Enhancing COVID-19 Diagnosis Accuracy and Transparency with Explainable Artificial Intelligence (XAI) Techniques

The COVID-19 pandemic has strained global healthcare systems, highlighting the need for efficient screening tools for timely SARS-CoV-2 detection and management. This research presents a novel approach using machine learning and Explainable AI (XAI) techniques to develop an effective screening tool. An XAI approach for COVID-19 detection using transfer learning with X-ray images. Heliyon, 9(4), e15137.). The model is trained on a dataset of over 50,000 individuals, utilizing features such as age, sex, contact history, and clinical symptoms to accurately estimate infection risk. The incorporation of XAI methods, including feature importance analysis and decision boundary visualization, enhances the model’s transparency and reliability, fostering greater confidence among healthcare professionals. This tool addresses comprehension challenges and leverages diverse knowledge for diagnosis, thereby empowering medical staff in triaging patients and optimizing resource allocation. This research significantly contributes to combating the COVID-19 pandemic and bolstering the resilience of healthcare systems against similar future challenges. The model achieved an auROC of 0.58 and a high accuracy score of 0.9282 on the test set, indicating its ability to discriminate between positive and negative COVID-19 test results and validating its predictive performance.