Brain tissue segmentation using q-entropy in multiple sclerosis magnetic resonance images

Area of Science:

• Medical Imaging
• Biophysics
• Computational Neuroscience

Background:

• Brain volume loss is a key indicator of neurodegenerative diseases like multiple sclerosis.
• Accurate segmentation of brain tissues (white matter, gray matter, cerebrospinal fluid) is crucial for monitoring disease progression.

Purpose of the Study:

• To introduce and evaluate a new method for brain tissue segmentation using generalized Tsallis entropy.
• To determine optimal statistical segmentation parameters (q values) for different brain tissue classes.
• To assess the method's efficacy in detecting annual brain volume loss in patients with neurodegenerative diseases.

Main Methods:

• Developed a tissue segmentation method based on pixel intensity thresholding and generalized Tsallis entropy.
• Compared the method's performance across various q parameters to find optimal values for white matter, gray matter, and cerebrospinal fluid.
• Applied the optimized method to analyze brain magnetic resonance images from 43 patients and 10 healthy controls.

Main Results:

• Identified optimal entropic q indices: 0.1 for white matter, 1.5 for gray matter, and 0.2 for cerebrospinal fluid.
• The algorithm detected an average annual brain volume loss of 0.98% in patients, consistent with existing literature.
• Demonstrated that generalized Tsallis entropy can differentiate structural correlations and scale-invariant similarities within tissue classes.

Conclusions:

• Generalized Tsallis entropy offers an effective approach for automatic brain tissue segmentation.
• This method provides valuable statistical parameters for characterizing distinct tissue classes.
• The technique shows promise for improving the assessment of neurodegenerative disease progression through accurate brain volume loss detection.