Lung disease classification using deep learning and genetic algorithm

Lung disorders are medical conditions that disrupt the lungs and their capacity to function normally. One fatal lung disease is a collapsed lung where the lung collapses partially or fully due to diseases like pneumothorax and atelectasis. The conventional approach to detecting such diseases is time-consuming, often requiring extensive manual analysis by trained experts, leading to delays in diagnosis and treatment. Computer-aided diagnostics have the potential to aid doctors in enhancing the consistency of diagnoses while also optimizing time efficiency. In this study, we enhance each lung X-ray image with three image enhancement techniques (i) contrast-limited adaptive histogram equalization (CLAHE), (ii) discrete wavelet transform (DWT), and (iii) gamma correction (GC) in parallel. A 3-channel convolutional neural network (CNN) then uses those enhanced images to extract features. The extracted features are further optimized using a genetic algorithm to improve the efficiency of the classification models. Our proposed model is validated with a dataset containing 9391 X-ray images to achieve an average precision, recall, and F1-score of 98, 97, and 98% and best accuracy of 99.33% which shows an improvement of 2.03% in terms of accuracy over the existing state-of-the-art methods.