Subject‐Level Segmentation Precision Weights for Volumetric Studies Involving Label Fusion
2024

Improving Brain Volume Measurement with Precision Weights

Sample size: 400 publication 10 minutes Evidence: high

Author Information

Author(s): Christina Chen, Sandhitsu R. Das, M. Dylan Tisdall, Fengling Hu, Andrew A. Chen, Paul A. Yushkevich, David A. Wolk, Russell T. Shinohara

Primary Institution: University of Pennsylvania

Hypothesis

Can subject-level precision weights improve the detection of differences in hippocampal volume between healthy individuals and those with cognitive impairments?

Conclusion

Incorporating subject-level precision weights significantly enhances the power to detect differences in hippocampal volume between groups.

Supporting Evidence

  • Incorporating precision weights improved the power to detect differences in hippocampal volume from 0.495 to 0.736.
  • The method was validated using data from the Alzheimer's Disease Neuroimaging Initiative (ADNI).
  • Precision weights were derived from bootstrapped atlas collections to quantify variability in volume estimates.
  • Statistical analysis showed that the weighted regression controlled type I error rates effectively.
  • Low precision weights were observed for outlier volume estimates, indicating effective downweighting.

Takeaway

This study shows that using special weights can help scientists get better measurements of brain size, which is important for understanding diseases like Alzheimer's.

Methodology

The study used multi-atlas segmentation and bootstrapping to derive subject-level precision weights for hippocampal volume estimates.

Potential Biases

Potential bias may arise from using atlases from healthy controls to segment images from disease cohorts.

Limitations

The method does not account for variability among multiple scans of the same subject under identical conditions.

Participant Demographics

The study included 200 controls (94 males, 106 females) and 200 AD subjects (120 males, 80 females), with ages ranging from 55 to 94.

Statistical Information

P-Value

p<0.01

Statistical Significance

p<0.01

Digital Object Identifier (DOI)

10.1002/hbm.70082

Want to read the original?

Access the complete publication on the publisher's website

View Original Publication