Intelligent System for Telemonitoring Healthcare in Pandemic Situations

Abstract

The COVID-19 pandemic has revealed critical weaknesses in global healthcare systems, particularly in their ability to ensure continuity of patient care under emergency conditions. This thesis addresses this challenge by proposing an intelligent, explainable, and generalizable framework for long-term remote patient monitoring in large-scale healthcare environments. Although COVID-19 serves as the primary case study, the proposed approach is designed to support a wide range of health monitoring scenarios requiring resilient and connected infrastructures. This research begins with an in-depth analysis of telemedicine models, intelligent healthcare systems, and health crisis management strategies, highlighting major limitations most notably the lack of continuous and personalized long-term monitoring for vulnerable populations. Based on these findings, a global architecture, AI-LMS, dedicated to long-term patient monitoring, is proposed. This architecture integrates several complementary components, including AI-RiskX, an explainable deep learning model combining CNN and LSTM architectures with SHAP-based analysis for risk stratification and early detection of critical onditions, as well as NS-Assist, a knowledge-based decision support system applied to pediatric nephrotic syndrome, demonstrating the practical applicability of the proposed framework in real-world clinical settings. Together, these components form a unified ecosystem that integrates data-driven intelligence, explainability, and clinical reasoning. The results demonstrate that the integration of explainable AI-driven remote monitoring systems can improve continuity of care, strengthen trust between patients and clinicians, and enhance the resilience of healthcare systems. This thesis thus contributes to the development of a new generation of intelligent healthcare systems that are explainable, ethical, and adaptable to diverse clinical contexts.