Diagnosis of pediatric pneumonia through different types of deep convolutional neural network in chest x-ray images

Abstract

Background: Pneumonia is a common fatal disease for children below 5 years. Early diagnosis and treatment can help decreasing mortality rate. Diagnosis of pneumonia in chest X-ray images can be difficult and error prone. The aim of this study is to evaluate a computer-aided pneumonia detection system to facilitate the diagnosis.

Materials and Methods: In this context, five well-known convolutional neural networks (CNN) models including basic CNN, Inception-V3, ResNet-50, VGG-16 and MobileNet-V2 were trained with appropriate transfer learning on the chest X-ray dataset.

Result: Among the five different models, MobileNet is the most successful algorithm with 89 % accuracy. The lower accuracy models, in descending order, are Inception 80.13 %, Basic CNN 78.85 %, VGG-16 77.08 %, and lastly ResNet 72.44 % Conclusion: Diagnosis of pediatric pneumonia with deep convolutional neural network in chest X-ray images demonstrate acceptable accuracy. Computer-aided diagnostic approach would be helpful to diagnose pneumonia.

Conclusion: Diagnosis of pediatric pneumonia with deep convolutional neural network in chest X-ray images demonstrate acceptable accuracy. Computer-aided diagnostic approach would be helpful to diagnose pneumonia.