An Improved Soft Subspace Clustering Algorithm Based on Particle Swarm Optimization for MR Image Segmentation

Area of Science:

• Data Mining
• Image Processing
• Machine Learning

Background:

• Soft subspace clustering (SSC) analyzes high-dimensional data, assigning weights to cluster classes for membership degree assessment.
• Enhanced SSC algorithms incorporate spatial information to improve intra-class compactness and inter-class separation.
• Existing SSC methods are sensitive to noisy data, leading to poor segmentation accuracy and local optima.

Purpose of the Study:

• To develop a robust SSC approach mitigating noise interference for improved image segmentation.
• To enhance the accuracy and reliability of soft subspace clustering in the presence of noisy data.
• To introduce a novel methodology for segmenting noisy images using an optimized clustering technique.

Main Methods:

• A novel soft subspace clustering (SSC) approach is proposed, integrating particle swarm optimization (PSO).
• PSO is employed to identify optimal clustering centers, enhancing data analysis.
• Spatial information is leveraged through increased geographical membership for precise inter-cluster quantification.
• An extended noise clustering method maximizes weights, with constraints shifted from equality to boundary to minimize noise impact.

Main Results:

• The proposed SSC-PSO method demonstrates reduced sensitivity to noisy data.
• Experimental results show superior segmentation accuracy on images with existing or introduced noise.
• The algorithm effectively segments noisy images, validating its efficacy.

Conclusions:

• The revised SSC approach based on PSO offers a novel and effective method for noisy image segmentation.
• This methodology significantly improves segmentation accuracy compared to traditional SSC algorithms.
• The study provides a robust solution for handling noise in high-dimensional data clustering and image analysis.