AI-Assisted PET Imaging Identifies Hidden Brain Tumour Lesion
Philipp Lohmann, Robin Gutsche, Jan-Michael Werner, N. Jon Shah, Karl-Josef Langen and Norbert Galldiks
February 15th 2024
A 43-year-old patient with inconclusive results from structural MRI scans underwent additional O-(2-[18F]fluoroethyl)-L-tyrosine (FET) PET imaging for further diagnosis of a suspected glioma, revealing unexpected insights into his condition. The PET scan showed a lesion with slightly increased FET uptake that was segmented by a human expert. Interestingly, besides this lesion, the artificial intelligence (AI) algorithm JuST_BrainPET identified an additional lesion, which did not show pathological tracer uptake and was hence not segmented by the human expert. This lesion subsequently progressed into a contrast-enhancing and metabolically active glioblastoma detected during follow-up scans four months later.
The AI tool correctly predicted a pathological process at an early stage of the disease, which could have influenced diagnostic and treatment decisions, such as guiding biopsy or determining the target volume for radiotherapy.
This incidental finding underscores the potential of AI-based decision support for patient management based on amino acid PET.
Original publication: Example of Artificial Intelligence–Based Decision Support for Amino Acid PET: Early Prediction of Suspected Brain Tumor Foci for Patient Management
JuST_BrainPET: https://github.com/MIC-DKFZ/nnUNet/tree/nnunetv1#useful-resources