Mp36-03 Decoding Benign Renal Cortical Lesions: A Machine Learning Approach From the Inmarc Registry

You have accessJournal of UrologyKidney Cancer: Epidemiology & Evaluation/Staging/Surveillance I (MP36)1 May 2024MP36-03 DECODING BENIGN RENAL CORTICAL LESIONS: A MACHINE LEARNING APPROACH FROM THE INMARC REGISTRY Cesare Saitta, Jonathan A. Afari, Mimi V. Nguyen, Kevin Hakimi, Dattatraya Patil, Hajime Tanaka, Julian Cortes, Margaret F. Meagher, Mirha Mahmood, Joshua Matian, Mariam Mansour, Dhruv Puri, Clara Cerrato, Kit L. Yuen, Masaki Kobayashi, Shohei Fukuda, Nicolò M. Buffi, Giovanni Lughezzani, Yasuhisa Fujii, Aditya Bagrodia, Viraj Master, and Ithaar H. Derweesh Cesare SaittaCesare Saitta , Jonathan A. AfariJonathan A. Afari , Mimi V. NguyenMimi V. Nguyen , Kevin HakimiKevin Hakimi , Dattatraya PatilDattatraya Patil , Hajime TanakaHajime Tanaka , Julian CortesJulian Cortes , Margaret F. MeagherMargaret F. Meagher , Mirha MahmoodMirha Mahmood , Joshua MatianJoshua Matian , Mariam MansourMariam Mansour , Dhruv PuriDhruv Puri , Clara CerratoClara Cerrato , Kit L. YuenKit L. Yuen , Masaki KobayashiMasaki Kobayashi , Shohei FukudaShohei Fukuda , Nicolò M. BuffiNicolò M. Buffi , Giovanni LughezzaniGiovanni Lughezzani , Yasuhisa FujiiYasuhisa Fujii , Aditya BagrodiaAditya Bagrodia , Viraj MasterViraj Master , and Ithaar H. DerweeshIthaar H. Derweesh View All Author Informationhttps://doi.org/10.1097/01.JU.0001008612.93052.9d.03AboutPDF ToolsAdd to favoritesDownload CitationsTrack CitationsPermissionsReprints ShareFacebookLinked InTwitterEmail Abstract INTRODUCTION AND OBJECTIVE: To create U.N.I.K. (Urologic Non-neoplastic Investigation of Kidneys), a machine learning (ML) model capable of predicting the probability of benign lesion (angiomyolipoma, oncocytoma and other benign neoplasia) at final histological report in patients suspected of renal cell carcinoma (RCC). METHODS: We queried the INMARC database for cT1/2 renal neoplasms who underwent surgery. Primary outcome was the development of a preoperative model capable of predicts benign neoplasia. Secondary outcome was to juxtapose ML performances to traditional logistic regression (LR) model. ML algorithm evaluated were, random forest, extreme gradient boosting (XGBoost), and light gradient boosting machine (LightGBM). Model's performances were evaluated with receiver operator characteristic curve (ROC) analysis using estimated area under the curve (AUC), accuracy, and F1 score. To evaluate differences between the ROC's curve the non-parametric DeLong's test was executed. RESULTS: Overall, 3,093 were analyzed [2,876 RCC (92.9%) vs. 217 benign histology (7.1%)]. LR revealed, female sex (OR 2.25, p<0.001), diabetes mellitus (OR 5.29, p<0.001), tumor size (OR 0.86, p<0.001), preoperative-CRP <1 mg/L (OR 1.57, p<0.001), preoperative calcium mg/dl (OR 1.87, p<0.001), Charlson Comorbidity Index (0.57, p<0.001), cystic lesion (OR 11.91, p<0.001), De Ritis ratio ≥0.9 (OR 1.50, p=0.024), preoperative proteinuria (OR 0.11, p=0.004) and Karnofsky Performance Status (OR 0.90, p<0.001) as independent predictors of benign cortical neoplasm (accuracy 0.92; F1 score 0.35; AUC 0.89). Every ML model demonstrated superior performances when compared to LR (Figure 1). Among ML model XGBoost demonstrated the best performances (Accuracy: 0.94; F1 score: 0.60; AUC: 0.94, Figure 1), and was used to develop U.N.I.K. CONCLUSIONS: Combining clinical features, serum biomarkers and radiographic findings, we have developed a point of care model (U.N.I.K.) capable of predicting, with significant accuracy, benign tumor in patients with cortical renal neoplasms being considered for surgical resection. U.N.I.K. may refine clinical decision making with respect to identification of patients with increased likelihood of benign tumor histology and reduce burden of surgical overtreatment. External validation is requisite. Download PPT Source of Funding: None © 2024 by American Urological Association Education and Research, Inc.FiguresReferencesRelatedDetails Volume 211Issue 5SMay 2024Page: e592 Advertisement Copyright & Permissions© 2024 by American Urological Association Education and Research, Inc.Metrics Author Information Cesare Saitta More articles by this author Jonathan A. Afari More articles by this author Mimi V. Nguyen More articles by this author Kevin Hakimi More articles by this author Dattatraya Patil More articles by this author Hajime Tanaka More articles by this author Julian Cortes More articles by this author Margaret F. Meagher More articles by this author Mirha Mahmood More articles by this author Joshua Matian More articles by this author Mariam Mansour More articles by this author Dhruv Puri More articles by this author Clara Cerrato More articles by this author Kit L. Yuen More articles by this author Masaki Kobayashi More articles by this author Shohei Fukuda More articles by this author Nicolò M. Buffi More articles by this author Giovanni Lughezzani More articles by this author Yasuhisa Fujii More articles by this author Aditya Bagrodia More articles by this author Viraj Master More articles by this author Ithaar H. Derweesh More articles by this author Expand All Advertisement PDF downloadLoading ...