In the realm of healthcare, early detection is everything. This project leverages the power of Deep Learning to classify brain tumors into four distinct types: glioma, meningioma, no tumor, and pituitary, based on the Kaggle Brain Tumor dataset.
Brain tumors can have life-altering consequences if not detected early. Early diagnosis significantly improves treatment options and patient outcomes. By integrating this model into healthcare systems, hospitals can:
- Revolutionize healthcare delivery.
- Provide proactive and precise treatment.
- Save countless lives by detecting tumors before significant harm occurs.
- Accuracy: Achieved 97% accuracy on training, 92% on validation, and 88% on the test set.
- Efficiency: The model uses only 4 million parameters, making it lightweight and efficient for real-world deployment.
This project uses the Kaggle Brain Tumor dataset, which consists of MRI images classified into four categories. The dataset was accessed using the opendatasets library for seamless integration.
- Architecture: A lightweight convolutional neural network (CNN) designed for efficiency and performance.
- Optimization: Balances parameter count and accuracy for practical applications.
- Performance: The model delivers remarkable accuracy while maintaining a compact size.
-
Clone this repository:
git clone https://github.com/Tusharkn12/brain-tumor-classification.git cd brain-tumor-classification -
Install dependencies:
pip install -r requirements.txt
-
Download the dataset:
python download_dataset.py
-
Run the notebook:
jupyter notebook brain-tumor-classification.ipynb
| Metric | Training | Validation | Testing |
|---|---|---|---|
| Accuracy | 97% | 92% | 88% |
- Data Preprocessing: The dataset is preprocessed with normalization and augmentation techniques to improve model generalization.
- Model Training: The CNN model is trained using TensorFlow, focusing on minimizing loss while keeping the architecture lightweight.
- Evaluation: The model is evaluated on training, validation, and test datasets, ensuring robustness.
One of the most exhilarating challenges was engineering the lightest weight model possible without compromising accuracy. This was achieved through careful tuning of hyperparameters and architecture optimization.
Contributions are welcome! Feel free to submit issues or pull requests for improvements. Let’s make this project even better together. 💪
- Integration: Deploy the model into a hospital management system for real-time diagnosis.
- Enhancements: Improve accuracy further with advanced architectures or additional data.
- Edge Deployment: Optimize the model for mobile and edge devices.
This project is licensed under the MIT License. See the LICENSE file for details.
Let’s revolutionize healthcare together. 💡🏥