Automating Landmark Selection for Brain Image Registration
Author Information
Author(s): Liu Yutong, Sajja Balasrinivasa R., Uberti Mariano G., Gendelman Howard E., Kielian Tammy, Boska Michael D.
Primary Institution: University of Nebraska Medical Center
Hypothesis
Can a technique be developed to automate landmark selection for nonlinear medical image registration?
Conclusion
The study found that automating landmark selection significantly improves registration accuracy compared to manual selection.
Supporting Evidence
- Automated landmark selection improved registration accuracy in most data sets.
- Trends towards improvement were observed in some cases with landmark optimization.
- Manual adjustments also showed trends towards improved accuracy.
Takeaway
This study created a way to automatically pick points on brain images to help match them up better, making it easier to see changes in the brain.
Methodology
The method involved generating contours on anatomical features, placing landmarks, and optimizing their positions using a cost function based on local curvature.
Potential Biases
Potential biases may arise from manual adjustments made by technicians.
Limitations
The technique still requires some user intervention and may suffer from inter- and intra-investigator inconsistencies.
Participant Demographics
Five mice were used in the imaging studies.
Statistical Information
P-Value
p<0.05
Statistical Significance
p<0.05
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website