Automatic brain extraction and hemisphere segmentation in rat brain MR images after stroke using deformable models.

Area of Science:

• Neuroimaging
• Computational Neuroscience
• Medical Image Analysis

Background:

• Experimental stroke models are vital for understanding cerebral ischemia.
• Accurate segmentation of rat brains in MRI is critical for lesion identification and injury analysis in preclinical research.
• Existing methods for rat brain extraction and hemisphere segmentation in DWI and T2WI images require improvement.

Purpose of the Study:

• To investigate and develop practical algorithms for rat brain extraction and hemisphere segmentation applicable to both DWI and T2WI MRI images.
• To enhance the accuracy and reliability of image segmentation for subsequent stroke injury analysis in rodent models.

Main Methods:

• An efficient geometric deformable model with an added image force was used for automatic rat brain extraction (skull stripping).
• A brain feature detection algorithm and gradient vector flow were employed for hemisphere segmentation.
• The developed algorithms were tested on 55 DWI and T2WI subjects.

Main Results:

• The proposed rat brain extraction method achieved high performance with average Dice scores of 97.13% (DWI) and 97.42% (T2WI).
• Hemisphere segmentation accuracy was demonstrated by low average Hausdorff distances of 0.17 mm (DWI) and 0.15 mm (T2WI).
• The developed schemes outperformed existing state-of-the-art methods both qualitatively and quantitatively.

Conclusions:

• The developed frameworks provide accurate and reliable rat brain extraction and hemisphere segmentation for DWI and T2WI images.
• These algorithms are advantageous for facilitating preclinical stroke investigation and neuroscience research.
• The study highlights the importance of precise image segmentation in advancing rodent stroke research.