Background and Aims Early identification of cardiac structural abnormalities indicative of heart failure is crucial to improving patient outcomes. Chest X-rays (CXRs) are routinely conducted on a broad population of patients, presenting an opportunity to build scalable screening tools for structural abnormalities indicative of Stage B or worse heart failure with deep learning methods. In this study, a model was developed to identify severe left ventricular hypertrophy (SLVH) and dilated left ventricle (DLV) using CXRs.Methods A total of 71 589 unique CXRs from 24 689 different patients completed within 1 year of echocardiograms were identified. Labels for SLVH, DLV, and a composite label indicating the presence of either were extracted from echocardiograms. A deep learning model was developed and evaluated using area under the receiver operating characteristic curve (AUROC). Performance was additionally validated on 8003 CXRs from an external site and compared against visual assessment by 15 board-certified radiologists.Results The model yielded an AUROC of 0.79 (0.76-0.81) for SLVH, 0.80 (0.77-0.84) for DLV, and 0.80 (0.78-0.83) for the composite label, with similar performance on an external data set. The model outperformed all 15 individual radiologists for predicting the composite label and achieved a sensitivity of 71% vs. 66% against the consensus vote across all radiologists at a fixed specificity of 73%.Conclusions Deep learning analysis of CXRs can accurately detect the presence of certain structural abnormalities and may be useful in early identification of patients with LV hypertrophy and dilation. As a resource to promote further innovation, 71 589 CXRs with adjoining echocardiographic labels have been made publicly available. Structured Graphical Abstract The deep learning model that was developed takes as input a pre-processed chest X-ray of dimension 224-by-224, age, and sex; the model outputs a probability for dilated left ventricle, severe left ventricular hypertrophy, and a composite label indicating the presence of either structural abnormality. The model outperforms all 15 board-certified radiologists in the task of detecting the presence of either abnormality (composite label). For a single point of comparison, we used the consensus vote amongst radiologists. The model achieves a sensitivity of 71% compared with the consensus vote sensitivity of 66% at a fixed specificity of 73%. Saliency maps demonstrate that at shallower layers in the network, the model is sensitive to the broader cardiac silhouette as well as structures in the left heart. AI, artificial intelligence; AUROC, area under the receiver operating characteristic curve; CXR, chest X-ray; DLV, dilated left ventricle; IVSd, interventricular septal thickness at end-diastole; LVIDd, left ventricular internal diameter at end-diastole; LVPWd, left ventricular posterior wall distance at end-diastole; SLVH, severe left ventricular hypertrophy.

PURPOSE:Preservation of renal parenchyma is a major goal when performing a partial nephrectomy. IRIS anatomical visualization software generates a segmented 3D model, allowing improved visualization of the tumor and surrounding structures. We hypothesize that using IRIS intraoperatively during partial nephrectomy on complex tumors increases the precision of surgical procedures and therefore may result in more tissue preservation.METHODS:We identified 74 non-IRIS and 19 IRIS patients who underwent partial nephrectomy, with nephrometry scores of 9, 10, and 11. Propensity scores were used to match 18 pairs of patients on nephrometry score, age, and tumor volume. Pre- and postoperative imaging (MRI/CT) was obtained. Volumes of the preoperative tumor and preoperative whole kidney were obtained to calculate predicted postoperative whole kidney volume and then compared to actual postoperative whole kidney volume.RESULTS:Mean differences between predicted and actual postoperative whole kidney volumes were 19.2 cm3 (SD=20.2) and 32 cm3 (SD=16.1, P = .0074) for IRIS and non-IRIS groups, respectively. The mean improvement in precision for the IRIS procedure was 12.8 cm3 (95% confidence interval, 2.5 to Inf; P = .02). There was no significant change in mean glomerular filtration rate from baseline to 6 months postoperatively between IRIS and non-IRIS groups (-6.39, SD=15.8 vs -9.54, SD=13.3; P = .5). No significant differences in complication rates (0 vs 1, P = .2), worsening glomerular filtration rate staging (5 vs 4, P = 1), and >25% decrease in glomerular filtration rate (3 vs 4, P = 1) were found between IRIS and non-IRIS groups.CONCLUSIONS:We demonstrated that using IRIS intraoperatively when performing partial nephrectomy on complex tumors is associated with improved surgical precision.

You have accessJournal of UrologyCME1 Apr 2023MP68-15 DOES UTILIZING IRIS, A SEGMENTED THREE-DIMENSIONAL MODEL, PRESERVE RENAL PARENCHYMAL VOLUME DURING ROBOTIC PARTIAL NEPHRECTOMY? Teona Iarajuli, Christina Caviasco, Tanner Corse, Katherine Kim, Jennifer Nguyen, Ruth Sanchez DE LA Rosa, Simon Gelman, Nick Spagnuolo, Hannah Sidoti, Mitchell Miller, and Michael Stifelman Teona IarajuliTeona Iarajuli More articles by this author , Christina CaviascoChristina Caviasco More articles by this author , Tanner CorseTanner Corse More articles by this author , Katherine KimKatherine Kim More articles by this author , Jennifer NguyenJennifer Nguyen More articles by this author , Ruth Sanchez DE LA RosaRuth Sanchez DE LA Rosa More articles by this author , Simon GelmanSimon Gelman More articles by this author , Nick SpagnuoloNick Spagnuolo More articles by this author , Hannah SidotiHannah Sidoti More articles by this author , Mitchell MillerMitchell Miller More articles by this author , and Michael StifelmanMichael Stifelman More articles by this author View All Author Informationhttps://doi.org/10.1097/JU.0000000000003331.15AboutPDF ToolsAdd to favoritesDownload CitationsTrack CitationsPermissionsReprints ShareFacebookLinked InTwitterEmail Abstract INTRODUCTION AND OBJECTIVE: Preservation of renal parenchymal volume is one of the major goals of surgeons performing a partial nephrectomy (PN). IRIS anatomical visualization software generates a segmented 3D model, allowing surgeons to better visualize the tumor, renal vasculature, collecting system, and their relationship. We hypothesize that using a segmental 3D model intra-operatively during PN on complex tumors would improve renal parenchymal volume preservation. METHODS: We identified 74 non-IRIS and 19 IRIS patients who underwent PN from January 1st, 2019 to January 1st, 2022, with nephrometry scores of 9, 10, 11, and had complete pre and postoperative axial imaging. We completed propensity score matching (PSM) based on nephrometry score, age, BMI, and tumor volume and finalized 18 PS matched pairs. Pre and postoperative imaging (MRI or CT) were obtained from patient charts on Fuji PACS and quality screened by our researchers. Volumes of the preoperative renal tumor mass and whole kidney volume were obtained and used to calculate predicted postoperative volume. Measured postoperative renal volumes were compared to the predicted postoperative volumes. Statistical analysis was performed using a paired one-sided t-test. All analyses were performed using R. RESULTS: The mean predicted postoperative kidney volume in the IRIS and non-IRIS groups were 194.1 cm3 (71.5) and 218.9 cm3 (SD=48.9), and postoperative whole kidney volume was 175.0 cm3 (SD=73.5) and 186.9 cm3 (SD=47.2). Mean difference between predicted and actual postoperative whole kidney volume in the IRIS group was statistically smaller, 19.2 cm3 (SD=20.2), than the non-IRIS group, 32 cm3 (SD=16.1, p=0.0074). Percent difference from predicted kidney volume in the IRIS group was 10.9% vs. 17.1% in the non-IRIS group. There was no significant change in mean GFR from baseline to 6-month postoperative period between IRIS and non-IRIS groups (-6.39, SD=15.8 vs. -9.54, SD=13.3; p=0.374). Additionally, no significant difference in complication rates (0 vs. 1, p=0.235), frequency of patients who experienced worsening GFR staging (5 vs 4, p=1.00) and >25% decrease in GFR levels (3 vs. 4, p=1.00) were found between IRIS and non-IRIS groups respectively. CONCLUSIONS: We demonstrated that using a segmental 3D model intra-operatively when performing a partial nephrectomy on complex tumors improves renal parenchymal preservation. The long-term clinical significance of this preservation remains to be determined. Source of Funding: No Funding © 2023 by American Urological Association Education and Research, Inc.FiguresReferencesRelatedDetails Volume 209Issue Supplement 4April 2023Page: e959 Advertisement Copyright & Permissions© 2023 by American Urological Association Education and Research, Inc.MetricsAuthor Information Teona Iarajuli More articles by this author Christina Caviasco More articles by this author Tanner Corse More articles by this author Katherine Kim More articles by this author Jennifer Nguyen More articles by this author Ruth Sanchez DE LA Rosa More articles by this author Simon Gelman More articles by this author Nick Spagnuolo More articles by this author Hannah Sidoti More articles by this author Mitchell Miller More articles by this author Michael Stifelman More articles by this author Expand All Advertisement PDF downloadLoading ...