Elterman Dean, Bhojani Naeem, Lee Kiwook, Jung Jiyoung, Doo Karen, Gressler Laura E, Chughtai Bilal
Division of Urology, Centre Hospitalier de l'Université de Montréal, Université de Montréal, Montreal, QC, Canada.
Division of Urology, University Health Network, University of Toronto, Toronto, ON, Canada.
Can Urol Assoc J. 2025 Jan;19(1):E50-E54. doi: 10.5489/cuaj.8870.
We sought to assess the performance of the proudP artificial intelligence (AI) algorithm, integrated into a mobile application, in estimating uroflow curves and parameters using recorded urination sounds.
A direct comparison was made between the peak flow rate (Qmax), voided volume (VV), and uroflow curves predicted by the proudP algorithm and those obtained through established validation methods. A hardware uroflow simulator replicated uroflow profiles by precisely controlling water flow rates and extracting corresponding sound data. Ten uroflow profiles, representing typical patterns observed in male subjects, were selected. Simulation experiments with proudP were conducted using a standard toilet setup. The uroflow simulator was calibrated to reproduce uroflow profiles, and validation was performed against a Flowmaster uroflowmetry device. Statistical analysis included descriptive summaries, Bland-Altman analysis, and concordance correlation coefficient (CCC) analysis.
The proudP accurately captured various uroflow patterns generated by the simulator, with low standard deviations in Qmax predictions and biases near zero. The standard deviations of voided volume were slightly larger, primarily due to uroflow patterns with extended voiding times. The study validated the accuracy of proudP against in-office uroflowmetry, demonstrating robustness across different smartphone models.
ProudP proved to be as accurate as in-office uroflowmetry in estimating uroflow rate across various patterns. Its convenience in home monitoring offers patients a means to observe their urination patterns accurately, while enabling healthcare professionals to gain detailed insights remotely. ProudP emerges as an essential solution for clinical practice and urological research.
我们旨在评估集成到移动应用程序中的proudP人工智能（AI）算法，利用记录的排尿声音来估计尿流曲线和参数的性能。
对proudP算法预测的最大尿流率（Qmax）、排尿量（VV）和尿流曲线与通过既定验证方法获得的结果进行直接比较。一种硬件尿流模拟器通过精确控制水流速率并提取相应的声音数据来复制尿流曲线。选择了代表男性受试者中观察到的典型模式的十条尿流曲线。使用标准马桶设置对proudP进行模拟实验。对尿流模拟器进行校准以再现尿流曲线，并与Flowmaster尿流计设备进行验证。统计分析包括描述性总结、Bland-Altman分析和一致性相关系数（CCC）分析。
proudP准确捕捉了模拟器生成的各种尿流模式，Qmax预测的标准差较低，偏差接近零。排尿量的标准差略大，主要是由于排尿时间延长的尿流模式。该研究验证了proudP相对于门诊尿流计的准确性，表明在不同智能手机型号上具有稳健性。
在估计各种模式下的尿流率方面，proudP被证明与门诊尿流计一样准确。它在家庭监测中的便利性为患者提供了一种准确观察其排尿模式的方法，同时使医疗保健专业人员能够远程获得详细的见解。proudP成为临床实践和泌尿外科研究的重要解决方案。
