Artificial Intelligence-Assisted Segmentation of Prostate Tumors and Neurovascular Bundles: Applications in Precision Surgery for Prostate Cancer.
Authors
Affiliations (6)
Affiliations (6)
- Department of Urology, Renmin Hospital of Wuhan University, 238 Jiefang Road, Wuchang District, Wuhan, 430060, Hubei, China.
- Institute of Urologic Disease, Renmin Hospital of Wuhan University, Wuhan, China.
- Department of Urology, Renmin Hospital of Wuhan University, 238 Jiefang Road, Wuchang District, Wuhan, 430060, Hubei, China. [email protected].
- Institute of Urologic Disease, Renmin Hospital of Wuhan University, Wuhan, China. [email protected].
- Department of Urology, Renmin Hospital of Wuhan University, 238 Jiefang Road, Wuchang District, Wuhan, 430060, Hubei, China. [email protected].
- Institute of Urologic Disease, Renmin Hospital of Wuhan University, Wuhan, China. [email protected].
Abstract
The aim of this study was to guide prostatectomy by employing artificial intelligence for the segmentation of tumor gross tumor volume (GTV) and neurovascular bundles (NVB). The preservation and dissection of NVB differ between intrafascial and extrafascial robot-assisted radical prostatectomy (RARP), impacting postoperative urinary control. We trained the nnU-Net v2 neural network using data from 220 patients in the PI-CAI cohort for the segmentation of prostate GTV and NVB in biparametric magnetic resonance imaging (bpMRI). The model was then validated in an external cohort of 209 patients from Renmin Hospital of Wuhan University (RHWU). Utilizing three-dimensional reconstruction and point cloud analysis, we explored the spatial distribution of GTV and NVB in relation to intrafascial and extrafascial approaches. We also prospectively included 40 patients undergoing intrafascial and extrafascial RARP, applying the aforementioned procedure to classify the surgical approach. Additionally, 3D printing was employed to guide surgery, and follow-ups on short- and long-term urinary function in patients were conducted. The nnU-Net v2 neural network demonstrated precise segmentation of GTV, NVB, and prostate, achieving Dice scores of 0.5573 ± 0.0428, 0.7679 ± 0.0178, and 0.7483 ± 0.0290, respectively. By establishing the distance from GTV to NVB, we successfully predicted the surgical approach. Urinary control analysis revealed that the extrafascial approach yielded better postoperative urinary function, facilitating more refined management of patients with prostate cancer and personalized medical care. Artificial intelligence technology can accurately identify GTV and NVB in preoperative bpMRI of patients with prostate cancer and guide the choice between intrafascial and extrafascial RARP. Patients undergoing intrafascial RARP with preserved NVB demonstrate improved postoperative urinary control.