You have accessJournal of UrologyCME1 Apr 2023PD37-07 A MACHINE LEARNING TOOL TO DETERMINE OBSTRUCTION IN CHILDREN WITH HYDRONEPHROSIS FROM ROUTINE SONOGRAPHIC FINDINGS Adree Khondker, Jethro Kwong, Jin Kyu Kim, Michael Chua, Margarita Chancy, Natasha Brownrigg, Joana Santos, Lauren Erdman, Neeta D'Souza, John Weaver, Mandy Rickard, and Armando Lorenzo Adree KhondkerAdree Khondker More articles by this author , Jethro KwongJethro Kwong More articles by this author , Jin Kyu KimJin Kyu Kim More articles by this author , Michael ChuaMichael Chua More articles by this author , Margarita ChancyMargarita Chancy More articles by this author , Natasha BrownriggNatasha Brownrigg More articles by this author , Joana SantosJoana Santos More articles by this author , Lauren ErdmanLauren Erdman More articles by this author , Neeta D'SouzaNeeta D'Souza More articles by this author , John WeaverJohn Weaver More articles by this author , Mandy RickardMandy Rickard More articles by this author , and Armando LorenzoArmando Lorenzo More articles by this author View All Author Informationhttps://doi.org/10.1097/JU.0000000000003335.07AboutPDF ToolsAdd to favoritesDownload CitationsTrack CitationsPermissionsReprints ShareFacebookLinked InTwitterEmail Abstract INTRODUCTION AND OBJECTIVE: For children with hydronephrosis (HN), the diuretic renogram is the gold standard in evaluating patients with suspected obstruction. Herein, we aimed to use routinely reported ultrasound findings, along with machine learning approaches, to predict the risk of ureteropelvic junction obstruction (UPJO) in infants with isolated HN. METHODS: We included patients less than 24 months of age at baseline with a renogram within 3 months of an ultrasound. Age, sex, and routinely reported ultrasound findings (laterality, kidney length, anteroposterior diameter [APD], SFU grade) were abstracted. T 1/2 washout periods were collected from renography and stratified as low risk (<20 minutes), unclear risk (20-60 minutes), and high risk of obstruction (>60 minutes). A random forest model was trained to classify obstruction risk, referred to as AI Evaluation of Renogram Obstruction (AERO). Model performance was determined by measuring area under the receiver-operator-characteristic curve (AUROC) and decision curve analysis. RESULTS: A total of 304 patients met inclusion criteria, with a median age of diuretic renogram at 4 months (IQR 2, 7). Of all patients, 48 (16%) were low-risk, 102 (33%) were unclear risk, 156 (51%) were high risk of obstruction based on diuretic renogram. AERO achieved a multi-class AUROC of 0.75 which was superior to logistic regression (Figure 1). The most important features for prediction included age, APD, and SFU grade. We deployed our model in an easy-to-use application (https://sickkidsurology.shinyapps.io/AERO/). At a threshold cutoff of 30%, AERO would allow 137 more patients per 1000 to safely avoid a renogram without missing significant obstruction compared to a strategy in which a renogram is performed for all patients with SFU grade 3 or above. CONCLUSIONS: Routine ultrasound findings can be used to determine if a diuretic renogram can be safely avoided for children with isolated hydronephrosis, thus offering the potential to minimize invasiveness of monitoring and exposure to radiation. Source of Funding: None © 2023 by American Urological Association Education and Research, Inc.FiguresReferencesRelatedDetails Volume 209Issue Supplement 4April 2023Page: e989 Advertisement Copyright & Permissions© 2023 by American Urological Association Education and Research, Inc.MetricsAuthor Information Adree Khondker More articles by this author Jethro Kwong More articles by this author Jin Kyu Kim More articles by this author Michael Chua More articles by this author Margarita Chancy More articles by this author Natasha Brownrigg More articles by this author Joana Santos More articles by this author Lauren Erdman More articles by this author Neeta D'Souza More articles by this author John Weaver More articles by this author Mandy Rickard More articles by this author Armando Lorenzo More articles by this author Expand All Advertisement PDF downloadLoading ...
Read full abstract