One in four men will develop benign prostatic hyperplasia within their lifetime which often manifests as lower urinary tract symptoms (LUTS). International guidelines recommend uroflowmetry in the evaluation of male LUTS. However, traditional in-clinic uroflowmeters are costly, may malfunction, and provide only a one-off reading which may not represent the patient’s typical voiding behaviour, especially under the stress of a clinical setting.
SonoCurve is a novel machine-learning algorithm developed to provide uroflowmetry metrics and a flow curve from the sound of the urinary void. If deployed in a patient’s smart device, potential advantages include lower cost, serial at-home testing (in a more physiological environment) and remote assessment.
This pilot validation study aims to compare SonoCurve with conventional uroflowmetry in a symptomatic patient population.

After obtaining institutional approval and registration with the Clinical Trials Registry - India, we conducted a prospective, within person comparative study in male patients recruited from a specialist LUTS clinic (Mar/May 24). Men unable to void spontaneously, with catheters in situ, or with neurogenic bladders were excluded. 
Participants voided into a gravimetric uroflowmeter (Status Medical Equipments) from standing. To simulate a toilet bowl, 500ml of water was placed into the urine receptable at baseline. Uroflowmetry metrics (maximum flow rate, average flow rate, voided volume and voiding time) and raw flow data were recorded. Simultaneously, high-fidelity waveform audio recordings were made using a smartphone (Galaxy M14, Samsung) fixed 80cm above and 40cm behind the receptacle. Data from voids of <150mls or with significant artefact were excluded. 
Audio data were analysed using the SonoCurve algorithm to obtain sonouroflowmetry outputs. Paired outputs were compared using Lin’s concordance correlation coefficient (Python 3.12).

Flows from 42 men were included (mean age: 54.9 years, range 19-85). Moderate correlations were observed for maximum flow rate (0.69, 95% CI 0.49-0.83) and average flow rate (0.71, 95% CI 0.54-0.83). Strong correlations were found for voided volume (0.92, 95% CI 0.84-0.96) and voiding time (0.95, 95% CI 0.87-0.98).

In symptomatic patients in a real world clinical environment the SonoCurve algorithm was able to predict uroflowmetry results with accuracy. Reasons for lesser correlation in maximum and average flow rates may be due to the high level of background noise present in the sound recordings, artefacts from standard uroflowmetry, variation in room sizes or variation in phone placement during data capture.

This pilot study demonstrates that SonoCurve provides comparable results to conventional uroflowmetry in men with symptomatic LUTS in a clinical environment. Large-scale patient evaluation is planned.