Analysis of Multiview Integration Deep Learning Framework for Early Lung Disease Detection and Classification
Аннотация
Abstract Detecting lung diseases through clinical diagnosis, particularly pneumonia, tuberculosis and COVID-19, a health problem of international importance, is one of the most important tasks for the detection of lung diseases. Current deep learning (DL) approaches have been effective in such scenarios, but at the same time have some drawbacks, like difficulty in capturing the location of illness through AI and others. In order to promote the widespread application of lung disorder diagnostics in medical facilities, the researchers present a novel MVIDL algorithm that takes advantage of multiview CXR images to obtain an extremely high accuracy level in lung disease detection and classification. The suggested approach incorporates the edges of the multiview CXR image into a deep convolutional network, thus enabling the acquisition of rich spatial or contextual information for a more comprehensive understanding of lung diseases, by the already existing methods. Besides, the use of multiview imaging in training gives rise to more generalized models, lowers the number of wrong predictions, and increases the robustness over a wider range of lung disease types. In addition, the proposed framework provides XAI mechanisms as part of its features to facilitate interpretability and build up clinical trust by making the decision process transparent with visual representations. The MVIDL system tackles the limitations of traditional single-view and nonexplainable deep learning methods, thus providing a scalable, interpretable, and clinically relevant pathway for lung disease detection and classification. The experimental outcomes confirmed that the MVIDL model really excelled all the other state-of-the-art representations very much, obtaining 99.72% test accuracy, 99.75% precision, 98.25% recall, and an F1-score of 98.97%. This study is a step towards the development of highly efficient, smart, and understandable diagnostic tools that will enable radiologists to make timely clinical decisions for the early detection of lung disorders.