Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{200461,
        author = {Siddhant Jadhav and Aliasgar Lakdawala and Cyjil Varghese and Dr. Kalpana Deorukhkar},
        title = {AarogyaMitra: Health Coach AI — A Wearable-Free Multimodal Framework for Personalized Chronic Disease Management Using Hybrid CNN–LSTM, Ensemble Anomaly Detection, and RAG-Augmented Conversational Coaching.},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {12},
        number = {12},
        pages = {2505-2515},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=200461},
        abstract = {Chronic non-communicable diseases (NCDs) principally Type 2 diabetes mellitus, hypertension, and cardiovascular disorders affect more than 1.3 billion individuals globally and account for 74% of all deaths worldwide. In India, over 101 million individuals live with diabetes and approximately 220 million suffer from hypertension, yet the national doctor-to-patient ratio stands at a critically inadequate 1:1,456. Existing digital health solutions either depend on prohibitively expensive wearable hardware or deliver narrow, non-personalised monitoring devoid of emergency response capability. This paper presents Aarogya Mitra: Health Coach AI, a wearable-free, AI-driven mobile health platform providing continuous, personalised chronic disease management using only a standard smartphone. The system integrates six modalities motion signals, GPS, manual vitals, HealthKit data, medical OCR, and behavioural context through a unified sense–analyse–act–learn pipeline. A hybrid CNN–LSTM architecture trained on 30–60 second sliding windows models biosignal patterns; a Random Forest ensemble of 100 decision trees provides personalised anomaly detection; and a RAG-enhanced LLaMA-3-8B-Instruct conversational agent grounded in WHO and ICMR clinical guidelines delivers context-aware health coaching with multilingual support for 5+ Indian languages. An OCR Smart Cam module digitises prescriptions and lab reports to 93% accuracy. A real-time SOS fall-detection subsystem dispatches emergency alerts within 10 seconds. Evaluated across the PhysioNet MIT-BIH Arrhythmia, WESAD, and DEAP benchmark datasets alongside an 8-week controlled pilot study (n=47 in the IJIRT paper; n=50+ in beta trials), Aarogya Mitra achieved 95.2% bio signal classification accuracy (AUC-ROC: 0.991), 96.1% ensemble accuracy, 93% OCR extraction accuracy, and 98% SOS alert success rate. A 2-month beta trial (n=50+) demonstrated a 68% improvement in medication adherence, 91% weekly engagement, and a 30% reduction in emergency hospital visits, yielding a user satisfaction score of 4.3/5.},
        keywords = {Chronic Disease Management, Wearable-Free Health Monitoring, CNN–LSTM Bio signal Classification, Retrieval-Augmented Generation (RAG), Large Language Model (LLM) Healthcare, Mobile Health (mHealth), Smartphone Sensing, Anomaly Detection, Random Forest, OCR Medical Document Digitisation, Fall Detection, SOS Emergency Alert, Multilingual NLP, Gemini AI, LLaMA-3, Preventive Healthcare, Medication Adherence, Multimodal Data Fusion.},
        month = {May},
        }