Related Experiment Video
Updated: Jun 16, 2026
09:06
Whole-brain Segmentation and Change-point Analysis of Anatomical Brain MRI—Application in Premanifest Huntington's Disease
Published on: June 9, 2018
Construction of multi-region-multi-reference atlases for neonatal brain MRI segmentation
Feng Shi1, Pew-Thian Yap, Yong Fan
1IDEA Lab, Department of Radiology and BRIC, University of North Carolina at Chapel Hill, NC 27599, USA.
Neuroimage
|February 23, 2010
Summary
This study introduces a new method for segmenting neonatal brain MRI scans, improving accuracy by creating personalized atlases from sub-populations. This enhances guidance for better brain tissue segmentation in infants.
Area of Science:
- Medical Imaging
- Neuroscience
- Computer Vision
Background:
- Neonatal brain MRI segmentation is difficult due to poor image quality.
- Traditional atlas-based methods average images, reducing inter-subject variability and segmentation accuracy.
- Existing atlases lack the specificity needed for precise neonatal brain analysis.
Purpose of the Study:
- To develop an improved atlas-based segmentation framework for neonatal brain MRI.
- To enhance segmentation accuracy by accounting for local inter-subject structural variability.
- To create a more personalized and effective atlas for guiding neonatal brain tissue segmentation.
Main Methods:
- Proposed a multi-region, multi-reference framework for atlas generation.
- Clustered pre-segmented images into sub-populations for each brain region to create regional probability atlases.
- Adaptively combined regional atlases based on query image similarity using a joint registration-segmentation strategy.
Main Results:
- The proposed method achieved higher tissue overlap rates compared to manual segmentation (GM: 0.86, WM: 0.83, CSF: 0.80).
- Demonstrated lower standard deviations in segmentation accuracy (GM: 0.02, WM: 0.03, CSF: 0.05).
- Outperformed two other average-shape atlas-based segmentation methods in experimental validation.
Conclusions:
- The novel framework generates more accurate and personalized atlases for neonatal brain MRI segmentation.
- This approach effectively addresses the limitations of traditional averaging methods.
- The method significantly improves segmentation quality for neonatal brain tissues.
