Quantification of Tumor Variability in Mice and Patients using PET-MRI Machine-Learning Models – Nature Biomedical Engineering

Researchers have found a way to better understand the heterogeneity of tumours in mice and patients through machine-learning models trained on PET-MRI data. In oncology, intratumoural heterogeneity, or the differences within a tumour, can impact the effectiveness of treatments. By using dynamic positron emission tomography (PET) and multiparametric magnetic resonance imaging (MRI), researchers were able to train classifiers to identify changes in tumour subtypes caused by targeted therapies in mice. They were also able to apply these classifiers to retrospective PET-MRI data from patients with liver metastases from colorectal cancer, which agreed with tumour histology. The ability to characterize intratumoural heterogeneity through machine-learning and multimodal imaging may lead to better precision oncology treatments.