eng Oriental Scientific Publishing Company Biomedical and Pharmacology Journal 0974-6242 2026-07-17 19 3 72725 article Umapathy Kannan 1 Bibhuti Bhusan Rath 2 Selvakumarasamy Kathirvelu 3 Sasi Govindrajulu 4 Govindaraju Sankaranarayan 5 Mageswari Narayanasamy 6 Department of ECE, Rajalakshmi Engineering College, Thandalam, Chennai. Department of Electrical Engineering, Aditya Institute of Technology and Management,Tekkali, AP, India. Department of ECE, Saveetha School of Engineering, Saveetha Institute of Medical and Technical Sciences, SIMATS, Chennai. Department of ECE, Chettinad Institute of Technology, Chettinad Academy of Research and Education, Tamil Nadu, India. Department of Computer Science and Humanities, St. Joseph University, Tindivanam. India. Department of ECE, Ashoka Women's Engineering College(Autonomous), Kurnool, India. Early detection of chronic and acute health conditions plays a crucial role in reducing hospitalization and mortality rates, especially among aging populations and individuals with comorbidities. The proposed study presents a Wearable IoT-Based Health Monitoring System that uses a Convolution Neural Network-Long Short-Term Memory (CNN-LSTM) hybrid deep learning model that allows real-time physiological data analysis for early prediction of disease onset. It consists of wearable biosensors such as Electrocardiogram (ECG), Photoplethysmogram(PPG), accelerator, and peripheral oxygen saturation (SpO2), an edge processing unit that uses Arduino Nano 33 BLE Sense and Zigbee to transmit wirelessly, and a two-tiered cloud-edge architecture to provide scalable analytics. Preprocessing and feature extraction recollected signals are subjected to Z-score normalization, noise removal, and Independent Component Analysis, and reduced in dimensions with Autoencoders. The obtained features are fed through the CNN-LSTM model that was trained on theMedical Information Mart for Intensive Care(MIMIC-III) dataset because it shows high-dimensional, real-world clinical variability. The model attained a classification accuracy of 97.12, precision of 96.84, recall of 97.33, and the F1-score of 97.08. An additional risk scoring system using fuzzy logic makes the health status evaluation even more personalized, and it provides real-time alert whenever predefined risk thresholds are exceeded. System performance was confirmed on 10 diseases with more than 92% accuracy in each and real-time responsiveness with less than 40ms latency and low power use appropriate capacities of wearable devices. The supported encryption and safe cloud protocols guarantee the adherence to Health Insurance Portability and Accountability Act (HIPAA) of the United States and General Data Protection Regulation (GDPR). The suggested system has a strong potential to enhance the aspects of preventive healthcare, optimize the time on emergency response, and provide the capability of continuously monitor the status in remote and under-resourced locations. https://biomedpharmajournal.org/vol19no3/wearable-iot-based-health-monitoring-for-early-disease-prediction/ CNN-LSTM Early Disease Prediction Edge Computing Fuzzy Logic GDPR Health Monitoring HIPAA MIMIC-III Risk Scoring Wearable IoT