Pneumonia Detection in Children Chest X-ray Images Using Convolutional Neural Networks

Pneumonia is the most common diagnosis found in cases of lung disease in children. Systematic early childhood disease diagnosis is often time-consuming and vulnerable to errors. Radiologists, on the other hand, have a difficult time identifying pneumonia by chest X-rays, which must be manually examined. The purpose of this study was to develop an automated classification system of pneumonia images using deep learning to assist clinical diagnosis. This study used the Convolutional Neural Network (CNN) method to classify normal lungs and pneumonia lungs in children. The data used are secondary data obtained from a retrospective cohort of pediatric patients aged one to five years from Guangzhou Women and Children's Medical Center, Guangzhou, China. The data that has been prepared undergoes a pre-processing process, namely performing data augmentation. This study used the VGG16, VGG19, InceptionV3 and ResNet50 of CNN models for recognition and classification pneumonia images. System evaluation is done using a Confusion matrix and ROC curve by calculating the Area Under Curve (AUC) value. VGG16 architecture with 100 epochs had the highest accuracy value, with accuracy of 95.51%, sensitivity of 90.6%, specificity of 98.46%, and AUC of 94.53%. The findings of this study will aid researchers who will perform medical research using the CNN technology and medical professionals in improving pneumonia diagnosis in children.