Classification of Bladder Emptying Patterns by LSTM Neural Network Trained Using Acoustic Signatures

Area of Science:

• Biomedical Engineering
• Urology
• Artificial Intelligence

Background:

• Non-invasive uroflowmetry is crucial for diagnosing lower urinary tract symptoms (LUTS).
• Current methods face temporospatial constraints.
• A smart system utilizing acoustic signatures of uroflow is proposed for unconstrained measurement.

Purpose of the Study:

• To develop and validate a deep-learning algorithm for measuring urine flowrate using acoustic signatures.
• To establish a smart, non-invasive uroflowmetry system for remote patient monitoring.

Main Methods:

• Collected uroflowmetry data from 27 patients with LUTS.
• Utilized a long short-term memory (LSTM) deep-learning algorithm.
• Analyzed uroflow sounds based on loudness and roughness parameters.

Main Results:

• The acoustic signature analysis showed signal patterns similar to clinical urological measurements.
• Consistent flowrate values were obtained, validating the model.
• Predicted flowrate accurately reflected actual physical characteristics.

Conclusions:

• Acoustic uroflowmetry using deep learning is a viable, non-invasive diagnostic tool.
• This technology can lead to smart devices for everyday use and remote monitoring.
• Objective, continuous measurement of bladder function can enable early diagnosis of urinary system diseases.