https://github.com/arshdeepyadavofficial/brain-tumor-detection-and-classification

Enhanced MRI Brain Tumor Detection using a Hybrid Deep Learning + Machine Learning model. Combines MobileNetV2 & SVM to classify tumors (Glioma, Meningioma, Pituitary, No Tumor) from contrast MRI. Achieves ~93% accuracy via transfer learning & augmentation.
https://github.com/arshdeepyadavofficial/brain-tumor-detection-and-classification

brain-tumor cnn computer-science deep-learning grad-cam healthcare hybrid-model image-classification keras medical-diagnosis medical-imaging mobilenet-v2 mri-classification multi-class-classification svm tensorflow transfer-learning tumor-classification tumor-detection

Last synced: 3 months ago
JSON representation

Enhanced MRI Brain Tumor Detection using a Hybrid Deep Learning + Machine Learning model. Combines MobileNetV2 & SVM to classify tumors (Glioma, Meningioma, Pituitary, No Tumor) from contrast MRI. Achieves ~93% accuracy via transfer learning & augmentation.

• Host: GitHub
• URL: https://github.com/arshdeepyadavofficial/brain-tumor-detection-and-classification
• Owner: arshdeepyadavofficial
• License: mit
• Created: 2025-06-25T15:51:53.000Z (3 months ago)
• Default Branch: main
• Last Pushed: 2025-06-27T11:37:34.000Z (3 months ago)
• Last Synced: 2025-06-27T12:36:11.876Z (3 months ago)
• Topics: brain-tumor, cnn, computer-science, deep-learning, grad-cam, healthcare, hybrid-model, image-classification, keras, medical-diagnosis, medical-imaging, mobilenet-v2, mri-classification, multi-class-classification, svm, tensorflow, transfer-learning, tumor-classification, tumor-detection
• Language: Jupyter Notebook
• Homepage:
• Size: 32.7 MB
• Stars: 0
• Watchers: 0
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

          
# 🧠 ENHANCED MRI BRAIN TUMOR DETECTION AND CLASSIFICATION USING HYBRID APPROACH

[![Last Commit](https://img.shields.io/github/last-commit/arshdeepyadavofficial/Brain-Tumor-Detection-And-Classification.svg)](https://github.com/arshdeepyadavofficial/Brain-Tumor-Detection-And-Classification/commits/)

[![Jupyter Notebook](https://img.shields.io/badge/Jupyter-Notebook-orange?logo=jupyter)](https://jupyter.org/)

[![Python](https://img.shields.io/badge/Python-3.x-blue?logo=python)](https://www.python.org/)




[![Markdown](https://img.shields.io/badge/Markdown-000000?logo=markdown&logoColor=white)](https://commonmark.org/help/)




This project presents a deep learning-based hybrid approach for the **automatic detection and classification of brain tumors** using **contrast-enhanced MRI scans**. The system classifies MRI images into four categories: **Glioma**, **Meningioma**, **Pituitary**, and **No Tumor**. The model achieves outstanding performance with a classification accuracy of **~99%**, validated through **cross-validation** and **AUC-ROC** metrics.

📩 **Want the research paper?**  

Feel free to reach out at: **2208390100017@reck.ac.in**

---

## 📂 Dataset Information

The MRI dataset used in this project was sourced from **Kaggle**:

🔗 [Brain Tumor MRI Dataset (Kaggle)](https://www.kaggle.com/datasets/navoneel/brain-mri-images-for-brain-tumor-detection)

### 📊 Dataset Summary

| Class      | Number of Images | Training Set | Testing Set |

| ---------- | ---------------- | ------------ | ----------- |

| Glioma     | 1621             | 1321         | 300         |

| Meningioma | 1656             | 1339         | 306         |

| Pituitary  | 1757             | 1457         | 300         |

| No Tumor   | 2000             | 1595         | 405         |

| **Total**  | **7023**         | **5712**     | **1311**    |

---

## 🧠 How It Works

This project implements an automated brain tumor detection and classification system using a **hybrid deep learning–machine learning** approach. The key stages are:

### 🔄 Preprocessing

- **Noise Reduction:** Gaussian filtering  

- **Normalization:** Intensity scaling  

- **Enhancement:** Contrast, brightness  

- **Inversion & Resizing:** To 299×299  

- **Augmentation:** Rotation, shear, zoom

### 📥 Feature Extraction

- **Model:** MobileNetV2 (pre-trained on ImageNet)  

- **Purpose:** Extract high-level visual features via transfer learning

### 🧮 Classification

- **Classifier:** SVM with linear kernel  

- **Classes:** Glioma, Meningioma, Pituitary, No Tumor

### 🔍 Interpretability

- **Grad-CAM:** Heatmaps for tumor region localization

### 📈 Evaluation

- **Metrics:** Accuracy, Precision, Recall, F1-score, Confusion Matrix  

- **Results:** ~93%  accuracy

---

### 🧩 Model Visualization

A visual summary of the pipeline and model flow:

![Model Visualization](visualselection.png)

---

## 📈 Performance Metrics

* ✅ **Accuracy**: ~93%  

* 🧮 **Loss**: Converges quickly, minimal overfitting  

* 📌 **AUC-ROC**: High discriminatory power across all tumor types

### 📊 Classification Results:

![Confusion Matrix and Classification Report](confusion_classification.jpg)

---

## 🧪 Cross-Validation

The model was evaluated using **5-fold cross-validation** to ensure its stability and reliability across different data splits. Results showed consistent performance with minimal deviation across folds.

---

## 🔍 Example Predictions

| MRI Sample                         | Prediction      |

| ---------------------------------- | --------------- |

| ![Glioma](predic_glioma.png)       | Glioma Tumor    |

| ![No Tumor](predic_notumor.png)    | No Tumor        |

| ![Pituitary](predic_pituitory.png) | Pituitary Tumor |

Additional samples with enhanced features:

![Enhanced Sample View](get_samples_with_enhancement.png)

---

## 🧠 Dependencies

To run this project, install the following Python libraries:

```bash

pip install tensorflow keras matplotlib scikit-learn numpy opencv-python

````

```bash

pip install -r requirements.txt

````

---

## 🚀 Getting Started

To reproduce the results:

1. Clone this repository:

```bash

git clone https://github.com/arshdeep-yadav/Brain-Tumor-Detection-Hybrid-Approach.git

cd Brain-Tumor-Detection-Hybrid-Approach

```

2. Download the dataset from Kaggle:

   👉 [Download MRI Dataset](https://www.kaggle.com/datasets/navoneel/brain-mri-images-for-brain-tumor-detection)

3. Activate your environment and run the notebook:

```bash

jupyter notebook brain-tumor-mri-accuracy-99-auc-roc-cv.ipynb

```

---

## 📜 Citations

Please consider citing the following sources if you use this work in your research:

> **[Arshdeep Yadav](https://github.com/arshdeepyadavofficial)**, 


>  C.S.E; B.Tech 3rd Year, 


>  REC Kannauj

> Navoneel Chakrabarty. *Brain MRI Dataset*, [Kaggle](https://www.kaggle.com/datasets/navoneel/brain-mri-images-for-brain-tumor-detection)

> Chollet, François. *Deep Learning with Python*. Manning Publications, 2018.

---

## 📌 Future Work

* [ ] Deploy the model via Streamlit or Flask

* [ ] Integrate Grad-CAM for visual explanations

* [ ] Real-time MRI scan classification interface

---

## 👨‍💻 Author

Developed with 🖤 by **Arshdeep Yadav** 


🎓 CSE B.Tech 3rd Year 


🏫 Rajkiya Engineering College, Kannauj

🌐 Connect with me:

* [GitHub](https://github.com/arshdeepyadavofficial)

* [LinkedIn](https://www.linkedin.com/in/arshdeep-yadav-827aa1257?utm_source=share&utm_campaign=share_via&utm_content=profile&utm_medium=android_app)

---

## 📄 License

This project is licensed under the [MIT License](LICENSE).  

Feel free to use, modify, and share it with proper attribution.