Hybrid Deep Maxout-VGG-16 model for brain tumour detection and classification using MRI images.

Area of Science:

• Medical Imaging Analysis
• Artificial Intelligence in Healthcare
• Computational Neuroscience

Background:

• Early brain tumor detection is critical for effective treatment and improved patient survival rates.
• Conventional brain tumor detection methods face limitations such as accessibility issues, radiation exposure, high costs, and potential for false negatives.
• There is a need for advanced, accurate, and accessible methods for brain tumor identification.

Purpose of the Study:

• To develop and evaluate a novel Deep Maxout-Visual Geometry Group-16 (DM-VGG-16) model for accurate brain tumor detection using Magnetic Resonance Imaging (MRI).
• To overcome the limitations of existing brain tumor detection techniques by leveraging deep learning.
• To improve the diagnostic performance in terms of accuracy, True Negative Rate (TNR), and True Positive Rate (TPR).

Main Methods:

• The proposed method utilizes Magnetic Resonance Imaging (MRI) images as input.
• Image pre-processing is performed using a Non-Local Mean (NLM) filter, followed by segmentation using Template-based K-means and improved Fuzzy C Means (TKFCM) algorithm.
• Feature extraction involves techniques like area, cluster prominence, Hybrid PCA- Normalized GIST (NGIST), and Improved Median Binary Pattern (IMBP), feeding into the integrated Deep Maxout Network (DMN) and Visual Geometry Group-16 (VGG-16) model (DM-VGG-16) for detection.

Main Results:

• The developed DM-VGG-16 model demonstrated superior performance compared to conventional methods.
• The model achieved high performance metrics: 90.76% accuracy, 90.65% True Negative Rate (TNR), and 90.75% True Positive Rate (TPR).
• The integration of Deep Maxout Network (DMN) with VGG-16 architecture proved effective for brain tumor identification.

Conclusions:

• The proposed DM-VGG-16 model offers a promising and effective approach for automated brain tumor detection from MRI scans.
• The model's high accuracy and reliability suggest its potential to enhance early diagnosis and treatment planning.
• This deep learning-based method addresses some limitations of traditional techniques, paving the way for improved neuro-oncological diagnostics.