Multimodal MRI Brain Tumor Image Segmentation Using Sparse Subspace Clustering Algorithm.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Biomedical Engineering

Background:

• Brain tumors are a leading cause of cancer mortality, necessitating advanced diagnostic tools.
• Accurate brain tumor segmentation from Magnetic Resonance Imaging (MRI) is crucial for effective diagnosis and treatment planning.
• Traditional segmentation methods struggle with multimodal MRI data due to variations in image acquisition and inherent data limitations.

Purpose of the Study:

• To introduce and evaluate a novel sparse subspace clustering (SSC) algorithm for brain tumor segmentation using multimodal MRI data.
• To assess the performance and noise robustness of the proposed SSC algorithm compared to existing methods.

Main Methods:

• Implementation of a sparse subspace clustering (SSC) algorithm tailored for multimodal MRI brain tumor images.
• Comparative analysis against traditional segmentation techniques and top-performing algorithms from the Brats 2015 competition.
• Evaluation of algorithm performance under varying levels of added Gaussian noise (5%, 10%, 15%, 20%).

Main Results:

• The proposed SSC algorithm shows competitive accuracy, ranking between 10 and 15 compared to leading methods in the Brats 2015 competition.
• Experimental results indicate superior noise immunity of the SSC algorithm when subjected to Gaussian noise.
• The algorithm effectively segments brain tumor structures from multimodal MRI, addressing limitations of single-mode imaging.

Conclusions:

• Sparse subspace clustering offers a promising approach for accurate and robust brain tumor segmentation from multimodal MRI.
• The developed algorithm provides a valuable tool for assisted diagnosis, outperforming comparable methods in noise resilience.
• Further research can explore the integration of SSC with other deep learning techniques for even greater diagnostic precision.