You can deploy this model and start earning money today!

joshchan1301/
x-ray_img_analysis_ai

Name: x-ray_img_analysis_ai
Availability: InStock

Views

No views yet

🫁 Pulmo Vision – Chest X-ray AI Analysis

Pulmo Vision is an AI project that applies the Swin Transformer to analyze chest X-ray images, supporting the detection of several conditions:

BACTERIAL Pneumonia
VIRAL Pneumonia
NORMAL (healthy)

Additionally, the system generates Grad-CAM heatmaps to explain the model’s decision, helping users understand which lung regions contribute to the diagnosis.

🚀 Features

📷 Upload X-ray images → the model analyzes and returns predicted label and probability.
🔥 Grad-CAM visualization → highlights the lung regions the model focuses on.
🧠 Swin Transformer backbone → state-of-the-art computer vision architecture.
🌐 FastAPI backend → easy to deploy (Vercel, Render, Railway…).
🤖 GPT-powered chatbot (OpenAI API) → explains results in a user-friendly way.

📂 Project Structure

pulmo-backend/
│── main.py # FastAPI server (API routes)
│── model_inference.py # Model loading, preprocessing, Grad-CAM
│── requirements.txt # Dependencies
│── .env.example # API keys & config
│── swin_best_model.pth # Model checkpoint

⚙️ Installation

Clone repo:

bash

git clone https://github.com/joshchan1301/pulmo-backend.git
cd pulmo-backend

Install dependencies:

pip install -r requirements.txt

Create .env file:

OPENAI_API_KEY=sk-xxxxxxx

Download pretrained model (Swin Transformer):
- Google Drive link:
  https://drive.google.com/uc?export=download&id=1VD2pXT9aDHmGwn2KiD3aSoXj5nBglVLw
- Save as swin_best_model.pth in the project folder.

▶️ Run API server

uvicorn main:app --reload

API docs (Swagger UI):
👉 http://127.0.0.1:8000/docs

📡 API Endpoints

1. Analyze Chest X-ray

POST /api/analyze-xray

Input: file (form-data, image .png/.jpg)
Output JSON:

json

{
  "label": "BACTERIAL",
  "probability": 87.5,
  "heatmap": "<base64-encoded PNG>"
}

2. Pulmo AI Chatbot

POST /api/chat

Input JSON:

{ "message": "What does my X-ray mean?" }

Output JSON:

json

{
  "reply": "The X-ray indicates signs of bacterial pneumonia. The red highlighted areas are where the model focused most."
}

📊 Demo

Do your own =)))

📖 Tech Stack

Python, FastAPI, httpx – API server
PyTorch, timm, torchvision – Deep learning model
OpenAI API – Chatbot
Grad-CAM – Explainable AI

⚠️ Disclaimer

This is not a certified medical tool.
For educational and research purposes only.
Not a substitute for professional medical diagnosis.

✍️ Author: Tran Hieu

joshchan1301/x-ray_img_analysis_ai