PSMA-Positive Prostatic Volume Prediction with Deep Learning Based on T2-Weighted MRI

Nobody wants to replace PSMA PET, but would it be feasible to estimate increased prostatic PSMA uptake based on T2 MRI alone? We know that high PSMA expression might correlate with structural characteristics such as histopathology growth patterns not recognized by the human eye on MRI images.

Differently, deep structural image analysis might be able to detect such differences and, eventually, predict if a lesion would be PSMA positive; indeed, we trained a neural network based on PSMA PET/MRI scans to predict increased prostatic PSMA uptake through the evaluation of axial T2-weighted sequence alone, reaching a dice similarity coefficient of 69.5 ± 15.6%. An algorithm with much more data, external validation, and of course, precision may improve (in a more than futurable vision) the assessment of doubtful/unclear prostatic MRI findings (i.e., PIRADS 3) determining if a patient requires further and more comprehensive examinations.

Written by: Riccardo Laudicella, MD, PhD, Nuclear Medicine Physician, Nuclear Medicine Unit, Department of Biomedical, Dental Sciences, and Morpho-Functional Imaging, Messina University, Italy

Read the Abstract

Login