voxels, which leads to less accurate, less reproducible, and less robust results. To address this, we developed a methodology for the automatic detection and quantification of metabolic abnormalities in 18-FDG PET/CT images. Specifically, we constructed an organ-by-organ normative atlas of 18-FDG uptake from a representative dataset of healthy subjects. This atlas enables the automatic identification of abnormal metabolic activity.

We validated this approach in patients with lung cancer undergoing immunotherapy. While immunotherapy has shown promising results across various cancers, including lung cancer, it is also associated with immune-related adverse events, emphasizing the need for their early detection. In this context, the normative atlas was applied to automatically identify both tumoral and inflammatory metabolic abnormalities in these patients.