Transformative Approach to Pandemic Response: Leveraging Explainable-AI and Deep Learning for Enhanced Smart Monitoring

Artificial Intelligence (AI) is widely used in the healthcare sector, including healthcare administration, predictive analytics for medical outcomes, decision-making processes, and diagnoses. While AI has advanced to rival human proficiency in certain healthcare tasks, its implementation is limited due to perceived opacity and lack of trust. To address this limitation, this study introduces an innovative approach to interpretable deep learning for detecting Respiratory Syncytial Virus (RSV). The proposed technique aggregates audio and image data features to generate a cumulative probabilistic outcome and employs the SHapley Additive exPlanations (SHAP) technique for transparent explanations. Two public datasets, Coswara and Virufy, were used to train and evaluate the approach. The technique outperformed existing techniques by achieving an overall Accuracy, Precision, Recall, and F1-Score of 97.56%, 98.78%, 95.87%, and 97.86%, respectively. After generating probabilistic predictions, each prognosis is explained using the SHAP technique that considers both localized and global perspectives. In this manner, the proposed approach offers explanations for predictions, increasing trust in AI applications in the healthcare sector.