Deep Learning Innovations for Lung Disease Diagnosis and Severity Evaluation: A Comprehensive Review
TL;DR
This study will discuss about the advances in improving practical clinical deployment through lightweight architectures such as GANs, solving the problem of data imbalance and the use of multimodal images including X-ray, CT, and histopathological images, which solves the problems associated with medical trust.
Abstract
The lung diseases like lung cancer, COVID-19, interstitial lung disease (ILD) and pneumonia still remain one of the leading causes of death in the world. There is a need for quick and precise diagnostic tools to overcome the existing disease diagnosis challenges. Despite the deep learning providing advances to traditional computer-aided diagnostics (CAD), still the issues such as computational complexity, lack of data, and explainability remain unaddressed. This review will present an overview of the recent research advances in the area of deep learning, where the hybrid architecture of convolutional neural networks with transformers is used to solve the problem. This study will discuss about the advances in improving practical clinical deployment through lightweight architectures such as GANs, solving the problem of data imbalance and the use of multimodal images including X-ray, CT, and histopathological images. The use of Explainable AI (XAI) includes visual heat maps like Grad-CAM and Bayesian uncertainty quantification, which solves the problems associated with medical trust. The importance of multi-center validation is highlighted in order to properly integrate efficient diagnostic assistants.
