Deep learning algorithms for detecting and visualising intussusception on plain abdominal radiography in children: a retrospective multicenter study

Gitaek Kwon,Junwon Bae,Jongwoo Lim,Dong Keon Lee,Jongbin Ryu,Bo-Kyeong Kang,Chiwon Ahn,Jaehoon Oh

doi:10.1038/s41598-020-74653-1

Abstract

This study aimed to verify a deep convolutional neural network (CNN) algorithm to detect intussusception in children using a human-annotated data set of plain abdominal X-rays from affected children. From January 2005 to August 2019, 1449 images were collected from plain abdominal X-rays of patients ≤ 6 years old who were diagnosed with intussusception while 9935 images were collected from patients without intussusception from three tertiary academic hospitals (A, B, and C data sets). Single Shot MultiBox Detector and ResNet were used for abdominal detection and intussusception classification, respectively. The diagnostic performance of the algorithm was analysed using internal and external validation tests. The internal test values after training with two hospital data sets were 0.946 to 0.971 for the area under the receiver operating characteristic curve (AUC), 0.927 to 0.952 for the highest accuracy, and 0.764 to 0.848 for the highest Youden index. The values from external test using the remaining data set were all lower (P-value < 0.001). The mean values of the internal test with all data sets were 0.935 and 0.743 for the AUC and Youden Index, respectively. Detection of intussusception by deep CNN and plain abdominal X-rays could aid in screening for intussusception in children.

Highlights

This study aimed to verify a deep convolutional neural network (CNN) algorithm to detect intussusception in children using a human-annotated data set of plain abdominal X-rays from affected children
Because the mean annual intussusception incidence rate is approximately 30 per 100,000 live births in the first 3 year of life[3], using ultrasound as a screening exam to rule out intussusception in all children who present with nonspecific signs and symptoms is difficult
In a study on the use of risk stratification in evaluating intussusception in children, it was found that abdominal radiography could be used as the initial diagnostic modality to identify children at risk with sensitivity and specificity values of 0.77 and 0.79, r espectively[22]

Summary

Introduction

This study aimed to verify a deep convolutional neural network (CNN) algorithm to detect intussusception in children using a human-annotated data set of plain abdominal X-rays from affected children. Detection of intussusception by deep CNN and plain abdominal X-rays could aid in screening for intussusception in children. Hydrostatic or pneumatic enemas were considered the gold standards for both diagnosing and treating intussusception[6] These are invasive radiologic procedures that must be performed by radiologists and are not always readily available[7]. There are no previous studies on the availability and external validity of deep learning in diagnosing intussusception using large data sets of plain abdominal radiographs. This study aimed to create a human-annotated data set of plain abdominal X-rays of children with intussusception for internal and external validation, and to verify a possibility of deep CNN to detect intussusception with this set

Objectives

Methods

Results

Conclusion