JoVE - Journal of Visualized Experiments
JoVE Visualize
Contact Us

Predicting Biochemical Recurrence of Prostate Cancer Post-Prostatectomy Using Artificial Intelligence: A Systematic Review

  • 0E.J. Whitten Prostate Cancer Research Centre, Epworth Healthcare, Melbourne 3002, Australia.
Cancers +

|

November 9, 2024

|

November 9, 2024

View abstract on PubMed

Summary

This summary is machine-generated.

Artificial intelligence (AI) shows promise in predicting biochemical recurrence (BCR) after radical prostatectomy (RP). AI models using radiological data achieved higher accuracy than those using pathological or clinicopathological data alone.

Area Of Science

  • Urology
  • Oncology
  • Medical Imaging
  • Artificial Intelligence

Background

  • Biochemical recurrence (BCR) after radical prostatectomy (RP) is a critical indicator of disease progression and mortality in prostate cancer (PCa).
  • Accurate prediction of BCR is essential for guiding post-operative management and improving patient outcomes.
  • Existing prediction models have limitations, necessitating exploration of advanced techniques like artificial intelligence (AI).

Purpose Of The Study

  • To systematically review and evaluate the accuracy of artificial intelligence (AI) algorithms in predicting biochemical recurrence (BCR) in prostate cancer (PCa) patients following radical prostatectomy (RP).

Main Methods

  • A comprehensive literature search was conducted across major scientific databases (Medline, Embase, Web of Science, IEEE Xplore) following PRISMA guidelines.
  • Included studies utilized AI for BCR prediction post-RP, excluding those involving radiotherapy or salvage RP.
  • Data from 24 studies involving 27,216 patients were analyzed, with 7,267 developing BCR.

Main Results

  • AI algorithms incorporating radiological parameters demonstrated superior predictive accuracy (median AUROC of 0.90) compared to those based solely on pathological (0.74) or clinicopathological (0.81) variables.
  • Seven studies reported AI models outperforming or matching traditional BCR prediction methods.
  • The overall risk of bias was unclear in three studies due to ambiguous criteria and missing follow-up data.

Conclusions

  • AI holds significant potential for predicting BCR post-RP, especially when leveraging radiological data.
  • Variability in AI performance and study methodologies underscores the need for standardized, prospective research with external validation before widespread clinical adoption.
  • Further research is required to optimize AI models and ensure their reliability in clinical practice.

Keywords: