Development of deep learning-based novel auto-segmentation for the prostatic urethra on planning CT images for prostate cancer radiotherapy

Abstract

Urinary toxicities are one of the serious complications of radiotherapy for prostate cancer, and dose-volume histogram of prostatic urethra has been associated with such toxicities in previous reports. Previous research has focused on estimating the prostatic urethra, which is difficult to delineate in CT images; however, these studies, which are limited in number, mainly focused on cases undergoing brachytherapy uses low-dose-rate sources and do not involve external beam radiation therapy (EBRT). In this study, we aimed to develop a deep learning-based method of determining the position of the prostatic urethra in patients eligible for EBRT. We used contour data from 430 patients with localized prostate cancer. In all cases, a urethral catheter was placed when planning CT to identify the prostatic urethra. We used 2D and 3D U-Net segmentation models. The input images included the bladder and prostate, while the output images focused on the prostatic urethra. The 2D model determined the prostate's position based on results from both coronal and sagittal directions. Evaluation metrics included the average distance between centerlines. The average centerline distances for the 2D and 3D models were 2.07 ± 0.87 mm and 2.05 ± 0.92 mm, respectively. Increasing the number of cases while maintaining equivalent accuracy as we did in this study suggests the potential for high generalization performance and the feasibility of using deep learning technology for estimating the position of the prostatic urethra.

Keywords: CT image; Deep learning technology; Prostate cancer; Prostatic urethra; Urinary toxicity.

Fig. 1

Workflow of our methods. We used 2D or 3D U-Net to estimate the prostatic urethra. Input data for deep learning model are the gray-scaled bladder and prostate. Two-dimensional data were created by projecting 3D data in the coronal and sagittal directions. Output from deep learning models could be irregularly shaped; therefore, the coordinates of the prostatic urethra were computed on each CT slice and a circle using the coordinates was created

Fig. 2

Box-and-whisker plots for the average centerline distance (CLD) and the percentage of segmented centerline lying within a radius (PWR). CLD is the average distance of the predicted urethra and ground truth obtained by the urinary catheter. It was calculated for the whole urethra and the top 1/3, middle 1/3, and bottom 1/3 of the prostate. The PWR is the percentage of sections where the difference is within a certain distance. We calculated the PWR for 3.5 mm and 5 mm

Fig. 3

Case of best prediction. Yellow lines show the ground truth, the blue line shows the prediction from the 2D model, and the red line shows the prediction from the 3D model. Various cases show the best prediction for the 2D and 3D models

Fig. 4

Case of worst prediction. Yellow lines show the ground truth, the blue line shows the prediction from the 2D model, and the red line shows the prediction from the 3D model. The same case shows the worst prediction for the 2D and 3D models. The CLD was 4.75 mm and 5.26 mm for the 2D and 3D models, respectively. Both these models could not predict the asymmetrical prostatic urethra and resulted in poor accuracy