🌟 NutriSight

NutriSight is a web-based health assistant that combines machine learning, image processing, and NLP-powered food analysis to support people with Diabetic Retinopathy (DR).

It detects the stage of DR from retinal fundus images and recommends diabetes-friendly meals from restaurant menus — empowering users to make informed, healthy dining choices.

---

🚀 Features

• 🧠 Diabetic Retinopathy Detection

  • Preprocessing: CLAHE, grayscale conversion
  • Feature extraction: GLCM, LBP, entropy, vessel density
  • Classification: Trained XGBoost model
  • Predicts 5 DR stages with confidence scores

• 🍲 Personalized Food Advisor

  • Analyzes restaurant menus via text input or web-scraping
  • Extracts nutritional data and evaluates dishes against DR-specific dietary rules
  • Recommends the most diabetes-friendly options

• 🔗 Integrated Health Management

  • Combines medical diagnosis with nutrition guidance
  • Provides personalized, stage-aware food recommendations

---

🛠️ Tech Stack

• Machine Learning: XGBoost
• Image Processing: OpenCV, scikit-image
• NLP & Data Handling: NLTK / difflib, pandas
• Frontend: Streamlit
• Backend: Python

---

📊 DR Detection Workflow

1. Input: Retinal fundus image
2. Preprocessing: CLAHE enhancement + grayscale
3. Feature Extraction:
   • Gray-Level Co-occurrence Matrix (GLCM)
   • Local Binary Patterns (LBP)
   • Image entropy
   • Vessel density
4. Classification: XGBoost model → predicts No DR, Mild, Moderate, Severe, or Proliferative
5. Output: Predicted stage + confidence score

---

🍽️ Nutrition Recommendation Workflow

1. Input: Restaurant menu (image/text/web-scraped)
2. Processing: NLP-based dish name matching with nutrition database
3. Analysis: Evaluates calories, carbs, sugars, sodium, etc.
4. Output: Safe vs unsafe dishes with personalized recommendations based on DR stage