Explainable Artificial Intelligence Approach for the Early Prediction of Ventilator Support and Mortality in COVID-19 Patients

Early prediction of mortality and risk deterioration in COVID-19 patients can reduce increase the opportunity for better more timely treatment. In current study, DL model explainable artificial intelligence (EAI) were combined to identify impact certain attributes on ventilatory support patients. Nevertheless, does not suffer from curse dimensionality, but order significant attributes, EAI feature importance method was used. The produced results; however, it lacks interpretability. study performed using COVID-19-hospitalized King Abdulaziz Medical City, Riyadh. dataset contains patients’ demographic information, laboratory investigations, chest X-ray (CXR) findings. used suffers an imbalance; therefore, balanced accuracy, sensitivity, specificity, Youden index, AUC measures investigate effectiveness proposed model. Furthermore, experiments conducted original SMOTE (over under sampled) datasets. outperforms baseline with a accuracy 0.98 0.998 predicting full-feature set. Meanwhile, ventilator highest 0.979 0.981 achieved. will assist doctors early that are at or improve management hospital resources.