Automated cerebrum segmentation from three-dimensional sagittal brain MR images

Area of Science:

• Medical Imaging
• Neuroscience
• Computer Vision

Background:

• Accurate segmentation of the cerebrum is crucial for neurological studies and clinical applications.
• Existing methods for 3D brain MR image segmentation can be labor-intensive and prone to inaccuracies.

Purpose of the Study:

• To develop and validate a fully automated algorithm for cerebrum segmentation in 3D sagittal brain Magnetic Resonance (MR) images.
• To improve the efficiency and accuracy of cerebrum segmentation compared to conventional methods.

Main Methods:

• The algorithm utilizes a midsagittal view for initial cerebrum segmentation based on anatomical landmarks and connectivity algorithms.
• Cerebrum segmentation in adjacent lateral slices is guided iteratively using the segmentation mask from the previous slice.
• A restoration step is incorporated to correct for potential truncations in segmented brain regions by analyzing boundary differences and applying connectivity algorithms.

Main Results:

• The automated algorithm achieved satisfactory cerebrum segmentation across full 3D sagittal brain MR images.
• The proposed method demonstrated superior performance compared to conventional edge detection algorithms in segmenting the cerebrum.
• Iterative masking and region restoration effectively handled variations between adjacent slices.

Conclusions:

• The developed algorithm provides a robust and fully automated solution for 3D cerebrum segmentation from sagittal brain MR images.
• This automated approach enhances the reliability and efficiency of neuroimaging analysis.
• The method shows significant potential for clinical and research applications requiring precise brain structure segmentation.