Enhanced COVID-19 Detection Through Combined Image Enhancement and Deep Learning Techniques

The rapid spread of COVID-19 has highlighted the need for automated patient data analysis to enable faster and more accurate diagnosis. Using pre-trained deep learning models on X-ray images has shown potential for effective COVID-19 detection. However, the performance of these models is highly dependent on the quality and quantity of training data. To address these challenges, enhancing the visual quality of X-ray images is critical for reliable virus detection. This study evaluates and combines three image enhancement techniques—Histogram Equalization, Contrast-Limited Adaptive Histogram Equalization (CLAHE), and Gamma Correction—to determine the optimal approach for improving detection accuracy. A dataset comprising 125 chest X-ray images from COVID-19-positive patients and 500 images from non-COVID-19 cases was used. The images were preprocessed using the enhancement techniques, and the enhanced datasets were employed to train ResNet50 and DenseNet201 models. Simulation results demonstrate that enhanced images consistently yield higher detection accuracy than unenhanced images. Among the techniques tested, combining Histogram Equalization, CLAHE, and Gamma Correction with the DenseNet201 model achieved the highest performance, attaining a remarkable accuracy of 99.03%. This outperforms previous methods, including the DarkCovidNet model, which achieved an accuracy of 98.08% on the same dataset.