Efficient Brain Tumor Classification using Filter-Based Deep Feature Selection Methodology

Description:

Methodology

• Implementation of a novel two-stage framework for classifying 3 kinds of brain tumors and healthy patients from structural MRI scans.
• In the first stage, a pre-trained Convolutional Neural Network has been used to extract relevant features from pre-processed images through transfer learning, considerably reducing training time and extensive hardware requirements.
• In the second stage, a filter-based deep feature selection methodology using Mutual Information has been applied to minimize the extracted, high-dimensional feature maps.
• Finally, a Support Vector Machine with a polynomial kerenl ruse has been used for multi-class classification.

Datasets used:

• Brain Tumor MRI Dataset
• Crystal Clean: Brain Tumors MRI Dataset
• Figshare: Brain Tumors MRI Dataset

Preprocessing of MRI scans:

Methodology Used:

Importance of Project:

• The promising results achieved underscore the potential of our lightweight framework’s robust nature and generalization capabilities.
• Suitable for deployment in real-time environments with limited technological resources.
• Assist medical professionals in making precise diagnoses and, ultimately enhance patient outcomes.

Basic Requirements:

To ensure consistent results, please use the following libraries:

Package Name	Version	Version Badge
pandas	1.5.3
scikit-learn	1.0.2
seaborn	0.11.2
matplotlib	3.5.1
opencv-python	4.5.3.20210927
numpy	1.21.2
tensorflow	2.8.0
pillow	9.1.0

Installation

To install the required packages, run:

pip install -r requirements.txt

Program files:

Source Codes:

• Dependencies
• Configuration File
• Data Preprocessing
• Data Loader
• Deep Learning Model
• Deep Learning Model Training
• Training Utilities
• Deep Feature Extraction
• Filter-based Deep Feature Selection + SVM Classifier
• Evaluation

Scripts:

• Mutual Information Analysis
• ML Pipeline
• Results Visualization

The implementation files of this project can be found here:

• Brain Tumor MRI Jupyter Notebook
• Crystal Clean Dataset Jupyter Notebook
• Figshare Dataset Jupyter Notebook

Credit(s) and Acknowledgement:

Collaborator: Utathya Aich Supervisor: Dr. Pawan Kumar Singh

Paper: Link

It'd be great if you could cite our manuscript if this code has been helpful to you: Kar, S., Aich, U. & Singh, P.K. Efficient Brain Tumor Classification Using Filter-Based Deep Feature Selection Methodology. SN COMPUT. SCI. 5, 1033 (2024).

Thank you very much!