Brain MR perfusion image segmentation using independent component analysis and hierarchical clustering

Area of Science:

• Neuroimaging
• Medical Image Analysis
• Cerebrovascular Physiology

Background:

• Dynamic-susceptibility-contrast (DSC) MRI is crucial for assessing brain perfusion.
• Analyzing temporal signal variations in brain tissues allows for segmentation of blood supply patterns, vital for understanding cerebral hemodynamics.

Purpose of the Study:

• To develop and validate an automated method for multi-tissue hemodynamic classification using DSC MRI data.
• To improve the differentiation between physiological and pathological hemodynamics in the brain.

Main Methods:

• Independent component analysis (ICA) was employed to decompose DSC MRI perfusion data into independent components.
• Hierarchical clustering (HC) was utilized to automatically segment distinct tissue clusters based on these independent components.
• The ICA-HC method was applied to data from five normal subjects and one patient with internal carotid artery stenosis.

Main Results:

• The ICA-HC approach successfully segmented brain tissues based on hemodynamic properties.
• The method demonstrated effectiveness in classifying multi-tissue hemodynamics.
• Improved differentiation between pathological and physiological hemodynamic patterns was observed.

Conclusions:

• The combined ICA-HC technique provides an effective tool for automated multi-tissue hemodynamic classification in DSC MRI.
• This approach enhances the ability to distinguish between normal and abnormal cerebral blood flow, aiding in the diagnosis and study of cerebrovascular diseases.