Brain tissue segmentation based on MP2RAGE multi-contrast images in 7 T MRI

Area of Science:

• Neuroimaging
• Medical Image Analysis
• Radiology

Background:

• Accurate segmentation of brain tissues is crucial for neuroimaging research and clinical diagnosis.
• Existing segmentation methods, such as FSL, FreeSurfer, and SPM12, have limitations in processing time or accuracy.
• High-field MRI, specifically 7 Tesla (7T), offers enhanced image resolution but requires optimized segmentation techniques.

Purpose of the Study:

• To develop and evaluate a novel method for segmenting brain tissues (gray matter, white matter, cerebrospinal fluid) using multi-contrast MP2RAGE images at 7T.
• To compare the proposed method's processing time and segmentation accuracy against established tools (FSL, FreeSurfer, SPM12).

Main Methods:

• Utilized multi-contrast images (T1 map, uniform T1-weighted image) from a 7T MP2RAGE sequence.
• Developed an intensity-based segmentation approach, similar to FSL, without relying on template images or atlases.
• Evaluated processing time and similarity of segmented masks against FSL, FreeSurfer, and SPM12.

Main Results:

• The proposed method achieved significantly faster processing times (28s) compared to FSL (444s) and SPM12 (159s).
• Segmentation similarity was higher with FSL compared to SPM12.
• Misclassification occurred in subcortical structures and large vessels, but segmentation was good in the cerebellum and medial white matter.

Conclusions:

• The MP2RAGE-based segmentation method offers a rapid and effective alternative for brain tissue segmentation at 7T.
• Its speed and comparable accuracy make it suitable for time-sensitive neuroimaging research and clinical applications.
• Further refinement may improve accuracy for complex structures like subcortical regions.