Benign prostatic hyperplasia affects approximately a third of men aged >50 years, frequently causing lower urinary tract symptoms (LUTS). Uroflowmetry is a standard diagnostic tool for male LUTS, but conventional in-clinic uroflowmeters are expensive, prone to malfunction, and provide only one-off readings.
To overcome these limitations, we developed a novel machine-learning algorithm (SonoCurve) that provides uroflowmetry outputs and a flow curve from the sound of the urinary void. If deployed on the patient’s smart device, this approach would permit serial at-home testing. This has the potential to streamline patient pathways and reduce burden on healthcare systems.
This study aimed to compare the performance of SonoCurve against traditional uroflowmetry in healthy males.

Two healthy male volunteers (authors CK and WM, both aged 35), performed serial voids into a gravimetric uroflowmeter from a standing position (Jan/Mar 24). To simulate a toilet bowl, 500ml of water was placed into the urine receptacle before each void. Uroflowmetry outputs (max flow rate, average flow rate, voided volume and voiding time), along with raw flow data, were exported. Simultaneously, high-quality waveform audio file (WAV) recordings were captured using a smartphone (Galaxy S8+, Samsung) placed at a fixed distance 80cm above and 40cm behind the urine receptacle to simulate placement on the toilet cistern. Voids with significant sound artefact were excluded.
The SonoCurve algorithm was used to analyse audio files. Paired uroflowmetry and SonoCurve urinary flow metrics were compared using Lin’s concordance correlation coefficient. Analyses were conducted using Python 3.12. This non-invasive self-experimental study was conducted in accordance with all principles of the Declaration of Helsinki.

41 paired urinary flows were included (CK: 35, WM: 6). Moderate correlation was observed in maximum flow rate (0.85, 95% CI 0.75-0.9). Strong correlation was observed in average flow rate (0.92, 95% CI 0.86-0.95), voided volume (0.91, 95% CI 0.85-0.95) and voiding time (1, 95% CI 0.99-1).

In healthy male volunteers under ideal conditions the SonoCurve algorithm is able to accurately predict uroflowmetry parameters. Variations between predicted and actual results are unlikely to be clinically significant in a symptomatic patient population due to the repeatability of the test.

Our findings demonstrate the feasibility of SonoCurve in predicting uroflowmetry metrics from smartphone audio recordings of the urinary void in healthy male volunteers. Patient evaluation has been initiated.