Related Experiment Videos
Fully automatic segmentation of the brain in MRI
1School of Computing Science, Simon Fraser University, Burnaby, BC, Canada. stella@cs.sfu.ca
IEEE Transactions on Medical Imaging
|June 9, 1998
Summary
A new automatic method accurately segments the brain in magnetic resonance (MR) images, even with radio frequency (RF) distortions. This robust technique reliably isolates brain tissue across various MRI scanners and sequences.
Area of Science:
- Medical Imaging
- Neuroimaging
- Image Processing
Background:
- Accurate brain segmentation is crucial for neurological studies.
- Existing methods struggle with radio frequency (RF) inhomogeneities and complex anatomical structures like the eyes.
- Automated segmentation simplifies and standardizes the analysis of magnetic resonance (MR) images.
Purpose of the Study:
- To develop a fully automatic and robust method for brain segmentation from head MR images.
- To address challenges posed by RF inhomogeneities and difficult-to-remove structures (e.g., eyes).
- To ensure reliable performance across different MRI scanners and imaging parameters.
Main Methods:
- An integrated approach combining anisotropic filters and "snakes" contouring techniques.
- Utilizing a priori knowledge to specifically remove ocular structures.
- A multistage process involving background noise removal, rough brain outline generation, and final mask refinement.
Main Results:
- Successful brain segmentation in every slice across diverse MR image datasets.
- Consistent performance irrespective of scanner type, image resolution, or echo sequences.
- Effective removal of background noise and ocular regions, leading to accurate brain masks.
Conclusions:
- The developed method offers a robust and fully automatic solution for brain segmentation in MR imaging.
- It overcomes limitations of previous techniques, particularly in the presence of RF inhomogeneities.
- This automated approach has significant potential for various brain studies and clinical applications.