Comparison of Data Augmentation Techniques for Training CNNs to Detect Pneumonia from Chest X-Ray Images

Pneumonia, a primary cause of mortality in children globally, is typically identified through chest X-ray imaging. This study examines the effectiveness of three different image augmentation techniques for training convolutional neural networks (CNNs) to accurately detect pneumonia using chest X-ray images. The researchers used a publicly accessible dataset consisting of 5,863 chest X-ray images, where 3,875 had indications of pneumonia, and 1,341 were normal. The dataset was partitioned into training, validation, and testing subsets to educate three CNN models employing various image augmentation approaches. Overall, the findings indicate that employing data augmentation methods can enhance the efficiency of CNNs in identifying pneumonia from chest X-ray images. However, the choice augmentation technique should depend on the specific performance metric of interest. This research underlines the possible advantages of image augmentation methods in enhancing the precision of CNNs for diagnosing pneumonia, which may aid in the prompt detection and treatment of this lethal illness.