Brain Tumor Analysis AI

An AI-powered web application for brain tumor analysis from MRI images. The system combines deep learning-based tumor classification and tumor segmentation to assist in identifying brain tumors while generating detailed prediction reports.

---

Overview

Brain Tumor Analysis AI is a deep learning application that allows users to upload MRI brain scans and receive automated predictions regarding tumor presence. The project integrates multiple computer vision models into a web application that performs inference, segmentation, visualization, and report generation.

The project was developed as part of an academic collaboration to demonstrate practical applications of Artificial Intelligence in medical image analysis.

---

Features

• Brain tumor classification
• Brain tumor segmentation
• MRI image upload interface
• Automatic preprocessing pipeline
• Prediction visualization
• Report generation
• Flask-based web application
• Modular inference pipeline

---

Tech Stack

• Python
• Flask
• TensorFlow / Keras
• OpenCV
• NumPy
• HTML
• CSS

---

Repository Structure

Brain-Tumor-Analysis-AI/

├── brain_tumor/
│   └── data.yaml
│
├── data/
│   └── selected_indices.json
│
├── models/
│   ├── classifier_model.py
│   ├── model_utils.py
│   └── unet_model.py
│
├── scripts/
│   ├── static/
│   ├── templates/
│   ├── app.py
│   ├── inference.py
│   ├── inference_web.py
│   ├── report_generator.py
│   └── ...
│
├── main.py
├── requirements.txt
├── README.md
└── LICENSE

---

Dataset

The training and validation datasets are not included in this repository.

This decision was made because:

• The dataset exceeds GitHub's recommended repository size.
• The dataset is publicly available from its original source.
• The repository is intended to focus on the implementation rather than redistributing datasets.

The dataset configuration is provided in:

brain_tumor/data.yaml

If you wish to retrain the models, recreate the following structure:

brain_tumor/
├── train/
└── valid/

and update the paths if necessary.

---

Trained Models

The pretrained .h5 model files are not included because they exceed GitHub's maximum file size limit.

Place the downloaded model files inside:

models/

before running inference.

---

Installation

Clone the repository

git clone https://github.com/<your-username>/Brain-Tumor-Analysis-AI.git

Install dependencies

pip install -r requirements.txt

---

Running the Application

Start the application:

python main.py

or

python scripts/app.py

(depending on your project configuration)

Open the browser and upload an MRI image for analysis.

---

Sample Output

Example generated reports are included in the repository to demonstrate the application's reporting capabilities.

---

Current Limitations

• Training dataset not included
• Pretrained model weights not included
• Designed primarily for educational and research purposes
• Performance depends on image quality

---

Future Improvements

• Multi-class tumor classification
• Cloud deployment
• DICOM image support
• Explainable AI (Grad-CAM)
• Patient history integration
• Improved reporting dashboard

---

Contributors

Developed collaboratively by:

• Kevin Koruthu George
• Arjun Ritvik Azhagesan

---

License

This project is licensed under the Apache 2.0 License.