Abstract #2570

mpMRI-based Machine-Learning Classifier Comparison for Gleason 4 Pattern Detection in Transition Zone and Peripheral Zone Prostate Lesions

Michela Antonelli^¹, Edward W Johnston^², Sebastien Ourselin*^1,3, and Shonit Punwani*²

¹Translational Imaging Group, CMIC, University College London, London, United Kingdom, ²Centre for Medical Imaging, University College London, London, United Kingdom, ³Dementia Research Centre, Department of Neurodegenerative Disease, UCL Institute of Neurology, London, United Kingdom

Multi-parametric MRI (mpMRI) can be used to non-invasively predict the presence of a Gleason 4 pattern in transition zone (TZ) and peripheral zone (PZ) prostate cancers. Here the performance of five machine-learning classifiers, which use mpMRI and clinical features, were compared. Analysis included a five-fold cross validation and a temporally separated validation to prove the generalisability of the classifiers. The results showed that PZ models can predict the presence of a Gleason 4 pattern better than TZ models. The statistically better PZ classifier is a linear regression model while for TZ the best classifier is Naïve Bayes model.

This abstract and the presentation materials are available to members only; a login is required.

Join Here