https://github.com/rahatmoktadir03/tumor-scope
Brain Tumor Classification project leveraging neural networks to classify MRI scans with high accuracy. Features include a Streamlit-based app for predictions, Gemini 1.5 Flash for interpretability, and advanced visualizations. It also includes model comparison, multimodal LLM integration, and real-time interactions.
https://github.com/rahatmoktadir03/tumor-scope
brain-tumor-classification deep-learning gemini keras machine-learning multimodal-llm neural-networks python streamlit tensorflow xception-model
Last synced: about 1 month ago
JSON representation
Brain Tumor Classification project leveraging neural networks to classify MRI scans with high accuracy. Features include a Streamlit-based app for predictions, Gemini 1.5 Flash for interpretability, and advanced visualizations. It also includes model comparison, multimodal LLM integration, and real-time interactions.
- Host: GitHub
- URL: https://github.com/rahatmoktadir03/tumor-scope
- Owner: rahatmoktadir03
- Created: 2024-11-21T02:15:12.000Z (over 1 year ago)
- Default Branch: main
- Last Pushed: 2025-07-18T14:59:02.000Z (11 months ago)
- Last Synced: 2025-07-18T19:15:40.278Z (11 months ago)
- Topics: brain-tumor-classification, deep-learning, gemini, keras, machine-learning, multimodal-llm, neural-networks, python, streamlit, tensorflow, xception-model
- Language: Jupyter Notebook
- Homepage: https://tumor-scope.streamlit.app
- Size: 2.59 MB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# 🧠 TumorScope – Brain Tumor Classification with Deep Learning & Gemini AI
---
## 📖 Overview
**TumorScope** is a web application that classifies **brain MRI scans** into four categories:
- **Glioma**
- **Meningioma**
- **Pituitary Tumor**
- **No Tumor**
The app uses:
- **Xception (Transfer Learning)**
- **Custom CNN Model**
- **Saliency Maps** for interpretability
- **Gemini 1.5 Flash (Google AI)** to generate concise **medical explanations** for predictions
The project is deployed on **Streamlit Cloud** and provides an interactive user experience for radiologists, researchers, and students.
---
## ✨ Key Features
- **Upload Brain MRI Images** for classification
- **Dual-Model Selection**: Transfer Learning (Xception) or Custom CNN
- **Saliency Map Generation** to visualize important image regions
- **AI-Powered Explanations** via Gemini 1.5 Flash
- **Interactive Probability Charts** with Plotly
- **Secure Deployment** on Streamlit Cloud
---
## 🛠 Tech Stack
| Component | Technology |
|-------------------|-------------------------|
| **Frontend** | Streamlit |
| **Model Training**| TensorFlow / Keras |
| **Explainability**| Gemini 1.5 Flash |
| **Visualization** | OpenCV, Plotly |
| **Deployment** | Streamlit Cloud |
---
## 🚀 Getting Started
### 1️⃣ Clone the Repo
```bash
git clone https://github.com/rahatmoktadir03/tumor-scope.git
cd tumor-scope
```
### 2️⃣ Install Dependencies
```bash
pip install -r requirements.txt
```
### 3️⃣ Add Gemini API Key
- Create a .env file in the project root:
```init
GOOGLE_API_KEY=your-gemini-api-key
```
### 4️⃣ Run Locally
```bash
streamlit run streamlit_app.py
```
---
## 📊 Models and Accuracy
- Xception (Transfer Learning) – ~99% accuracy
- Custom CNN – ~98% accuracy
- Saliency Maps highlight critical regions for model interpretability
---
## 🔮 Future Improvements
- Chat with MRI scan (using multimodal LLMs)
- Compare multiple models side-by-side
- Generate detailed diagnostic reports for doctors