Related Experiment Video
Updated: Jul 9, 2026
06:48
Automated Segmentation of Cortical Grey Matter from T1-Weighted MRI Images
Published on: January 7, 2019
Probabilistic brain tissue segmentation in neonatal magnetic resonance imaging
Petronella Anbeek1, Koen L Vincken, Floris Groenendaal
1Department of Radiology, University Medical Center Utrecht, 3584 CX Utrecht, The Netherlands. P.Anbeek@umcutrecht.nl
Pediatric Research
|December 20, 2007
Summary
A new automated method accurately segments neonatal brain structures like white matter and gray matter using MRI scans. This technique provides reliable measurements for large-scale population studies.
Area of Science:
- Neuroimaging
- Medical Image Analysis
- Developmental Neuroscience
Background:
- Accurate segmentation of neonatal brain structures is crucial for understanding development and disease.
- Existing methods may lack automation or precision for large-scale studies.
- Routine diagnostic MRI data is often underutilized for detailed neonatal brain analysis.
Purpose of the Study:
- To develop and validate a fully automated method for segmenting four key neonatal brain structures: white matter (WM), central gray matter (CEGM), cortical gray matter (COGM), and cerebrospinal fluid (CSF).
- To assess the accuracy and reliability of the automated segmentation using quantitative metrics compared to a gold standard.
- To evaluate the utility of probabilistic segmentations for accurate neonatal brain tissue volume estimation.
Main Methods:
- A K nearest neighbor (KNN) classification technique was employed, utilizing features from spatial information and voxel intensities derived from T2-weighted (T2-w) and inversion recovery (IR) MRI scans.
- Probabilistic segmentation maps were generated for each tissue type, from which binary segmentations were obtained by applying thresholds.
- Quantitative validation was performed by comparing automated segmentations against a gold standard, calculating sensitivity, specificity, and Dice similarity index (SI).
Main Results:
- The automated method achieved high accuracy, with sensitivity >0.82 and specificity >0.9 for all segmented neonatal brain tissue types.
- Probabilistic segmentation volumes closely matched the gold standard volumes, demonstrating the method's reliability for tissue quantification.
- The KNN-based approach proved effective for segmenting WM, CEGM, COGM, and CSF in neonatal brains.
Conclusions:
- The developed automated KNN method offers a robust and accurate approach for neonatal brain segmentation using routine diagnostic MRI.
- Probabilistic segmentation outcomes provide a valuable tool for precise tissue volume measurements in neonatal populations.
- This method is suitable for large-scale population studies, facilitating research into neonatal brain development and disorders.
