Predicting Recurrence of Urinary Tract Infection in Paediatric Patients by Deep Learning Analysis of Tc-99m DMSA Renal Scan using Convolutional Neural Network

Introduction: Tc-99m dimercaptosuccinic acid (^99mTc-DMSA) renal scan is an important tool for assessment of childhood urinary tract infection (UTI), vesicoureteral reflux (VUR), and renal scarring. Recurrent UTI is associated with consecutive decline in renal function. Thus, it is important to predict recurrence of UTI especially in paediatric patients in terms of selecting appropriate therapeutic options, determining a follow-up plan, and preventing aggravation of renal dysfunction. We evaluated whether a deep learning (DL) analysis of ^99mTc-DMSA renal scans could predict the recurrence of UTI better than conventional clinical factors.

Methods: Subjects were 180 paediatric patients diagnosed with UTI, who underwent immediate post-therapeutic ^99mTc-DMSA renal scan. The primary outcome was the recurrence of UTI during the follow-up period. For the DL analysis, a convolutional neural network (CNN) model was used. Briefly, manually cropped ^99mTc-DMSA renal scan images were fed into the 2D-CNN model. Three sequential convolution layers, rectified linear unit, and pooling layers were applied, and the features were fed into a fully connected layer. Dropout layers were applied after the fully connected layers. The outputs of the network were probabilistic scores for the presence of recurrent UTI with values that ranged from zero to one. In addition, a stratified k-fold cross-validation was performed three times. Age, sex, presence of VUR, presence of cortical defect on ^99mTc-DMSA renal scan, split renal function (SRF), and DL prediction results were used as independent factors for predicting recurrent UTI. Diagnostic accuracy for predicting recurrent UTI was statistically compared between independent factors.

Results: During clinical follow-up, recurrent UTI occurred in 27 of 180 patients (15.0%). Significant differences were found in the presence of VUR, presence of cortical defect and split renal function (SRF) on ^99mTc-DMSA renal scan between patients with recurrent UTI and those without. The sensitivity, specificity and accuracy for predicting recurrent UTI were 44.4% (12/27), 88.9% (136/153) and 82.2% (148/180) by the presence of VUR; 44.4% (12/27), 76.5% (117/153) and 71.7% (129/180) by the presence of cortical defect; 74.1% (20/27), 80.4% (123/153) and 79.4% (143/180) by SRF (optimal cut-off = 45.93%); and 70.4% (19/27), 94.8% (145/153) and 91.1% (164/180) by the DL prediction. There were no significant difference in sensitivity between all independent factors (p > 0.05, for all). The specificity and accuracy of the DL prediction results were significantly higher than those of the other factors.

Conclusions: DL analysis of ^99mTc-DMSA renal scans may be useful for predicting recurrent UTI in paediatric patients. It is an efficient supportive tool to predict poor prognosis without visually demonstrable cortical defect in ^99mTc-DMSA renal scans.