IoT-Based Deep Learning Approach for Alzheimer's Disease Prediction, Patient Monitoring, and Safety

In response to the increasing prevalence of Alzheimer's disease and the burden of healthcare management, this study presents a novel approach to predict Alzheimer's disease using deep learning as well as patient monitoring with the help of the IoT to represent an innovative way to address health issues and disease management. To select the best-performing model for disease classification, we evaluated four resourceconstrained deep learning models for classifying Alzheimer's disease from brain MRI scans. Among them, EfficientNetB2, a highly efficient model optimized for edge devices, achieved a classification accuracy of $\mathbf{9 8. 6 9 \%}$, demonstrating its robustness in disease diagnosis. In addition to deep learning-based classification, we integrated patient monitoring by collecting realtime environmental and physiological data, such as temperature, humidity, pulse rate, and oxygen saturation, to determine the patient's condition and ensure its reliability. The collected data were securely stored in the database to ensure accessibility through FastAPI for immediate analysis and response. Also, live location tracking ensures patient safety, particularly during wandering episodes. By integrating deep learning-driven classification with IoT-based patient monitoring, this study enhances predictive capabilities, facilitates proactive healthcare management, and aims to improve patients' quality of life while reducing healthcare burden through early disease detection and tailored interventions.