Pneumonia Detection Using Deep Neural Networks Based on Chest X-ray Images

Pneumonia is a serious condition that infects one or both of an individual's lungs where air sacs, scientifically known as alveoli, get inflamed or filled with fluid, causing difficulty in breathing. If not diagnosed at an early stage, such a disease can result in instant death. Radiologists must possess deep knowledge and go through thorough examinations like CBC tests and X-ray scans to proceed with treatment. However, the task is challenging because experts have varied levels of proficiency, and X-ray imaging is prone to subjective variations. Deep Learning domain, especially convolutional neural networks, has yielded promising results in predicting pneumonia. This paper focuses on implementing CNN, VGG16, and ResNet50 to detect pneumonia. After performance evaluation, it was found that ResNet50 provided the most accurate results with a validation accuracy of 87.5%. The confusion matrix was computed in the end as an evaluation metric to analyze the results.