An advanced deep learning model for brain tumor classification using MRI scans, leveraging EfficientNetB1 and Grad-CAM for high accuracy and interpretability.
NeuroScanNet is a deep learning-powered medical imaging project that classifies brain tumors from MRI scans into four categories:
- Glioma
- Meningioma
- Pituitary Tumor
- No Tumor
Using EfficientNetB1 with transfer learning from ImageNet (6.58M parameters), the model achieves 98.85% test accuracy and a 99% F1-score, outperforming standard CNN architectures.
Grad-CAM provides interpretability by visualizing critical tumor regions, making the model more explainable for medical professionals.
✅ State-of-the-Art Performance: Achieves 98.85% accuracy with an optimized deep learning pipeline.
✅ Transfer Learning: Fine-tuned EfficientNetB1 with ImageNet-pretrained weights.
✅ Model Optimization: Implements data augmentation, ReduceLROnPlateau, and EarlyStopping, reducing training time by 30%.
✅ Explainability with Grad-CAM: Highlights key tumor regions for enhanced clinical interpretability.
📌 Dataset: Brain Tumor MRI Dataset from Kaggle
📌 Total Images: 6,170 MRI scans categorized into four tumor types.
📌 Data Preprocessing: Cropping, augmentation, and normalization applied for robust learning.
| Class | Precision | Recall | F1-Score | Support |
|---|---|---|---|---|
| 0 (Glioma) | 0.99 | 0.98 | 0.99 | 300 |
| 1 (Meningioma) | 0.97 | 0.99 | 0.98 | 306 |
| 2 (No Tumor) | 1.00 | 1.00 | 1.00 | 405 |
| 3 (Pituitary) | 1.00 | 0.99 | 0.99 | 300 |
Overall Metrics
| Metric | Score |
|---|---|
| Accuracy | 0.99 |
| Macro Avg | 0.99 |
| Weighted Avg | 0.99 |
- Model Architecture: EfficientNetB1 (pretrained on ImageNet).
- Loss Function: Categorical Crossentropy.
- Optimizer: Adam (LR scheduling with ReduceLROnPlateau).
- Regularization: Dropout (0.5) to prevent overfitting.
- Evaluation Metrics: Accuracy, F1-Score, Confusion Matrix.
- Python
- TensorFlow / Keras
- EfficientNetB1
- OpenCV
- Grad-CAM
- Jupyter Notebook
📧 Email: utkarshranaa06@gmail.com
🔗 GitHub: utkarshranaa
🔗 LinkedIn: www.linkedin.com/in/utkarshranaa
🔗 X/Twitter: @utkarshranaa
🚀 If you found this project useful, please ⭐ star the repository!