Gluco Vision – Non-Invasive Glucose Estimation Through Video

1. THE DREAM: Accessible Health for Everyone

One of the major growing diseases in the world,especially in India is Diabetes.Thousands and Millions of people live with diabetes or are at risk, but regular monitoring remains a challenge. The most traditional methods require finger-prick tests, which are painful, inconvenient, and costly.

For poor people who can afford the bare minimum, the financial burden is significant: a glucometer can cost between ₹800 and ₹1500, and test strips cost  between ₹15 and ₹30 each. This can reach between ₹1,000 and ₹2,500 per month for patients needing multiple daily tests. Due to the cost,most low income people frequently skip testing,which then increases the health complications such as heart diseases etc. which can even lead to death.

tThe idea with Gluco Vision is to make glucose monitoring affordable for everyone: a solution that is non-invasive, low-cost, and widely accessible, requiring only a smartphone or laptop camera.

2. The Problem: Painful, Expensiv

e, and Fragmented Monitoring

There are many problems created due to the traditional glucose monitoring:

• Invasiveness: Frequent pricking is painful and discourages adherence.
  
  
• High Cost: particularly for low income and poor people,continuous use of strips is expensive,
  
  
• Limited Insight: Finger prick tests only provide a snap
  shot of glucose levels,which are missing fluctuations throughout the day.
  
  
• Stress and Anxiety: it can lead to decrease of quality of life due to daily monitoring diseases

These limitations contribute to poor disease management, leading to hospitalization and diseases which can be avoidable.This mainly affects the poor.

3. Our Solution: Gluco Vision

Gluco Vision uses the power of photoplethysmography (PPG) which are video based  and machine learning to estimate blood glucose non-invasively, using just a camera and a smartphone.

Key Technical Highlights:

• • PPG Signal Extraction: Captures subtle color changes in fingertips from video, corresponding to blood volume changes.
    

• Signal Processing: to reduce noise from motion or lighting we apply Butterworth bandpass filtering and Savitzky-Golay smoothing
  
  
• Feature Engineering: To relate PPG signals with glucose dynamics,it computes heart rate, entropy, signal area, and other physiological features
  
  
• Data Augmentation: Generates synthetic glucose data to enhance training, combining real and simulated data for a better result .
  
  
• Machine Learning Model:To predict glucose levels with associated risk categories (normal, elevated, high),we Use Gradient Boosting Regressor 

Example Output:

 Model trained with enhanced real and synthetic data. Test MSE: 92.7
  Estimated Glucose: 142.3 mg/dL  Elevated glucose – Monitor or consult physician

Gluco Vision significantly reduces costs and makes regular monitoring feasible for low-income patients because we are eliminating the need for physical test strips.

4. Socio-Economic Impact

• • Budget friendly Healthcare: Eliminates the recurring expenses of strips and lancets.
    
    
  • Improved Compliance: encourages regular checks due to non-invasive monitoring.
    
    
  • Early Detection:reduces complications. Continuous or frequent monitoring which  allows for early intervention
    
    

• Accessible Anywhere: makes it feasible in remote or underserved areas as only a camera and internet-enabled device are required,
  
  

5. Future Roadmap

• Face and Fingertip Tracking: For easier, automated signal capture.
  
  
• Multiple Video Formats Support: Ensures compatibility across devices.
  
  
• Integration with Real-World Datasets: For better accuracy and better results we use real world datasets
  
  
• Advanced AI Models:  To improve predictive performance and accuracy we use Neural networks and deep learning
• Continuous Monitoring Potential: Enable long-term trends for proactive healthcare.