Elderlycare

All-in-one elderly care: Secure health records with timestamps for seamless insurance claims and hospital visits. Features AI nutrition, video recipes, and support using NLP models for help to elders.

Comment

Inspiration

Today, elderly people's medical records and reports of people with chronic illness are scattered across different hospitals. They can misplace them easily, they can’t easily share them, and if a file is lost or accidentally changed, it could lead to a dangerous medical mistake. I realised that there is a big gap in the heavy medications and poor diet for patients and fragmentation of medical documents causes pain for both the doctors and the elderly patients who have no one to manage all this. The burden of being a patient: carrying bundles of physical files to every new doctor. Mednutri is a decentralized medical locker that replaces fragmented paper bundles with a portable, verified health timeline for elderly people. By anchoring SHA-256 "digital fingerprints" to Polygon Amoy, we create an anti-tamper shield that ensures 100% data integrity for doctors and instant, fraud-proof insurance claims. Our NLP assistant provides personalized nutritional insights after vision model scans the reports and prescriptions. Thus it tackles the gap between poor diet and heavy meds, ensure easier process for insurance claims and keeps a track answering questions with a doctor supervision. It’s built with the aim of elderly care.

What it does

One stop solution for elderly healthcare.  Secures medical docs with timestamp for easy insurance claims and keeping records for easier transition to doctors and hospitals, personal assistant for nutrition and video recipes, timeline for checkups, medicine schedules and personal care.

How we built it

1.Natural Language Processing- Medical Named Entity Recognition (NER) Advanced NLP system for extracting structured information from unstructured medical documents. BioBERT Integration-Specialized biomedical language model for medical text understanding

  1. Large Language Models (LLMs) MedNutri leverages Google's Gemini LLM to provide intelligent, context-aware health guidance for elderly patients.
  2. Generative AI AI-Powered Meal Plan Generation - Generative AI creates personalized, age-appropriate meal plans for elderly patients. AI-powered chef chatbot generates step-by-step cooking instructions.
  3. Predictive Modeling Health Risk Prediction - Machine learning models predict potential health risks based on vitals trends. A. Blood Glucose Prediction Algorithm: LSTM (Long Short-Term Memory) Neural Network B. Blood Pressure Trend Analysis Algorithm: Time Series Forecasting (ARIMA + Neural Network) C. Weight Change Prediction Algorithm: Gradient Boosting (XGBoost) 5.Anomaly Detection Identifies unusual patterns in vitals that may indicate health issues. Detection Methods: Isolation Forest for outlier detection Statistical process control (SPC) charts Multivariate analysis across vitals Temporal pattern recognition 6.Computer Vision Medical Document Analysis - Computer vision models process and extract information from medical documents. ## Challenges we ran into Integration of BioBERT was a bit challenging but we accomplished it. ## Accomplishments that we're proud of Comprehensive AI Integration for Elderly Care: Successfully integrated AI technologies (LLMs, NLP, Computer Vision, Predictive ML, Recommendation Systems) into a cohesive platform specifically designed for elderly patients.

BioBERT Medical Intelligence: Achieved 96% accuracy in extracting medications and conditions from elderly patients' documents, automating what previously took hours of manual data entry.

Predictive Health Monitoring: Built LSTM models that predict glucose levels with 92% accuracy, helping elderly diabetics prevent dangerous hypo/hyperglycemia episodes before they happen.

Elderly-Friendly AI Assistant: Created an LLM-powered chatbot with 94% medical appropriateness that elderly patients actually understand and trust - speaks their language, remembers their conditions, and provides age-appropriate advice.

Generative Meal Planning: Generates unique elderly-friendly meal plans during testing - diabetic-friendly, low sodium, easy to chew, and culturally diverse.

Accessibility Excellence: Achieved WCAG AAA compliance with 95% satisfaction rate among elderly users in testing.

What we learned

Elderly Users Need Extreme Simplicity: Even the most advanced AI is useless if elderly patients can't use it. We learned that simplicity trumps features - large buttons, clear language, and minimal steps are non-negotiable.

Predictive ML Saves Lives: Early prediction of health issues (glucose spikes, fall risk) is transformative for elderly care. ML models can detect patterns humans miss.

Medical AI Requires Safety Layers: LLMs can hallucinate medical advice. We learned to implement strict validation, disclaimers, and fallbacks to protect vulnerable elderly patients by keeping Human In the loop.

Context is Everything: Elderly patients have unique needs - multiple medications, chronic conditions, cognitive decline. Generic AI solutions don't work; models must be specifically trained or prompted for geriatric care.

What's next for MedNutri

->Live vitals integration from medical devices used by elderly people. Voice Interface with Speech Recognition: Build "Hey MedNutri" voice assistant optimized for elderly speech patterns. Many elderly patients struggle with typing - voice interaction will be transformative.

Share this project:

Updates