Convolutional Neural Network for Automatic Pneumonia Detection in Chest Radiography

Abstract

X-ray imagery is a non-invasive method that involves exposure to small doses of ionizing radiation to parts of the body to help doctors diagnose diseases, including pneumonia. Detecting pneumonia on a chest X-ray image can be difficult for radiologists because X-ray images are often unclear, overlap with other diagnoses, and approach many other abnormalities. The automated method was developed as a decision support tool to help doctors diagnose pneumonia. This paper proposes different deep convolution neural network architectures with an augmentation strategy to classify the pneumonia detection from the chest X-ray images. We use three convolution layers and three classification layers (fully connected). Resize, flip, and rotation augmentation strategy to avoid overfitting. The experiment result shows that the augmentation strategy on the proposed CNN's architecture results in an accuracy value of 83,38% while on without augmentation result accuracy value 80,25%. The small difference between prediction results with the augmentation strategy and without the augmentation strategy shows that the proposed CNN's architecture can train small datasets.