Home Page – Provides an overview of AI-powered anemia detection and guides users to start the diagnosis
Input Page – Allows users to upload eye, lip, and skin images along with a PPG video for anemia analysis
Processing Page – Analyzes uploaded images and PPG data using AI models to detect anemia in real-time
Results Page – Displays anemia diagnosis, hemoglobin estimation, and detailed analysis of the uploaded data
Results Page – Displays anemia diagnosis, hemoglobin estimation, and analysis, along with an AI-powered chatbot for medical insights

🩸 AI-Powered Non-Invasive Anemia Detection System

🌟 Overview

The AI-Powered Non-Invasive Anemia Detection System is an innovative healthcare solution designed to analyze 👁️ eye conjunctiva, 👄 lip, and 🖐️ skin images, along with 📊 PPG-based hemoglobin estimation, to detect anemia in a fast, non-invasive, and efficient manner. The system leverages 🤖 deep learning models and a 🌲 Random Forest classifier to classify anemia risk levels and recommend nearby 👨‍⚕️ doctors based on the user's 📍 location.

✨ Inspiration

Anemia affects over 1.6 billion people worldwide 🌍, yet many cases go undiagnosed due to the lack of accessible and affordable diagnostic tools. Traditional blood tests require 🏥 lab infrastructure, trained personnel, and time, making them inaccessible for many rural and underserved populations.

Our inspiration came from the urgent need for an alternative, non-invasive, and AI-driven solution that could provide instant anemia detection using only a smartphone 📱. By leveraging deep learning, PPG-based hemoglobin estimation, and AI-powered decision-making, we sought to revolutionize anemia diagnosis and make healthcare more inclusive. ❤️

🔎 What It Does

📷 Captures and analyzes images of the eye, lips, and skin to detect pallor.
📹 Uses PPG-based hemoglobin estimation from a smartphone camera video.
🤖 AI-powered chatbot provides real-time medical insights.
📍 Doctor recommendation system suggests nearby specialists.
⚡ Runs in real-time on both web and mobile platforms for instant results.

🏗️ How We Built It

We structured our system into multiple AI models 🏥, each focusing on a specific biomarker:

🖥️ Deep Learning Models (TF Lite) process eye, lip, and skin images to detect anemia-related pallor.
🌲 A Random Forest Classifier estimates hemoglobin levels using PPG signal extraction from a smartphone camera video. 📱
The 💬 LangChain-powered AI chatbot offers real-time medical insights and advice based on the test results.

All models were optimized for ⚡ real-time execution, ensuring smooth performance on both web and mobile platforms. 🌐

🚧 Challenges We Ran Into

1️⃣ Ensuring Model Accuracy & Reliability

Training deep learning models to distinguish subtle anemia symptoms from normal variations in skin and eye color was challenging. 🎭
We used 📈 diverse datasets and augmentation techniques to improve model generalization.

2️⃣ Extracting Meaningful PPG Data

Processing 📹 finger-on-camera videos for PPG signal extraction required filtering out noise and inconsistencies.
We implemented ⚙️ frame-wise averaging and adaptive signal processing techniques to enhance PPG data quality.

3️⃣ Optimizing for Real-Time Execution

Running deep learning models on-the-fly required converting them into TensorFlow Lite (TF Lite)⚡ for efficiency.
We minimized latency and memory usage while maintaining high accuracy. 📊

4️⃣ Integrating Multiple AI Components

Combining computer vision 🖼️, PPG analysis 📊, chatbot intelligence 🤖, and doctor recommendations 🏥 was complex.
We carefully structured our 🔄 data pipeline to ensure seamless interaction between components.

🏆 Accomplishments That We're Proud Of

✅ Successfully built an AI model capable of detecting anemia with high accuracy.
✅ Developed an intuitive, real-time system that runs efficiently on mobile and web.
✅ Integrated PPG-based hemoglobin estimation, pushing the boundaries of non-invasive diagnostics.
✅ Created an AI chatbot to offer real-time health insights.
✅ Designed a doctor recommendation system for immediate medical assistance.

📚 What We Learned

🧠 AI in Healthcare Requires Precision – Model validation is critical to avoid false positives/negatives.
📊 PPG-based Diagnostics Are Powerful – This technique has potential beyond anemia detection, including cardiovascular health monitoring. ❤️
🎯 User Accessibility Is Key – Making the tool simple and intuitive ensures wider adoption, especially in non-technical communities.
🤝 Collaboration Enhances Innovation – The integration of ML engineers, software developers, and medical experts led to a well-rounded solution.

🔮 What's Next for AnemAI

🔹 🧬 Expand AI models to detect other blood disorders like iron deficiency & jaundice.
🔹 ⌚ Integrate wearable device support for continuous hemoglobin tracking.
🔹 📲 Develop a full mobile app to make real-time anemia screening more accessible.
🔹 🏥 Collaborate with healthcare institutions for clinical validation & large-scale adoption.

🛠 Tech Stack

Component	Technology Used
Frontend UI	Streamlit (Python)
Image Processing	OpenCV, PIL
Machine Learning Models	TensorFlow Lite (TF Lite), MobileNetV2, CNN
PPG Hemoglobin Estimation	Random Forest (Scikit-Learn)
AI Chatbot	LangChain + Llama 3.3-70B
Doctor Recommendations	Google Maps API

🎯 Final Thoughts

Our AI-powered anemia detection system embodies technology for social good 💙, offering a scalable, non-invasive, and accessible solution to a major global health issue. This project is not just an idea—it’s a step toward transforming early disease detection worldwide 🌍🚀!