Pediatric Pneumonia Detection in Chest X-ray Images using Convolutional Neural Network

Pneumonia is an inflammation of the lung which is a major cause of morbidity and mortality worldwide, especially among infants under the age of five. WHO reported there were 1.4 million pneumonia-related deaths among children under the age of five in 2018. Rapid and accurate diagnosis of pneumonia is essential to accelerate treatment and improve patient outcomes. Machine learning can be used to assist the screening process to detect pneumonia. Machine learning can be divided in two categories: traditional machine learning method and deep learning. The most popular deep learning method for image classification is Convolutional Neural Network. In this paper, a custom convolutional neural network based binary image classification is developed to automatically detect pneumonia on pediatric X-ray image data. Four different configurations are investigated and successful in detecting pneumonia with best result obtained by a CNN with 4 convolutional layers utilizing image augmentation with accuracy $\mathbf{9 2. 1 5 \%}$, precision $\mathbf{9 2. 6 8 \%}$, recall $\mathbf{9 2. 1 5 \%}$ and $\mathrm{F}-1$ score $91.98 \%$.