We proposed a patch-based convolutional neural network (CNN) model to distinguish prostate cancer using multi-parametric magnetic resonance images (mp-MRI). Our CNN model was trained in 182 patients including 193 cancerous (CA) vs. 259 normal (NC) regions, and tested independently in 21 patients including 21 CA vs 31 NC regions. The model produced an area under the receiver operating characteristic curve of 0.869, sensitivity of 90.5% and specificity of 67.7% for the differentiation of CA from normal regions, showing its potential for the diagnosis of prostate cancer in clinical application.
This abstract and the presentation materials are available to members only; a login is required.