Welcome to the repository for "An Experimental Study on Brain Tumor Detection Using Deep Learning Techniques." This project explores cutting-edge machine learning and deep learning methodologies to enhance the detection and classification of brain tumors.
The increasing incidence of brain tumors highlights an urgent need for precise diagnostic methods. This study introduces an innovative convolutional model to improve brain tumor identification accuracy. Using Inception V3, GoogLeNet, and VGG-19 models, we meticulously evaluate each model’s performance in classifying brain tumor images, achieving remarkable results:
- Inception V3: 98.25% accuracy
- GoogLeNet: 95.36% accuracy
- VGG-19: 91.24% accuracy
Our novel approach utilizes decision-level fusion to combine outputs from multiple models, enhancing the robustness and accuracy of our detection system by leveraging the unique strengths of each model.
- Transfer Learning Techniques: Leverages pre-trained models to improve performance.
- Fusion-Based Approach: Decision-level fusion aggregates diverse outputs for reliable tumor classification.
- Advanced Image Processing: Incorporates Otsu Thresholding and CLAHE to improve image quality and classification precision.
This research encompasses a variety of tumor types, including:
- Meningiomas
- Gliomas
- Pituitary Tumors
- Malignant Tumors
- Medulloblastomas
- Lymphomas
| Model | Accuracy (%) |
|---|---|
| Inception V3 | 98.25 |
| GoogLeNet | 95.36 |
| VGG-19 | 91.24 |
The datasets used in this study are outlined below:
| Name | Description | Web Link |
|---|---|---|
| BRATS | Contains BRATS 2018, 2019, 2020, and 2021 datasets | BRATS Segmentation |
| TCGA LGG | MRI scans: Around 500 T1-weighted and T2-weighted MRI | TCGA Portal |
| Figshare | A smaller dataset with well-defined classes | Figshare Dataset |
| Harvard Medical | Includes a variety of specific data images | Harvard Dataset |
| BrainWeb | 32 T1-weighted, 32 T2-weighted, 32 FLAIR, and 32 DTI volumes | BrainWeb |
| OASIS | Structural MRI from 416 subjects | OASIS Dataset |
- Deep Learning Frameworks: TensorFlow, Keras
- Image Processing Techniques: CLAHE (Contrast Limited Adaptive Histogram Equalization), Otsu Thresholding, Gaussian Filter
- Fusion Technique: Decision-level fusion for enhanced accuracy and robustness
For the complete research paper, including detailed methodologies and results, visit the IEEE link below:
Access Full Paper on IEEE (Published Part of IIT-Mandi) 🔗
Chaitanya Sai Nutakki
Department of Computer Science and Engineering,
SRM University – AP, Andhra Pradesh, India
✉️ Email: csnutakki@gmail.com