Lung Disease Classification from Chest X-ray Images Using Convolutional Neural Network and Long Short-Term Memory Model

The global expansion of lung illness necessitates the development of a reliable diagnostic technique. A lung illness diagnosis might be difficult for doctors to make. The goal of this approach is to develop a means of promptly identifying lung issues and halting the virus’s transmission. Five diseases are distinguished in the proposed framework: COVID-19, lung opacity, normal, tuberculosis, and viral pneumonia. The proposed system consists of three main steps: the preprocessing step, the feature extraction step, and the classification step. In the preprocessing step, to improve the quality of raw chest-Xray images using the image preprocessing method without data loss, this system proposes modelling the feature map and vector by using pre-trained Alexnet and LSTM. Machine learning techniques (such as SVM) are used for the categorization of lung diseases. Two publicly available datasets from Kaggle websites were used to create a dataset of COVID-19 3616, lung opacity 350, tuberculosis 500, normal 10192, and viral pneumonia 1345 chest X-ray images for this study. The dataset is divided into 70% training images and 30% testing images. The performance of the classifier is measured in experiments using accuracy, true positive, false negative, positive predictive, and false discovery rates. The purposed lung illness categorization method is built using MATLAB 2021b, which is used with an Intel Core i5 CPU running Windows 10 64-bit and 8GB of Random Access Memory (RAM). Finally, this system compares the proposed feature vector by adding the convolution and pooling layers of the pre-trained Alexnet and LSTM with the state-of-the-art pretrained Alexnet and LSTM.