Clinically Significant Prostate Cancer Classification Using Anatomical T2W MRI Sequences and Deep Learning

The precise identification and categorization of clinically significant prostate cancer (csPCa) is crucial for making effective treatment decisions. The utilization of various imaging sequences in biparametric MRI (bpMRI) enhances diagnostic accuracy. The present paper aims to examine the capacity of deep learning frameworks, such as DenseNet-201, ResNet-50, and VGG-19, to differentiate between csPCa and non-csPCa by utilizing various imaging sequences of biparametric MRI. Through this study, we aim to assess the effectiveness of deep learning frameworks with respect to diffusion-weighted imaging (DWI), apparent diffusion coefficient (ADC) mapping, and axial T2-weighted imaging (T2W). The results of the study show that when it comes to csPCa classification accuracy and AUC, the axial T2W is the most effective among the biparametric MRI sequences analyzed. In the classification of csPCa in bpMRI, these findings have underscored the critical nature of axial T2W imaging.