DEVELOPMENT OF A PNEUMONIA DETECTION SYSTEM USING CONVOLUTIONAL NEURAL NETWORKS

Abstract

ABSTRACT

Pneumonia is one of the world’s most lethal and life-threatening diseases today, affecting people of all ages. Early proactive treatment can significantly reduce the likelihood of death from this illness and also prevent circumstances from worsening. Chest X-rays imaging has been one of the most well-liked and well-known clinical approaches. However, diagnosing the condition using X-rays has become increasingly challenging due to pneumonia resembling other lung disorders. As a result, this study developed a system to detect pneumonia disease in chest X-ray images using convolutional neural network-based approach. The datasets used for this study consists of 5856 chest X-ray images was obtained from Kaggle to train three pre-trained CNN architectures: VGG16, ResNet50, and MobileNetV3. The dataset was cleaned, preprocessed, and divided using the 80:20 data split ratio. Early stopping and learning rate reduction were applied to each model to prevent overfitting. The performance of each model was evaluated using accuracy, precision, recall, and f1-score on the test data. The VGG16 model outperformed others with 94% accuracy, 91% precision, 95% recall, and 93% f1-score. The MobileNetV3 model, which had the second-best performance, had an accuracy of 93%, precision of 90%, recall of 94%, and f1 score of 92%, while ResNet50 had 92% accuracy, 89% precision, 93% recall, and 91% f1-score. The best performing model of the three which is VGG16 was chosen and implemented on a web application. This system will serve as a tool for the medical practioners in detecting pneumonia earlier and accurately for proper treatment.