Brain Tumor Detection and Categorization with Segmentation of Improved Unsupervised Clustering Approach and Machine...

Area of Science:

• Medical Imaging and Diagnostics
• Artificial Intelligence in Healthcare
• Computational Pathology

Background:

• Brain tumors are a significant global health concern, with conventional diagnostic methods like biopsies having limitations such as low sensitivity and procedural risks.
• Traditional brain tumor identification relies on subjective interpretation, increasing the potential for human error and escalating healthcare costs due to labor-intensive processes.
• Advancements in medical imaging, including Magnetic Resonance Imaging (MRI) and Computed Tomography (CT), coupled with computer-aided diagnostic (CAD) systems, are crucial for improved visualization and detection.

Purpose of the Study:

• To propose an enhanced Fuzzy C-Means (FCM) segmentation algorithm specifically designed for Magnetic Resonance Imaging (MRI) of brain tumors.
• To improve the accuracy and efficiency of automatic tumor segmentation and classification by reducing computational complexity and selecting relevant features.
• To evaluate the performance of the proposed algorithm against existing models for reliable brain tumor identification.

Main Methods:

• Development of an improved Fuzzy C-Means (FCM) algorithm for segmenting brain tumors in MRI images.
• Feature selection focusing on the most relevant shape, texture, and color characteristics to minimize complexity.
• Classification of segmented tumors using an improved Extreme Learning Machine (ELM) model.

Main Results:

• The improved ELM classifier achieved high performance metrics: 98.56% accuracy, 99.14% precision, and 99.25% recall.
• The proposed algorithm demonstrated superior accuracy compared to existing models, with improvements ranging from 1.21% to 6.23%.
• On benchmark datasets (Fig share and Kaggle), the algorithm achieved excellent results (e.g., 99.42% accuracy on Kaggle), particularly excelling in detecting glioma grades.

Conclusions:

• The enhanced FCM segmentation algorithm, combined with the improved ELM classifier, offers a robust and accurate method for brain tumor classification from MRI data.
• The proposed approach significantly outperforms existing models, showing potential for more reliable and efficient clinical diagnosis.
• Despite challenges like artifacts and computational demands, the refined algorithm represents a notable advancement in precise brain tumor identification.