Utilising Advanced AI Technology for the Accurate Evaluation of Brain Tumours in PET Imaging
17th October 2023
Talking to the Society of Nuclear Medicine and Molecular Imaging (SNMMI), INM-4 scientist Dr Philipp Lohmann describes a deep learning-based segmentation algorithm developed by the team. The state-of-the-art method enables a robust and fully automated volumetric evaluation of amino acid PET data in patients with gliomas, potentially improving diagnosis and treatment monitoring.
The developed algorithm was trained and fine-tuned on a large dataset consisting of 699 FET PET scans obtained at initial diagnosis or during follow-up from 555 brain tumour patients to automatically assess the metabolic tumour volume. The algorithm was then subsequently used to analyse data from a recent FET PET study on response assessment in glioblastoma patients undergoing adjuvant temozolomide chemotherapy. The algorithm's evaluation was compared to that of an experienced physician, as described in the study published in The Journal of Nuclear Medicine.
The results of the study demonstrate the impressive accuracy of the algorithm, which was able to accurately identify 92 percent of lesions. Furthermore, the algorithm's detection of changes in metabolic tumour volume was found to significantly correlate with disease-free and overall survival, consistent with the physician's evaluation.
In the interview, Dr Lohmann highlights the significance of the segmentation tool developed in the study, highlighting its potential to advance amino acid PET further and strengthen its clinical utility. He suggests that this tool could serve as a valuable resource, providing critical diagnostic information to benefit patients with brain tumours.
The full interview can be found here
Original publication: Automated Brain Tumor Detection and Segmentation for Treatment Response Assessment Using Amino Acid PET