An Integrative Hybrid Deep Learning Framework for Complex Multifaceted Lung Pathologies Classification using Chest X-Rays

Timely diagnosis and patient survival are contingent upon the early detection of lung illnesses. We offer a strategy that employs both convolutional neural networks (CNNs) and capsule networks for the categorization of lung diseases from chest x-ray pictures. The suggested model makes use of the hierarchical feature representation provided by capsule layers as well as the feature extraction capabilities of DenseNet, GoogleLeNet, and SENet architectures. The hybrid architecture effectively captures and utilizes the complex spatial relationships found in medical images, which has improved the diagnostic power and resilience of the classification system. The study employs a stochastic ensemble classifier that combines predictions from various CNNs and capsule networks, achieving a total accuracy of 95.56 % in detecting different lung diseases. The use of advanced deep learning, as demonstrated in the context of this study, paves the way to improving medical image analysis, which initiates a new medium for further developments.