<![CDATA[Pneumonia is a respiratory infection that remains a leading cause of death, especially in children, requiring an automatic detection system based on chest X-ray images. The main challenge in automatic classification is low image quality, such as suboptimal contrast and unclear lung details, which can affect the feature extraction process by deep learning models. To address these issues, this study applies Contrast Limited Adaptive Histogram Equalization (CLAHE) to enhance image contrast and a sharpening filter to clarify lung edge details. The study aims to analyze the effect of preprocessing on classification performance using EfficientNet-B0 based on Transfer Learning with a full fine-tuning strategy. The dataset used is Chest X-Ray Pneumonia from Kaggle with 5,856 images consisting of Normal and Pneumonia classes. Experiments compare the Baseline model, CLAHE, and a combination of CLAHE and sharpening in three data sharing scenarios. Evaluation is carried out using accuracy, precision, recall, and image quality metrics PSNR, SSIM, and CII. The results of the study showed that the combination of CLAHE and sharpening in the 80:10:10 scenario produced the best performance with an accuracy of 97.61%, precision of 0.97, recall of 0.99, and an increase in image quality based on a CII value of 1.157.]]>
Copyrights © 2026
