Exploring Ensemble Architectures for Lung X-Ray Multi-Class Image Classification using CNN-LSTM

Early diagnose of lung diseases is important as it aids In the treatment of the diseases and increment of the patient’s quality of life. Chest X-rays are a widely used and cost-effective diagnostic tool for examining the lungs and detecting abnormalities. However, interpreting chest X-ray images can be challenging, even for experienced radiologists, due to the complexity of lung anatomies and the subtle variations in disease manifestations. Thus, the ensemble-based approach to Lung X-Ray Multi-Class image classification has been proposed and used Convolutional Neural Network and Long Short-term Memory (CNN-LSTM). The proposed model is compared with Several models such as ensemble of DenseNet and InceptionV3, Vgg16 and Mobilenet V2 and deep bidirectional LSTM(DBLSTM). The proposed model applied for multiclass classification on lung disorders such as TB, normal, pneumothorax, pneumonia, COVID-19, and cardiomegaly. The proposed method provides a testing F1-Score of 94%, Recall of 99%, Precision of 95% and accuracy of 89.31% compared to other ensemble models. The proposed model can be utilized for medicos as a second opinion for diagnosis and treatment planning.