Explainable AI and Computational Modeling for Real-Time COVID-19 Detection Using Medical Imaging

Abstract

The global rise in population has made disease monitoring a critical challenge, highlighting the need for automated detection systems to improve diagnostic accuracy and reduce mortality. The COVID-19 pandemic has emphasized the importance of rapid diagnostic tools. This study proposes an explainable framework for COVID-19 detection using CT scans and chest X-rays, combining deep learning and machine learning. A CNN extracts features from images, which are classified using an ensemble of DT, RF, GNB, LR, KNN, and SVM models. A Susceptible-Infectious-Recovered (SIR) model is integrated to estimate virus transmission and support interpretability. Grad-CAM and t-SNE analyses validate feature importance and separability. Tested on two datasets (1,646 and 2,481 images), the proposed method achieved 98.5% accuracy, 99.2% precision, and 99.4% recall. Comparative analyses demonstrate superior performance, and explainable AI experiments confirm the robustness and transparency of the framework.

Ähnliche Arbeiten