Abstract:Objective:To determine the feasibility of using the U-Net deep learning network for automatic segmentation and localization of prostate cancer (PCa) on apparent diffusion coefficient (ADC) map.Methods:This study retrospectively collected clinical and imaging data from male patients who were scheduled for prostate biopsy due to clinically suspected PCa at Peking University First Hospital from June 1, 2017 to March 31, 2019. A total of 245 patients were included. All the patients were divided into PCa group ( n=184) and non-PCa group ( n=61) according to pathological results, and were randomly allocated into the dataset of the trainning (129 PCa, 19 non-PCa), validation (18 PCa, 3 non-PCa), and testing (37 PCa, 39 non-PCa). Ground truth of PCa lesion segmentation was obtained using manual annotation by one experienced radiologist on ADC map, referring to the pathological findings. After image preprocessing of automatic segmentation of the whole prostate gland, a U-Net was trained for the PCa lesion segmentation. With manual annotation as gold standard, dice similarity coefficient (DSC) was used to evaluate the segmentation effect. Diagnostic efficiency of PCa detection of the model was evaluated at lesion and patient level, and localization of the PCa lesions was also evaluated. Results:In the testing dataset, 51 PCa lesions were annotated and the largest dimension was (1.9±0.7) cm; 456 sectors were included based on sextant (122 PCa sectors and 334 non-PCa sectors). At lesion level, the mean DSC of the model was (70.2±21.7)% and the sensitivity of PCa lesion detection was 94.1% (48/51). At patient level, the sensitivity, specificity and accuracy of PCa detection were 97.3% (36/37), 66.7% (26/39), and 81.6% (62/76), respectively. The accuracy of automatic PCa localization based on sextant was 90.8% (414/456), the sensitivity was 95.1% (116/122), and the specificity was 89.2% (298/334).Conclusions:The U-Net deep learning model has high sensitivity for PCa detection on ADC map and high accuracy for PCa sextant localization. The prediction results can be automatically filled in the structured report, which further improves the clinical work efficiency.

Fully automatic segmentation on prostate MR images based on cascaded fully convolution network.

Prostate Ultrasound Image Segmentation Based on DSU-Net

Fully automated segmentation of prostate whole gland and transition zone in diffusion-weighted MRI using convolutional neural networks

Multi-scale discriminative network for prostate cancer lesion segmentation in multiparametric MR images

Automatic Segmentation of Prostate Magnetic Resonance Imaging Using Generative Adversarial Networks

3D Multi-Scale Discriminative Network with Multi-Directional Edge Loss for Prostate Zonal Segmentation in Bi-Parametric MR Images

A dual attention-guided 3D convolution network for automatic segmentation of prostate and tumor

Fully automated detection and localization of clinically significant prostate cancer on MR images using a cascaded convolutional neural network

A Cascaded Residual UNET for Fully Automated Segmentation of Prostate and Peripheral Zone in T2-weighted 3D Fast Spin Echo Images

A novel approach for automatic segmentation of prostate and its lesion regions on magnetic resonance imaging

Attention-guided multi-scale learning network for automatic prostate and tumor segmentation on MRI

Automatic Segmentation of Prostate Structures Using a Convolutional Neural Network from Multiparametric MRI

Automatic MRI Prostate Segmentation Using 3D Deeply Supervised FCN with Concatenated Atrous Convolution

A Novel Residual and Gated Network for Prostate Segmentation on MR Images.

Automatic Segmentation of Prostate MRI Based on 3D Pyramid Pooling Unet

A Fully Automatic Multi-Atlas Based Segmentation Method for Prostate MR Images

Automatic Segmentation of the Prostate on MR Images Based on Anatomy and Deep Learning.

CDA-Net: A Contrastive Deep Adversarial Model for Prostate Cancer Segmentation in MRI Images.

U-Net deep learning network for automatic segmentation and localization of prostate cancer on MRI apparent diffusion coefficient map

Cascade Dense-Unet for Prostate Segmentation in MR Images.