<p>Lung disease nowadays is a very common disease around the globe, and it encompasses conditions, such as pneumonia, obstructive pulmonary disease, asthma, fibrosis, and tuberculosis. Hence, prompt and accurate diagnosis of lung diseases is significant. Numerous approaches using image processing and machine learning have been designed for this crucial purpose. In this work, a classifier driven by optimization is introduced for the detection of lung disease. Initially, image pretreatment is done by Gaussian filter, and afterward, segmentation is done by improved attention U-Net mechanism. The final step involves lung disease detection by employing the AlexNet, which is trained using the proposed optimization approach, known as Serial Exponential Chameleon Swarm Algorithm (SExpCSA). The proposed algorithm is designed by integration of Serial Exponential Weighted Moving Average concept into Chameleon Swarm Algorithm. Finally, the experimentation evaluation is done for adopted approach over existing techniques using measures, like sensitivity, accuracy, and specificity. The evaluation depicts that SExpCSA-based AlexNet model achieved a maximum accuracy of 0.953, maximum specificity of 0.988, and maximal sensitivity of 0.947.</p>