Meeting Banner
Abstract #1047

Support Vector Neural Networks versus Logistic Regression MR based diagnostic model for classification of transition zone prostate cancer

Nikolaos Dikaios 1,2 , Jokha Alkalbani 2 , Alex Kirkham 3 , Clare Allen 3 , Hashim Ahmed 4 , Mark Emberton 4 , Alex Freeman 5 , Steve Halligan 2 , Stuart Taylor 2 , David Atkinson 2 , and Shonit Punwani 2

1 Medical Physics, UCL, London, Greater London, United Kingdom, 2 Centre of Medical Imaging, UCL, Greater London, United Kingdom, 3 Radiology, UCL, Greater London, United Kingdom, 4 Urology, UCL, Greater London, United Kingdom, 5 Histopathology, UCL, Greater London, United Kingdom

Multi-parametric MRI (mp-MRI) facilitates identification of transition zone cancers, yet its overall diagnostic accuracy is likely lower in this part of the prostate compared with the peripheral zone. Benign hyperplastic nodules within the transition zone likely make the localisation of cancer difficult. Logistic regression (LR) models1 for classifying transition zone (TZ) prostate cancer (PCa) on mp-MRI were previously derived and validated. Here we explore whether the application of support vector machine (SVM) neural network (SVNN) algorithms can improve classification accuracy. The proposed SVNN algorithm is trained on 70 patients and temporally validated on a second independent cohort of 85 patients.

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

Join Here