Customized convolutional neural network for pulmonary multi-disease classification using chest x-ray images

The development of accurate and reliable diagnostic tools is crucial for the timely and effective treatment of pulmonary diseases. However, in the midst of a pandemic, it is also important to have faster diagnosis of other types of pneumonia using reliable assistive technology to ensure effective treatment. To address this need, a novel DWTMBConvNet Convolutional Neural Network (CNN) model has been proposed. This model aims to detect multiple pulmonary diseases, including normal/healthy conditions, tuberculosis, Covid-19, bacterial pneumonia, and non-Covid-19 viral pneumonia. By analyzing Chest X-Ray (CXR) images, the model extracts relevant features for the classification of these diseases. The DWTMBConvNet model incorporates wavelet features and EfficientNet-inspired MBConv blocks to achieve impressive accuracy. With an accuracy of 0.955, specificity of 0.989, and sensitivity of 0.955 for five-way classification, the model outperforms several state-of-the-art CNN models. The discrete wavelet transform is employed to extract frequency and location-based features, which are informative for detecting pulmonary diseases. The EfficientNet-inspired MBConv blocks allow the model to learn highly discriminative features while using fewer parameters, thereby improving its performance. The model's accuracy for binary, three way, four way and five is analyzed which outperforms the various state-of-the-art models. These results showcase the effectiveness of the proposed DWTMBConvNet model in detecting pulmonary diseases with a high degree of accuracy. Consequently, it serves as a valuable tool for clinical diagnosis, aiding in the timely and effective treatment of patients.