Using machine learning to construct the diagnosis model of female bladder outlet obstruction based on urodynamic...

Area of Science:

• Urology
• Medical Informatics
• Biomedical Engineering

Background:

• Bladder outlet obstruction (BOO) is a common condition affecting women.
• Accurate diagnosis is crucial for effective treatment, especially in women with preserved detrusor contractility.
• Current diagnostic methods may require further refinement for specific patient populations.

Purpose of the Study:

• To develop and validate a machine learning-based diagnostic model for female bladder outlet obstruction (BOO).
• To utilize urodynamic study (UDS) data, focusing on patients with adequate detrusor contraction.
• To enhance the accuracy and reliability of BOO diagnosis in this specific demographic.

Main Methods:

• Retrospective analysis of urodynamic study (UDS) data from 134 female patients.
• Calculation of eleven urinary flow indicators during the voiding phase.
• Development of eight back propagation neural network models using varying combinations of UDS indicators.
• Evaluation of model stability and performance using five-fold cross-validation and a test dataset.

Main Results:

• The optimal model, utilizing 9 UDS indicators, achieved an Area Under the Curve (AUC) of 0.949 ± 0.060.
• The diagnostic model demonstrated high accuracy (94.4%), sensitivity (100%), and specificity (89.3%) in the testing phase.
• The study identified 9 significant UDS indicators crucial for diagnosing female BOO.

Conclusions:

• A machine learning model effectively diagnoses female bladder outlet obstruction (BOO) using specific urodynamic study (UDS) indicators.
• The developed model exhibits excellent classification accuracy and stability.
• This approach provides a promising tool for the intelligent diagnosis of BOO in women with preserved detrusor function.