Cascaded Deep Learning Model for Detecting Lung Infections Using Chest X-Rays

Abstract

This work focuses on efforts for accurately predicting lung diseases like omicron and pneumonia using chest X-ray imaging, a reliable method in this domain. The work adopts a transfer learning model for lung infection predictions from chest X-ray images. The proposed architecture encompasses both training and testing functions, with key steps including pre-processing, deep feature extraction, and classification. Initially, each X-ray image is enhanced through digital filtering for quality improvement. These processed images are then input into a robust, step-wise learning model that efficiently facilitates the automatic learning of features. The highlight of this approach is the Cascaded learning model, which not only achieves a high accuracy rate of 99% but also significantly reduces computational complexity. This is evidenced by a lower number of training parameters, making the model both more efficient and lightweight, and hence more practical for clinical applications in differentiating between omicron and pneumonia.