Mp13-04 AI Model for Predicting Male Infertility Risk From Serum Hormone Levels Without Semen Analysis

You have accessJournal of UrologyInfertility: Epidemiology & Evaluation I (MP13)1 May 2024MP13-04 AI MODEL FOR PREDICTING MALE INFERTILITY RISK FROM SERUM HORMONE LEVELS WITHOUT SEMEN ANALYSIS Hideyuki Kobayashi, Masato Uetani, Shinjiro Takeuchi, Yozo Mitsui, Koichi Nakajima, and Koichi Nagao Hideyuki KobayashiHideyuki Kobayashi , Masato UetaniMasato Uetani , Shinjiro TakeuchiShinjiro Takeuchi , Yozo MitsuiYozo Mitsui , Koichi NakajimaKoichi Nakajima , and Koichi NagaoKoichi Nagao View All Author Informationhttps://doi.org/10.1097/01.JU.0001008832.14212.d6.04AboutPDF ToolsAdd to favoritesDownload CitationsTrack CitationsPermissionsReprints ShareFacebookLinked InTwitterEmail Abstract INTRODUCTION AND OBJECTIVE: Semen analysis is essential for the diagnosis of male infertility. However, it can have personal, social, and cultural stigma. Therefore, it is necessary to establish a system for ease of screening for male infertility risk and with this aim, we investigated a system using only serum hormone levels and AI predictive analysis. METHODS: We obtained data from 3,662 patients who visited our hospital between 2011 and 2020 and underwent endocrinological and semen analysis. Regarding male infertility, they were classified according to the results of semen analysis. Rates for NOA (non-obstructive azoospermia), OA (obstructive azoospermia), cryptozoospermia, oligozoospermia and/or asthenozoospermia, normal, and ejaculation disorder were 12.23 % (n=448), 5.73 % (n=210), 1.26 % (n=46), 44.21 % (n=1619), 36.40 % (n=1333), and 0.16 % (n=6), respectively. We extracted age, LH, FSH, PRL, testosterone, E2, and T/E2 from medical records. Total motility sperm count was calculated based on semen volume, sperm concentration, and sperm motility. According to the WHO 2021 guideline, a total motility sperm count of 9.408×106 (1.4 ml×16×106/ml×42%) was calculated as the lower limit of normal and applied to each patient to determine male infertility risk. AI predictive analysis models were created using Prediction One, a product of Sony Biz Networks, Inc and AutoML Tables, Google Cloud Platform. RESULTS: Characteristics were age, LH, FSH, testosterone, E2, T/E2, and PRL. Risk of male infertility (0: normal or 1: abnormal) was used as the objective variable for creating an AI prediction model with an AUC of 74.42% using Prediction One. For the AI prediction model using AutoML Tables, AUC ROC (receiver operating characteristic) and AUC PR (precision–recall) were 74.2% and 77.2%. In a ranking of feature importance from 1st to 3rd, FSH came a clear 1st. T/E2 and LH were ranked 2nd and 3rd, respectively, for both Prediction One and AutoML Tables. In addition, we used sperm analysis and serum hormone level data from 2021 and 2022 to verify the AI prediction model generated using Prediction One for male infertility risk. In 2021 and 2022, there were 24 and 28 cases of NOA. The predicted and actual results for NOA were 100% matched in these 2 years. CONCLUSIONS: We succeeded in making an AI model for evaluating male infertility using serum hormone levels without conventional sperm analysis. Although FSH was the feature with the most important contribution, adding T/E2 and LH further improved the accuracy of the AI prediction model. Source of Funding: a Grant-in-Aid for Scientific Research (C) from the Japan Society for the Promotion of Science (JSPS) (JSPS KAKENHI Grant Number JP22K09486) © 2024 by American Urological Association Education and Research, Inc.FiguresReferencesRelatedDetails Volume 211Issue 5SMay 2024Page: e211 Advertisement Copyright & Permissions© 2024 by American Urological Association Education and Research, Inc.Metrics Author Information Hideyuki Kobayashi More articles by this author Masato Uetani More articles by this author Shinjiro Takeuchi More articles by this author Yozo Mitsui More articles by this author Koichi Nakajima More articles by this author Koichi Nagao More articles by this author Expand All Advertisement PDF downloadLoading ...