Brain-Tumor-Classification-using-Deep-Learning

Description:

Brain Tumor Classification from Mutlisequence MRI (T1, T1C and T2) and Mutlimodal CT and MRI using EfficientNetV2B0 with Mutliheaded Self Attention and Hyperparameter Fine-Tuning

Methodology

• Implementation of a novel framework for classifying various kinds of brain tumors and healthy patients from structural MRI scans of T1, T1C and T2 sequences as well CT scans.
• In the first stage, a pre-trained EfficientNetV2 architecture has been used followed by Mutli-Head Self Attention Mechanism on the extracted, high-dimensional sequential feature maps.
• Global Average Pooling, Batch Normalization, L1, L2 Regularization and Dropout along with fine-tuned hyperparameters have been applied before mutli-class classification through softmax activation function.

Datasets used:

• Brain Tumor MRI Images 44 Classes
• Brain Tumor MRI Images 17 Classes
• Brain tumor multimodal image (CT & MRI)

Workflow Used:

Classification Task [15 Class, 6 Class, 2 class]

Installation

To install the required packages, run:

pip install -r requirements.txt

Program Files:

• Dependencies
• Data Preprocessing
• Model Architecture
• Training
• Evaluation
• K-Fold Cross Validation
• Grad-CAM Analysis

Importance of Project:

• The promising results achieved underscore the potential of our framework’s robust nature and generalization capabilities across various modalities.
• Assist medical professionals in making precise diagnoses and, ultimately enhance patient outcomes.

Credit(s) and Acknowledgement:

Supervisor: Dr. Pawan Kumar Singh

Paper:

It'd be great if you could cite our paper (under review) if this code has been helpful to you.

Thank you very much!