Skip to content

README.md

Latest commit

 

History

History
215 lines (140 loc) · 5.98 KB

README.md

File metadata and controls

215 lines (140 loc) · 5.98 KB

🧠 NeuroAssist-AI

A Deep Learning-based Smart System for Brain Tumor Detection and Glioma Stage Prediction

🔗 Live Demo

📌 Overview

NeuroAssist-AI is a two-stage deep learning pipeline that automates:

  1. Brain Tumor Detection from grayscale MRI/CT images using a custom CNN.
  2. Glioma Stage Prediction using numerical gene mutation data via a custom ANN.

This system is inspired by real clinical practices and aims to provide intelligent support to radiologists and neurologists.

🧪 Live Testing

You can test the full system online:

🌐 🔗 Live Web App – NeuroAssist-AI

📚 Research Basis

📄 Inspired by: “Brain Tumor Classification and Glioma Stage Prediction Using Deep Learning”

🔗 Read the original paper

Note: Original paper had no public dataset or code - we implemented it from scratch.

📂 Dataset

Dataset: Brain Tumor MRI Dataset – Kaggle

  • 4 Classes: Glioma, Meningioma, Pituitary, No Tumor
  • Format: Grayscale .jpg images categorized in folders

🧠 Model Architecture

🔷 CNN – Brain Tumor Detection

Layer Type Description
Input Grayscale MRI/CT image
Conv Blocks 3 × Conv2D + ReLU + MaxPooling
FC Layers Flatten → Dense → Softmax
Output 4 classes (No Tumor, Meningioma, Pituitary, Glioma)

✅ Trained from scratch in PyTorch ❌ No dropout (no overfitting observed)

🟢 ANN – Glioma Stage Classification

Layer Type Description
Input Gene mutation test results
Dense 2–3 Fully Connected Layers
Activation ReLU + Softmax/Regression Output
Output Glioma Stage (I–IV)

💾 Model Files

File Name Purpose Availability
BTD_model.pth Brain Tumor Detection (CNN) 🔗 Download from Google Drive
glioma_stages.pth Glioma Stage Classification (ANN) ✅ Included in models/ directory

📍 Note: Due to GitHub’s 100MB limit, BTD_model.pth is stored externally. After downloading, manually place it inside the models/ folder like this:

models/BTD_model.pth

🔄 Optional: Auto-Download Script

Use this code to download the CNN model automatically if missing:

import os, urllib.request

model_url = "https://drive.google.com/uc?export=download&id=19SVLCD3DTa1aBZ9PI4TTgNkvJKgL2LSY"
model_path = "models/BTD_model.pth"

if not os.path.exists(model_path):
    os.makedirs("models", exist_ok=True)
    print("Downloading model...")
    urllib.request.urlretrieve(model_url, model_path)
    print("Model downloaded.")

⚙️ Tech Stack

Category Tools / Libraries
Language Python 3.10+
DL Framework PyTorch
Others OpenCV, NumPy, scikit-learn, Matplotlib
Training Env Jupyter Notebook, NVIDIA GPU
Deployment FastAPI + Vercel (Frontend)
Storage Google Drive (for model hosting)

📁 Folder Structure

NeuroAssistAI/
├── main.py                  # Entry point
├── API.py                   # FastAPI backend
├── utils.py                 # Helper functions
├── models/
│   ├── BTD_model.pth
│   └── glioma_stages.pth
├── images/
├── README.md
└── .gitignore

🚀 How to Run Locally

1️⃣ Clone the Repo

git clone https://github.com/fewgets/NeuroAssistAI.git
cd NeuroAssistAI

2️⃣ Install Dependencies

pip install torch torchvision opencv-python matplotlib scikit-learn fastapi uvicorn

Or use:

pip install -r requirements.txt

3️⃣ Run the App

python main.py

✨ Features

✅ Tumor classification using CNN

✅ Glioma stage prediction using ANN

✅ Web interface for real-time inference

✅ Pre-trained models included

✅ End-to-end modular system

✅ Easy to extend and deploy

🔭 Future Enhancements

  • 🤖 Chatbot integration for medical Q&A (Gemini/GPT)
  • 🧬 Integration with real-time genetic APIs
  • 📊 Real-time dashboard for doctors
  • 📱 Mobile version using React Native

📩 Contact

👨‍💻 Usama Shahid 📧 Email: dev.usamashahid@gmail.com 🔗 GitHub: @fewgets

For training notebooks, collab requests, or guidance — feel free to connect.

⚠️ License

This project is intended for academic and educational purposes only. Re-use is allowed with proper credit and citation. Not intended for clinical deployment without validation.

🔗 🌐 Live App • 📁 Dataset • 📜 Paper