Development of an Automatic Interpretation Algorithm for Uroflowmetry Results: Application of Artificial...

Area of Science:

• Urology
• Medical Informatics
• Artificial Intelligence

Background:

• Uroflowmetry (UFM) is a key diagnostic tool for lower urinary tract symptoms.
• Accurate interpretation of UFM results is crucial for timely diagnosis and treatment.
• Current interpretation relies on expert urologists, which can be time-consuming and subject to variability.

Purpose of the Study:

• To develop and validate an automated interpretation system for UFM results using machine learning (ML) and artificial intelligence (AI).
• To assess the performance of ML and deep learning (DL) algorithms in classifying UFM results as normal, borderline, or abnormal.

Main Methods:

• A dataset of 1,574 UFM results was prospectively collected and labeled by three urologists.
• Four ML and two DL algorithms were trained using UFM parameters and 2D curve images.
• Model performance was evaluated based on accuracy and area under the curve (AUC) for clinical discrimination.

Main Results:

• The automated system achieved high accuracy: up to 88.9% in males and 90.8% in females using six key parameters.
• Deep learning models, particularly convolutional neural networks, demonstrated superior performance in females (95.4% accuracy).
• The models showed excellent clinical discrimination, with AUC values up to 0.957 (males) and 0.974 (females).

Conclusions:

• An AI-powered algorithm for automatic UFM interpretation was successfully developed.
• The system effectively utilizes key UFM parameters and curve morphology for accurate classification.
• The developed algorithm demonstrates strong agreement with specialist urologist assessments, offering a promising tool for clinical practice.