---

🫁 HealthAI – AI-Powered Chest X-Ray Disease Detection System

A deep-learning powered platform for multi-disease medical imaging analysis, built with TensorFlow, FastAPI, Streamlit, and deployed on Render.

---

🏷️ Tech Stack Badges

---

🚀 Overview

HealthAI is an end-to-end medical imaging project that detects multiple chest diseases from X-ray images using a deep learning model trained on the CheXpert dataset. It features:

• 📸 Multi-Disease Chest X-Ray Classification (supports 5+ pathologies)
• 🧠 CNN/Transfer Learning Model (EfficientNetB0)
• ⚙️ FastAPI Backend for AI Inference
• 🧩 Streamlit Dashboard for Real-time Diagnosis
• ☁️ Cloud Deployment on Render
• 🔒 Clean architecture, modular API, scalable structure

This project demonstrates real-world deep learning workflow: data pipeline → model training → evaluation → API → UI → deployment.

---

🎥 Demo Video

healthai.mp4

---

🧠 Features

✔ 1. Multi-Label Disease Detection

Predicts probability for diseases such as:

• Atelectasis
• Cardiomegaly
• Consolidation
• Edema
• Pleural Effusion

Example output:

{
  "Atelectasis": 0.14,
  "Cardiomegaly": 0.02,
  "Consolidation": 0.51,
  "Edema": 0.09,
  "Pleural Effusion": 0.78
}

---

✔ 2. FastAPI Backend

Endpoints:

POST /predict-xray

• Accepts an uploaded X-ray image
• Returns normalized probabilities for all diseases

GET /health

Health check endpoint

---

✔ 3. Streamlit Dashboard

• Upload X-ray image
• Call backend API
• Display disease probabilities in clean UI
• Visual + interactive

---

✔ 4. Cloud Deployment

Backend deployed on Render, accessible publicly.

---

📁 Project Structure

HealthAI-Project/
│
├── backend/
│   ├── main.py
│   ├── model/
│   │   ├── healthai_model.keras
│   │   └── diseases.json
│   └── requirements.txt
│
├── dashboard/
│   ├── app.py
│   └── assets/
│
├── notebooks/
│   ├── 01_train_healthai_model.ipynb
│
├── .gitignore
├── README.md
└── LICENSE

---

🌍 Live API (Render Deployment)

Base URL:

https://healthai-backend-oj25.onrender.com/

Swagger Docs:

https://healthai-backend-oj25.onrender.com/docs

Streamlit Interface:

https://healthai-frontend-aldddrxsrswa59csqqrkij.streamlit.app/

---

📡 API Reference (Quick Guide)

🔹 POST /predict-xray

Request: multipart/form-data field name = file

Response:

{
  "Atelectasis": 0.14,
  "Cardiomegaly": 0.02,
  "Consolidation": 0.51,
  "Edema": 0.09,
  "Pleural Effusion": 0.78
}

🔹 GET /health

Simple API check.

---

🏗 Architecture

graph TD
    A[CheXpert Dataset] --> B[Model Training - EfficientNet]
    B --> C[Saved Model - Keras]
    C --> D[FastAPI Backend]
    D --> E[Streamlit Dashboard]
    D --> F[Mobile App - Future]
    D --> G[Cloud Deployment - Render]

Loading

---

📥 How to Run Locally

1️⃣ Create environment

conda create -n healthai python=3.10
conda activate healthai

2️⃣ Install backend dependencies

cd backend
pip install -r requirements.txt

3️⃣ Run FastAPI backend

uvicorn main:app --reload --host 0.0.0.0 --port 8000

Docs: http://127.0.0.1:8000/docs

---

4️⃣ Run Dashboard

Open a second terminal:

cd dashboard
streamlit run app.py

Streamlit UI: http://localhost:8501/

---

📸 Screenshots

Add these once screenshots are uploaded:

DASBOARD

PNEUMONIA PREDICTION

MULTIPLE DISEAS PREDICTION

---

📊 Model Training Details

• Dataset: CheXpert small (5 diseases)

• Input size: 224x224

• Base model: EfficientNetB0

• Optimizer: Adam (lr=1e-4)

• Loss: Binary Crossentropy

• Metrics: AUC, Accuracy

• Achieved Validation Metrics:

  • Accuracy: ~88%
  • Loss: ~0.27

---

🧩 Future Improvements

• Add more diseases (14-class CheXpert full)
• Add attention heatmaps (Grad-CAM)
• Add mobile app (Flutter)
• Add hospital-level data logging
• Improve UI/UX of dashboard

---

💼 Why This Project?

This project demonstrates:

• Deep learning model building
• Medical imaging pipeline
• Multi-label classification
• Clean backend architecture
• Real-world API integration
• Deployment experience
• Professional-grade UI

A perfect portfolio project for Machine Learning, AI Engineering, and Healthcare AI roles.

---

📄 License

This project is licensed under the MIT License.

---

👤 Author

Akhil Saurabh

---