A Hybrid Deep Learning Approach for the Automated Classification of Lung Diseases Utilizing Chest X-ray Images

The increase in respiratory diseases such as COVID-19, pneumonia, and tuberculosis continues to pose a treatment challenge. This paper introduces a deep learning framework that combines CNN, MLP, & RNN architectures to improve classification accuracy. This study used a dataset comprising 538 COVID-19 images, 1341 normal images, 3875 pneumonia images, and 650 tuberculosis images with a training-to-test ratio of 80:20. The performance metric values of individual models of CNN, MLP, and RNN are compared with the hybrid model. Hybrid outperforms all other methods with 90% accuracy, outperforming CNN (83%), MLP (81%), and RNN (84%). The confusion matrix shows how the hybrid model reduces the classification error, resulting in better understanding and specificity. The training/testing and real schemes also ensure the strength and extensibility of the model. The results showed that integrating multiple deep architectures can improve the detection of lung diseases; future work will focus on optimal selection and multivariate integration to improve classification.