Automated Segmentation of White Matter Hyperintensities in MRI Scans
Author Information
Author(s): Sean D. Smart, Michael J. Firbank, John T. O'Brien
Primary Institution: Institute for Ageing and Health, Newcastle University
Hypothesis
Can automated methods accurately segment white matter hyperintensities from MRI scans compared to manual segmentation?
Conclusion
The in-house automated segmentation method demonstrated the best performance in accurately identifying white matter hyperintensities.
Supporting Evidence
- The in-house program achieved an average overlap of 62.2% with manual segmentation.
- The Wu program had a mean overlap of only 36.81%, indicating poorer performance.
- SPM5 and SPM8 methods produced overlaps of 14.4% and 11.9%, respectively.
Takeaway
This study looked at different computer programs to help doctors find certain brain problems in older people using MRI scans, and one program worked better than the others.
Methodology
MRI scans were manually and automatically segmented using different software, and the overlap between methods was compared.
Potential Biases
Potential misclassification of non-WMH regions as WMHs due to the nature of the imaging techniques.
Limitations
The automated methods were less accurate for smaller or less defined white matter hyperintensities.
Participant Demographics
30 subjects aged 60+, including 10 without dementia and 20 with mild-to-moderate dementia.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website