A new CNN-based system for early diagnosis of prostate cancer

Abstract

We propose a convolutional neural network (CNN) based computer-aided diagnosis (CAD) system for early diagnosis of prostate cancer from diffusion-weighted magnetic resonance imaging (DWI). The proposed CNN-based CAD system begins by segmenting the prostate in a DWI dataset. Segmentation is achieved using our previously developed approach based on a geometric deformable model whose evolution is guided by first- and second-order appearance models. The spatial maps of apparent diffusion coefficients (ADCs) within the prostate, calculated for each 6-value, are used as image-based markers for the blood diffusion of the scanned prostate. For the purpose of classification/diagnosis, a three dimensional CNN has been trained to exact the most discriminatory features of these ADC maps for distinguishing malignant from benign prostate tumors. The proposed CNN-based CAD system is tested on DWI acquired from 23 patients using seven distinct 6-values. These experiments on in-vivo data confirm the high accuracy of the proposed CNN-based CAD system compared with our previously published results.