Related Experiment Video
Updated: Apr 12, 2026
06:48
Automated Segmentation of Cortical Grey Matter from T1-Weighted MRI Images
Published on: January 7, 2019
9.6K
Fully automated whole-head segmentation with improved smoothness and continuity, with theory reviewed
1Department of Biomedical Engineering, City College of the City University of New York, New York, NY, USA.
Plos One
|May 21, 2015
Summary
Accurate whole-head segmentation is crucial for brain stimulation and source reconstruction. This new tool uses Bayesian inference and MRFs for improved anatomical modeling, enhancing precision in neuroimaging applications.
Area of Science:
- Neuroimaging and Computational Neuroscience
- Medical Image Analysis
- Biophysics
Background:
- Accurate individualized head models are essential for precise transcranial electrical or magnetic stimulation (TES/TMS) and current-source reconstruction in electroencephalography (EEG) and magnetoencephalography (MEG).
- Existing automated segmentation tools for magnetic resonance imaging (MRI) typically focus only on brain tissues, have limited fields of view, and lack the necessary tissue continuity and smoothness for reliable current-flow estimations.
Purpose of the Study:
- To develop and validate a novel automated tool for accurate, whole-head anatomical segmentation.
- To improve the morphological accuracy and continuity of segmented tissues, specifically cerebrospinal fluid (CSF), skull, and soft tissues, for enhanced biophysical modeling.
Main Methods:
- A Bayesian inference framework combining an image intensity model, an anatomical prior (atlas), and morphological constraints via Markov Random Fields (MRFs).
- Evaluation using 20 simulated and 8 real whole-head MRI datasets at 1 mm³ resolution.
- Comparison with segmentation algorithms integrated into Statistical Parametric Mapping (SPM).
Main Results:
- The developed tool demonstrated improved surface smoothness and continuity compared to existing SPM segmentation algorithms.
- The method successfully segmented whole-head anatomy, including brain, CSF, skull, and soft tissues, with high morphological fidelity.
- The tool provides a feasible solution for routine, accurate, and morphologically correct whole-head modeling for individual subjects.
Conclusions:
- The presented tool significantly advances the capability for creating individualized head models essential for TES/TMS and EEG/MEG applications.
- The integration of Bayesian inference with MRFs offers a robust approach to whole-head MRI segmentation, addressing limitations of current methods.
- Public availability of the code and data within SPM facilitates widespread adoption and further research in computational neuroimaging.
