AI-assisted cryo-dose estimation and tissue injury modeling for assessing side effects in MRI-guided prostate cryoablation: a retrospective study

Phys Med Biol. 2026 Mar 3;71(5):10.1088/1361-6560/ae4847. doi: 10.1088/1361-6560/ae4847.

Abstract

Objective.Focal therapy offers a middle ground between active surveillance and whole-gland treatment for patients with low- and intermediate-risk prostate cancer, aiming to eliminate tumor tissue while preserving surrounding structures. Among available focal therapy approaches, magnetic resonance imaging (MRI)-guided focal cryoablation offers some technical advantages, including real-time monitoring of iceball formation and intraoperative images with excellent soft-tissue contrast. Despite its minimally invasive nature, prior studies have reported postoperative urinary and sexual dysfunction, raising concerns about unintended injury to critical anatomical structures such as the external urethral sphincter (EUS) and neurovascular bundles (NVBs). However, the direct relationship between anatomical injury and postoperative symptoms remains unclear. We hypothesize that damage to the EUS or NVBs during MRI-guided focal cryoablation increases the risk of postoperative urinary or erectile dysfunction.Approach.We analyzed 402 multi-slice MRI image sets from 42 procedures. We developed an AI-assisted tool for automatic segmentation of anatomical structures and iceballs. Quantitative metrics were proposed to estimate tissue damage based on spatial overlap, exposure time, and temperature dynamics. These metrics were then statistically correlated with postoperative urinary and erectile outcomes.Main results.Significant associations were observed between injury metrics and postoperative dysfunctions, with EUS involvement predictive of urinary dysfunction in both time-based(p<0.05)and temperature-proxy models(p<0.05), and NVB involvement associated with erectile dysfunction in the time-based model(p<0.05).Significance.Our findings demonstrate that EUS and NVB injuries are quantifiable predictors of postoperative morbidity with medium-to-large effect sizes. The developed AI-assisted framework supports objective evaluation of cryo-dose and establishes a preliminary basis for a quantitative injury metric designed to minimize side effects and enhance the safety of focal cryoablation.

Keywords: MRI; artificial intelligence; cryoablation; focal therapy; prostate cancer.

MeSH terms

  • Cryosurgery* / adverse effects
  • Cryosurgery* / methods
  • Humans
  • Magnetic Resonance Imaging*
  • Male
  • Prostate* / diagnostic imaging
  • Prostate* / surgery
  • Prostatic Neoplasms* / diagnostic imaging
  • Prostatic Neoplasms* / surgery
  • Radiation Dosage*
  • Retrospective Studies
  • Surgery, Computer-Assisted* / adverse effects
  • Surgery, Computer-Assisted* / methods