A pneumothorax (PTX) is a serious condition that can cause death in people because of breathing difficulties. Therefore, the identification of lesions in the lungs, such as pneumothorax, is a significant and challenging task. However, this challenge can be addressed through artificial intelligence techniques. On this subject, chest X-ray (CXR) images can be utilized for diagnosing this disease. This research offers a structure for the automated and effective detection of pneumothorax in CXR images by leveraging pre-trained Deep Convolutional Neural Networks (DCNN). Moreover, pre-processing techniques have been utilized for X-ray images to improve contrast and remove noise by enhancing as well as filtering the images. Additionally, Pre-trained network architectures including DenseNet-169, ResNet-50, and EfficientNet-B3 have been employed for classification. Performance assessment was carried out by utilizing evaluation parameters to compare the effectiveness of the deep CNN models. After a series of experiments, the proposed framework achieved diagnostic accuracy of 0.9517, 0.9639, and 0.9777 for ResNet-50, EfficientNet-B3, and DenseNet-169 respectively, using the publicly available SIIM-ACR dataset. The proposed framework holds promise for dependable utilization in clinical settings, aiding doctors and clinicians in the diagnosis of lung conditions.