Key points are not available for this paper at this time.
You have accessJournal of UrologyKidney Cancer: Epidemiology & Evaluation/Staging/Surveillance I (MP36)1 May 2024MP36-18 ARTIFICIAL INTELLIGENCE BASED PERSONALIZED ONCOLOGICAL OUTCOME PREDICTION MODEL FOR UPPER URINARY TRACT UROTHELIAL CARCINOMA AFTER RADICAL NEPHROURETERECTOMY: DEVELOPMENT AND MULTI-CENTER VALIDATION Homin Kang, Jaeyoung Cho, Bumjin Lim, Dalsan You, Cheryn Song, In Gab Jeong, Bumsik Hong, Jun Hyuk Hong, Hanjong Ahn, Hwanik Kim, and Ja Hyeon Ku Homin KangHomin Kang , Jaeyoung ChoJaeyoung Cho , Bumjin LimBumjin Lim , Dalsan YouDalsan You , Cheryn SongCheryn Song , In Gab JeongIn Gab Jeong , Bumsik HongBumsik Hong , Jun Hyuk HongJun Hyuk Hong , Hanjong AhnHanjong Ahn , Hwanik KimHwanik Kim , and Ja Hyeon KuJa Hyeon Ku View All Author Informationhttps://doi.org/10.1097/01.JU.0001008612.93052.9d.18AboutPDF ToolsAdd to favoritesDownload CitationsTrack CitationsPermissionsReprints ShareFacebookLinked InTwitterEmail Abstract INTRODUCTION AND OBJECTIVE: This study aimed to develop and validate an artificial intelligence (AI)-based personalized oncological outcome prediction model for upper-urinary tract urothelial carcinoma (UTUC) patients undergoing radical nephroureterectomy (RNU). METHODS: Data from UTUC patients who underwent RNU between 2010 to 2020 across three hospitals were retrospectively analyzed. The model was developed using data from one tertiary center and externally validated with data from two other tertiary referral hospitals. Primary outcomes were progression-free survival (PFS) and overall survival (OS). The AI model utilized XGBoost for risk score estimation. These scores were subsequently fed into a bootstrapped Weibull Accelerated Failure Time (AFT) model for survival curve calibration. Hyperparameter tuning was executed using the Optuna method. Model efficacy was gauged using the Concordance index (c-index), and a dockerized version of the model was deployed on Google Cloud Run. RESULTS: Of the 678 patients considered, 628 qualified for model training and internal validation, with a 9:1 division for training and validation sets. External validation involved data from 108 and 256 patients from two distinct hospitals. The median age was 70.0 years (IQR: 62.0-76.0) with 28.8% (181/628) being female. From the training data, 37.4% (235/628) had a pathologic T3 stage above, 7.4% (47/628) had node-positive disease, and distant metastasis was evident in 25.8% (162/628). Statistically significant parameters across the three hospitals included pre-operative GFR (p<0.001), hydronephrosis presence(p=0.013), pathologic T (p<0.001) and N stages(p<0.001), concurrent CIS(p<0.001), disease progression(p<0.001), and survival rate(p<0.001). The PFS model's c-index on training and internal validation were 0.889 and 0.789, respectively. External validations yielded c-index values of 0.725 and 0.724. For the OS model, training and internal validation c-indexes were 0.853 and 0.809, with external validation scores at 0.757 and 0.745. CONCLUSIONS: The AI-based model effectively predicts PFS and OS outcomes for UTUC patients post-RNU, showcasing robust performance across multicentre settings. Access to the demo version of developed model is available at https://utuc-surv-b2s45pnlyq-an.a.run.app/. Source of Funding: None © 2024 by American Urological Association Education and Research, Inc.FiguresReferencesRelatedDetails Volume 211Issue 5SMay 2024Page: e599 Advertisement Copyright & Permissions© 2024 by American Urological Association Education and Research, Inc.Metrics Author Information Homin Kang More articles by this author Jaeyoung Cho More articles by this author Bumjin Lim More articles by this author Dalsan You More articles by this author Cheryn Song More articles by this author In Gab Jeong More articles by this author Bumsik Hong More articles by this author Jun Hyuk Hong More articles by this author Hanjong Ahn More articles by this author Hwanik Kim More articles by this author Ja Hyeon Ku More articles by this author Expand All Advertisement PDF downloadLoading ...
Kang et al. (Mon,) studied this question.