Open Access

ARTICLE

Transformation of MRI Images to Three-Level Color Spaces for Brain Tumor Classification Using Deep-Net

Fadl Dahan^*
Department of Management Information Systems, College of Business Administration-Hawtat Bani Tamim, Prince Sattam bin Abdulaziz University, Al-Kharj, 11942, Saudi Arabia
* Corresponding Author: Fadl Dahan. Email:

Intelligent Automation & Soft Computing https://doi.org/10.32604/iasc.2024.047921

Received 22 November 2023; Accepted 27 February 2024; Published online 27 March 2024

Abstract

In the domain of medical imaging, the accurate detection and classification of brain tumors is very important. This study introduces an advanced method for identifying camouflaged brain tumors within images. Our proposed model consists of three steps: Feature extraction, feature fusion, and then classification. The core of this model revolves around a feature extraction framework that combines color-transformed images with deep learning techniques, using the ResNet50 Convolutional Neural Network (CNN) architecture. So the focus is to extract robust feature from MRI images, particularly emphasizing weighted average features extracted from the first convolutional layer renowned for their discriminative power. To enhance model robustness, we introduced a novel feature fusion technique based on the Marine Predator Algorithm (MPA), inspired by the hunting behavior of marine predators and has shown promise in optimizing complex problems. The proposed methodology can accurately classify and detect brain tumors in camouflage images by combining the power of color transformations, deep learning, and feature fusion via MPA, and achieved an accuracy of 98.72% on a more complex dataset surpassing the existing state-of-the-art methods, highlighting the effectiveness of the proposed model. The importance of this research is in its potential to advance the field of medical image analysis, particularly in brain tumor diagnosis, where diagnoses early, and accurate classification are critical for improved patient results.

---

Keywords

Camouflage brain tumor; image classification; weighted convolutional features; CNN ResNet50

---