Related Experiment Video
Updated: Jan 4, 2026
11:50
A Standardized Pipeline for Examining Human Cerebellar Grey Matter Morphometry using Structural Magnetic Resonance Imaging
Published on: February 4, 2022
4.5K
Hierarchical statistical shape analysis and prediction of sub-cortical brain structures.
Anil Rao1, Paul Aljabar, Daniel Rueckert
1Visual Information Processing, Department of Computing, Imperial College London, 180 Queens's Gate, London SW7 2BZ, UK. anil.w.rao@gsk.com
Medical Image Analysis
|August 11, 2007
Summary
This study uses statistical methods like canonical correlation analysis and partial least squares regression to analyze brain structure relationships. Adjacent sub-cortical structures show the strongest correlations, improving prediction accuracy.
Area of Science:
- Neuroimaging
- Multivariate Statistics
- Computational Neuroscience
Background:
- Understanding statistical relationships between brain structures is crucial for neuroscience.
- Existing methods may not fully capture complex inter-structural variations.
Purpose of the Study:
- To apply and evaluate two multivariate statistical techniques for analyzing brain structure variability.
- To quantify and predict correlated behavior between sub-cortical structures using 3D MR images.
Main Methods:
- Canonical Correlation Analysis (CCA) to quantify correlated behavior between sets of variables.
- Partial Least Squares Regression (PLSR) to predict variables from other sets.
- Application to 178 sets of 3D MR images of 18 sub-cortical structures.
Main Results:
- CCA revealed correlation coefficients between 0.51 and 0.67 for pairwise structures, with adjacent structures showing strongest correlations.
- PLSR improved prediction accuracy (sum squared error of 4.26 mm²) compared to using the mean shape (6.75 mm²).
Conclusions:
- Multivariate statistical techniques effectively quantify and predict inter-structural relationships in the brain.
- Adjacent brain structures exhibit stronger statistical correlations.
- A hierarchical approach combining PLSR with model fitting can further enhance prediction accuracy.
