Improved Statistical Power with a Sparse Shape Model in Detecting an Aging Effect in the Hippocampus and Amygdala

Area of Science:

• Medical image analysis
• Computational anatomy
• Statistical shape modeling

Background:

• Sparse regression is common in medical imaging but underutilized in anatomical studies.
• Laplace-Beltrami (LB) eigenfunctions traditionally represent shapes, but discarding terms can lose information.
• Existing methods may not optimally select significant eigenfunctions for shape reconstruction.

Purpose of the Study:

• To present a sparse shape modeling framework utilizing LB eigenfunctions for anatomical studies.
• To improve statistical power in shape analysis by selectively filtering eigenfunctions.
• To investigate the influence of age on amygdala and hippocampus shapes.

Main Methods:

• Developed a sparse shape modeling framework based on LB eigenfunctions.
• Applied a sparse penalty to filter significant eigenfunctions, acting as a smoothing process for dense anatomical data.
• Utilized the framework to analyze age-related shape changes in the amygdala and hippocampus.

Main Results:

• The proposed LB-based sparse regression framework enhances statistical power.
• The method effectively filters significant eigenfunctions, reducing false negatives in dense anatomical data.
• Demonstrated increased statistical power in analyzing age effects on brain structures.

Conclusions:

• The LB sparse shape modeling framework offers improved statistical power for anatomical studies.
• This approach provides a more effective way to utilize LB eigenfunctions for shape representation and analysis.
• The method shows promise for investigating anatomical variations in populations, such as age-related changes.