Mp36-14 How to Perform the Optimal Renal Biopsy: A Machine Learning-Based Indication

You have accessJournal of UrologyKidney Cancer: Epidemiology & Evaluation/Staging/Surveillance I (MP36)1 May 2024MP36-14 HOW TO PERFORM THE OPTIMAL RENAL BIOPSY: A MACHINE LEARNING-BASED INDICATION Federico Belladelli, Chiara Re, Francesco Cei, Giacomo Musso, Giuseppe Rosiello, Daniele Cignoli, Daniela Canibus, Francesco Fiorio, Roberto Bertini, Andrea Salonia, Francesco De Cobelli, Giorgio Brembilla, Antonio Esposito, Anna Palmisano, Ciro Piccolo, Marco Gambirasio, Roberta Lucianò, Nazario Tenace, Alberto Briganti, Francesco Montorsi, Alessandro Larcher, and Umberto Capitanio Federico BelladelliFederico Belladelli , Chiara ReChiara Re , Francesco CeiFrancesco Cei , Giacomo MussoGiacomo Musso , Giuseppe RosielloGiuseppe Rosiello , Daniele CignoliDaniele Cignoli , Daniela CanibusDaniela Canibus , Francesco FiorioFrancesco Fiorio , Roberto BertiniRoberto Bertini , Andrea SaloniaAndrea Salonia , Francesco De CobelliFrancesco De Cobelli , Giorgio BrembillaGiorgio Brembilla , Antonio EspositoAntonio Esposito , Anna PalmisanoAnna Palmisano , Ciro PiccoloCiro Piccolo , Marco GambirasioMarco Gambirasio , Roberta LucianòRoberta Lucianò , Nazario TenaceNazario Tenace , Alberto BrigantiAlberto Briganti , Francesco MontorsiFrancesco Montorsi , Alessandro LarcherAlessandro Larcher , and Umberto CapitanioUmberto Capitanio View All Author Informationhttps://doi.org/10.1097/01.JU.0001008612.93052.9d.14AboutPDF ToolsAdd to favoritesDownload CitationsTrack CitationsPermissionsReprints ShareFacebookLinked InTwitterEmail Abstract INTRODUCTION AND OBJECTIVE: Renal biopsies (RB) may not yield pathologic diagnosis and even in a larger proportion it is not possible to diagnose tumor grading. The aim of the study was to define by machine learning algorithms an optimal RB strategy. METHODS: Within a prospectively maintained database, patients with indeterminate renal masses who underwent RB at a single tertiary center were identified. We recorded and analyzed the guidance method (US/CT), number of biopsy cores (NoC), and total tissue length (LoC) to evaluate their influence on diagnostic outcomes, including histological characterization and grading. The Boruta algorithm was used to select relevant variables. Subsequently, the K-Nearest Neighbors (KNN), a non-parametric supervised machine learning model, predicted the probability of obtaining a pathological diagnosis including tumor grading based on the chosen variables. Decision boundaries from the KNN model guided the creation of diagnostic recommendations, which were tested using multivariable logistic regression. RESULTS: Overall, 197 patients underwent RB. Median tumor size was 2.5 cm (Interquartile range 2-3.2). The rate of non-diagnostic RB was 8.6% (n=17/197), and the rate of tumour grading not assessable was 35% (69/197). The median (IQR) NoC and LoC were 2 (1-3) and 1.2 (0.6-1.6) cm, respectively. When Boruta algorithm was used only NoC and LoC were selected. By employing a KNN algorithm (optimal K value=9), a predictive model was constructed using these variables, achieving an accuracy of 75%. The decision boundaries used by the model (Fig. 1) indicate that the minimum number of cores to be taken should be 2 and that these cores should provide at least 0.8 cm of tissue or, alternatively, in case of RB with more than two cores no minimum tissue threshold is required. At multivariable logistic regression analysis, these indications resulted associated with higher probability of histological characterization (p=0.02) and grading (p<0.01) after adjusting for BMI, lesion size, location, and year of biopsy. CONCLUSIONS: Employing various machine learning algorithms, we defined an optimal renal biopsy strategy based on at least 2 cores and at least 0.8 cm of tissue or at least 3 cores and no minimum tissue threshold. These recommendations should be implemented in clinical practice to reduce non-diagnostic RB. Download PPT Source of Funding: None © 2024 by American Urological Association Education and Research, Inc.FiguresReferencesRelatedDetails Volume 211Issue 5SMay 2024Page: e597 Advertisement Copyright & Permissions© 2024 by American Urological Association Education and Research, Inc.Metrics Author Information Federico Belladelli More articles by this author Chiara Re More articles by this author Francesco Cei More articles by this author Giacomo Musso More articles by this author Giuseppe Rosiello More articles by this author Daniele Cignoli More articles by this author Daniela Canibus More articles by this author Francesco Fiorio More articles by this author Roberto Bertini More articles by this author Andrea Salonia More articles by this author Francesco De Cobelli More articles by this author Giorgio Brembilla More articles by this author Antonio Esposito More articles by this author Anna Palmisano More articles by this author Ciro Piccolo More articles by this author Marco Gambirasio More articles by this author Roberta Lucianò More articles by this author Nazario Tenace More articles by this author Alberto Briganti More articles by this author Francesco Montorsi More articles by this author Alessandro Larcher More articles by this author Umberto Capitanio More articles by this author Expand All Advertisement PDF downloadLoading ...