You have accessJournal of UrologyUrodynamics/Lower Urinary Tract Dysfunction/Female Pelvic Medicine: Non-neurogenic Voiding Dysfunction I (PD21)1 Apr 2020PD21-12 APPLICATION OF MACHINE LEARNING ALGORITHMS TO CLASSIFY STORAGE LOWER URINARY TRACT SYMPTOMS Kai Dallas*, Ashley Caron, Jennifer Anger, Karyn Eilber, and A. Lenore Ackerman Kai Dallas*Kai Dallas* More articles by this author , Ashley CaronAshley Caron More articles by this author , Jennifer AngerJennifer Anger More articles by this author , Karyn EilberKaryn Eilber More articles by this author , and A. Lenore AckermanA. Lenore Ackerman More articles by this author View All Author Informationhttps://doi.org/10.1097/JU.0000000000000871.012AboutPDF ToolsAdd to favoritesDownload CitationsTrack CitationsPermissionsReprints ShareFacebookLinked InTwitterEmail Abstract INTRODUCTION AND OBJECTIVE: Lower urinary tract symptoms (LUTS) are a set of symptoms that encompass problems with the holding (storage) and emptying (voiding) of urine. Storage LUTS (urinary urgency, frequency, nocturia, painful urination, and bladder discomfort) are categorized into several conditions with symptomatic overlap (e.g. interstitial cystitis/painful bladder syndrome (IC/PBS) and overactive bladder (OAB)). As no objective diagnostic criteria exists for these conditions, we applied machine learning algorithms to generate novel diagnostic groupings. METHODS: 514 patients referred to a urogynecology clinic at a tertiary referral center completed the Interstitial Cystitis Symptom and Problem Indices (ICSI/ICPI), Overactive Bladder Questionnaire (OABq), Genitourinary Pain Index (GUPI), and Pelvic Floor Disability Index (PFDI-20). The most common complaints at presentation were bladder or pelvic pain, urinary frequency, incontinence, and hematuria. Machine learning unsupervised clustering (k-means) was used to define patients groups based on similar patient phenotype (patient demographics and questionnaire data, see Figure 1). All analysis was performed in R version 3.6.1. RESULTS: The algorithm identified 5 patient clusters: 142 controls (asymptomatic scores on all indices), 85 patients with a broad range of high responses (Bladder Pain Syndrome-BPS), 55 patients with urgency and urge incontinence (UUI), 111 with vaginal/pelvic pain unrelated to voiding (Non-Urologic Urogenital Pain-NUPP) and 121 with pelvic floor dysfunction (PFD) (Figure 1). The specific groupings can be compared in depth with our interactive application: https://drlackerman-lab.shinyapps.io/luts-kmeans/ CONCLUSIONS: Machine learning identifies phenotypic-specific patient cluster based on validated questionnaires, refining our classic diagnostic scheme of LUTS to more specifically define OAB, IC/PBS, and asymptomatic patients and distinguish these from new groups of pelvic floor myofascial-derived urgency/frequency and non-urologic pelvic pain. This is a promising novel approach to categorizing these patients with LUTS with likely prognostic implications, which we will explore in prospective studies to follow. Source of Funding: n/a © 2020 by American Urological Association Education and Research, Inc.FiguresReferencesRelatedDetails Volume 203Issue Supplement 4April 2020Page: e458-e458 Advertisement Copyright & Permissions© 2020 by American Urological Association Education and Research, Inc.MetricsAuthor Information Kai Dallas* More articles by this author Ashley Caron More articles by this author Jennifer Anger More articles by this author Karyn Eilber More articles by this author A. Lenore Ackerman More articles by this author Expand All Advertisement PDF downloadLoading ...