Machine learning in female urinary incontinence: A scoping review

Area of Science:

• Urology
• Medical Informatics
• Artificial Intelligence

Background:

• Female urinary incontinence (UI) is a prevalent condition impacting quality of life.
• Machine learning (ML) offers potential for improving diagnosis and management of UI.
• A comprehensive understanding of ML applications in female UI is needed.

Purpose of the Study:

• To conduct a scoping review of literature on ML applications in female UI over the past decade.
• To identify trends, methodologies, and outcomes of ML models used for female UI.

Main Methods:

• Systematic literature search across Medline, Google Scholar, PubMed, and Web of Science.
• Keywords included "Urinary incontinence" and "Machine learning" or "Predict" or "Prediction model".
• Data extraction focused on application area, ML type, input variables, and model validation.

Main Results:

• 23 studies met inclusion criteria from 798 identified papers.
• Logistic regression was the predominant ML method (91.3%).
• ML primarily predicted postpartum UI (39.1%) and de novo incontinence post-surgery (34.8%), with variable predictive performance (AUC 0.56-0.95).

Conclusions:

• ML models show potential for predicting and diagnosing female UI.
• Lack of standardization, transparency, and external validation limits current applicability and reproducibility.
• Future research should prioritize robust validation and transparent reporting of ML models for female UI.