Machine Learning Models for the Noninvasive Diagnosis of Bladder Outlet Obstruction and Detrusor Underactivity in Men With Lower Urinary Tract Symptoms - Study overview

Aug 6, 2026

This study develops and evaluates machine learning models, including CatBoost and extreme gradient boosting (XGBoost), to differentiate bladder outlet obstruction (BOO) from detrusor underactivity (DUA) in men with lower urinary tract symptoms (LUTS). The models use patient-reported outcomes, uroflowmetry measurements, and ultrasound-derived features collected from men aged 40 and older. Results highlight CatBoost’s higher sensitivity for screening and XGBoost’s higher specificity and overall diagnostic performance (AUROC 0.826 for BOO and 0.819 for DUA).

LanguageEnglish

CategoryHealthcare

FormatPDF

Pages8

Reading Time20 min

AccessFree to download

Purpose

Methods

Results

Conclusions

Keywords

Introduction

Original Article Int Neurourol J 2024;28(Suppl 2):S74-81 [https://doi.org/10.5213/inj.2448360.180](https://doi.org/10.5213/inj.2448360.180)[ ](https://doi.org/10.5213/inj.2448360.180)pISSN 2093-4777 · eISSN 2093-6931 Machine Learning Models for the Noninvasive Diagnosis of Bladder Outlet Obstruction and Detrusor Underactivity in Men With Lower Urinary Tract Symptoms Hyungkyung Shin1,*, Kwang Jin Ko2,*, Wei-Jin Park1, Deok Hyun Han2, Ikjun Yeom1,3, Kyu-Sung Lee2 1Acryl Advanced AI Research Center, Acryl Inc., Seoul, Korea 2Department of Urology, Samsung Medical Center, Sungkyunkwan University School of Medicine, Seoul, Korea 3College of Computing and Informatics, Sungkyunkwan University, Suwon, Korea Purpose: This study aimed to develop and evaluate machine learning models, specifically CatBoost and extreme gradient boosting (XGBoost), for diagnosing lower urinary tract symptoms (LUTS) in male patients. The objective is to differentiate between bladder outlet obstruction (BOO) and detrusor underactivity (DUA) using a comprehensive dataset that includes patient-reported outcomes, uroflowmetry measurements, and ultrasound-derived features. Methods: The dataset used in this study was collected from male patients aged 40 and older who presented with LUTS and sought treatment at the urology department of Samsung Medical Center. We developed and trained CatBoost and XGBoost models using this dataset. These models incorporated features like prostate size, voiding parameters, and responses from questionnaires. Their performance was assessed using standard metrics such as accuracy, precision, recall, F1-score, and area under the receiver operating characteristic curve (AUROC). Results: The results indicated that the CatBoost models displayed greater sensitivity, rendering them effective for initial screenings by accurately identifying true positive cases. Conversely, the XGBoost models showed higher specificity and precision, making them more suitable for confirming diagnoses and reducing false positives. In terms of overall performance for both BOO and DUA, XGBoost surpassed CatBoost, achieving an AUROC of 0.826 and 0.819, respectively. Conclusions: Integrating these machine learning models into the diagnostic workflow for LUTS can significantly enhance clinical decision-making by offering noninvasive, cost-effective, and patient-friendly diagnostic alternatives. The combined application of CatBoost and XGBoost models has the potential to improve diagnostic accuracy and provide customized treatment plans for patients, ultimately leading to better clinical outcomes. Keywords: Artificial intelligence; Bladder outlet obstruction; Diagnosis; Lower urinary tract symptoms; Urinary bladder, Underactive • Grant/Fund Support: This work was supported by National IT Industry Promotion Agency (NIPA) grant funded by the Korean government (MSIT) (No.H0401-24-1001, Development of AI Precision Medical Solution (Doctor Answer 2.0)). • Research Ethics: The data for this study were collected from Samsung Medical Center, following approval from the Institutional Review Board (IRB) under File No. SMC 2021-08-116. • Conflict of Interest: No potential conflict of interest relevant to this article was reported. Corresponding author: Kyu-Sung Lee [https://orcid.org/0000-0003-0891-2488](https://orcid.org/0000-0003-0891-2488)[ ](https://orcid.org/0000-0003-0891-2488)Department of Urology, Samsung Medical Center, Sungkyunkwan University School of Medicine, 81 Irwon-ro, Gangnam-gu, Seoul 06351, Korea [Email: ksleedr@skku.edu](Email: ksleedr@skku.edu) Co-corresponding author: Ikjun Yeom [https://orcid.org/0000-0001-7883-2905](https://orcid.org/0000-0001-7883-2905)[ ](https://orcid.org/0000-0001-7883-2905)College of Computing and Informatics, Sungkyunkwan University, 2066 Seoburo, Jangan-gu, Suwon 16419, Korea Email: [ikjun@skku.edu](ikjun@skku.edu) *Hyungkyung Shin and Kwang Jin Ko contributed equally to this study as co-first au

FAQs

What was the main purpose of this study?

To develop and evaluate machine learning models (CatBoost and XGBoost) for diagnosing lower urinary tract symptoms by distinguishing BOO from DUA using noninvasive data.

How were CatBoost and XGBoost models built and assessed?

The models were trained on a dataset from men aged 40 and older, using patient-reported outcomes, uroflowmetry, and ultrasound-derived features. Performance was measured with metrics such as accuracy, precision, recall, F1-score, and AUROC.

How did CatBoost and XGBoost differ in diagnostic behavior?

CatBoost showed greater sensitivity, supporting initial screening by identifying true positives. XGBoost showed higher specificity and precision, making it more suitable for confirming diagnoses and reducing false positives, with overall AUROC values reported for both BOO and DUA.

Aug 6, 2026

Upload by Emma Wilson