Mp30-12 Artificial Intelligence Convolutional Neural Network for Efficient and Accurate Ct-Based Stone Volume Determination

You have accessJournal of UrologyImaging/Uroradiology II (MP30)1 May 2024MP30-12 ARTIFICIAL INTELLIGENCE CONVOLUTIONAL NEURAL NETWORK FOR EFFICIENT AND ACCURATE CT-BASED STONE VOLUME DETERMINATION Andrei D. Cumpanas, Chanon Chantaduly, Kalon L. Morgan, Wei Shao, Candice M. Tran, Yi Xi Wu, Amanda McCormac, Jacob C. Tsai, Jaime Altamirano Villaroel, Zachary E. Tano, Bruce Gao, Roshan M. Patel, Pengbo Jiang, Peter Chang, Jaime Landman, and Ralph V. Clayman Andrei D. CumpanasAndrei D. Cumpanas , Chanon ChantadulyChanon Chantaduly , Kalon L. MorganKalon L. Morgan , Wei ShaoWei Shao , Candice M. TranCandice M. Tran , Yi Xi WuYi Xi Wu , Amanda McCormacAmanda McCormac , Jacob C. TsaiJacob C. Tsai , Jaime Altamirano VillaroelJaime Altamirano Villaroel , Zachary E. TanoZachary E. Tano , Bruce GaoBruce Gao , Roshan M. PatelRoshan M. Patel , Pengbo JiangPengbo Jiang , Peter ChangPeter Chang , Jaime LandmanJaime Landman , and Ralph V. ClaymanRalph V. Clayman View All Author Informationhttps://doi.org/10.1097/01.JU.0001009416.90901.7b.12AboutPDF ToolsAdd to favoritesDownload CitationsTrack CitationsPermissionsReprints ShareFacebookLinked InTwitterEmail Abstract INTRODUCTION AND OBJECTIVE: Renal stones are often irregularly shaped and as such, their true volume may be incompletely characterized when relying on only linear dimensions. Accordingly, we sought to compare the precision of a University of California Irvine (UCI) artificial intelligence (AI) algorithm for stone volume determination to the three "best-fit" ellipsoid formulas available for volumetric stone assessment among 322 calculi. METHODS: A total of 322 non-contrast computerized tomography (NCCT) scans were retrospectively obtained from patients with a urolithiasis diagnosis. The largest stone was designated as the "index stone". To assess the convolutional neural network's (CNN) performance in determining stone volume, both a Pearson correlation coefficient (R) and a Dice Overlap Score were calculated. Using a CT-based 3D slicer program, the "ground truth" volume of the index stone was determined by an experienced research physician fellow (expertise validated by comparison with volumetric gas pycnometry, with an R of 0.99). The AI-calculated index stone volume was compared with the "ground truth" volume as well as the volume determined by the scalene, prolate and oblate-ellipsoid formulas (Figure 1). RESULTS: The UCI AI algorithm had an R value of 0.99 and a Dice score of 0.96, indicating a high level of accuracy and preciseness. The CNN outperformed the 3-ellipsoid formula-based volume predictions for stones of all sizes (Table 1). The algorithm's accuracy and precision improved when measuring larger, often more irregularly shaped stones; in contrast, the ellipsoid-determined volumes displayed an opposite trend. Even the best-fit ellipsoid formulas overestimated the true stone volume by a mean of 27% to 89%, depending on the urolith's size. CONCLUSIONS: Ellipsoid formulas fail to fully account for the irregular shape of a kidney stone, thereby rendering an inaccurate overestimated volume, especially for larger stones. In contrast, the UCI AI algorithm proved to be both accurate and precise in assessing stone volume, regardless of stone size. Download PPT Source of Funding: None © 2024 by American Urological Association Education and Research, Inc.FiguresReferencesRelatedDetails Volume 211Issue 5SMay 2024Page: e496 Advertisement Copyright & Permissions© 2024 by American Urological Association Education and Research, Inc.Metrics Author Information Andrei D. Cumpanas More articles by this author Chanon Chantaduly More articles by this author Kalon L. Morgan More articles by this author Wei Shao More articles by this author Candice M. Tran More articles by this author Yi Xi Wu More articles by this author Amanda McCormac More articles by this author Jacob C. Tsai More articles by this author Jaime Altamirano Villaroel More articles by this author Zachary E. Tano More articles by this author Bruce Gao More articles by this author Roshan M. Patel More articles by this author Pengbo Jiang More articles by this author Peter Chang More articles by this author Jaime Landman More articles by this author Ralph V. Clayman More articles by this author Expand All Advertisement PDF downloadLoading ...