Atlas selection for hippocampus segmentation: Relevance evaluation of three meta-information parameters

Area of Science:

• Neuroimaging
• Medical Image Analysis

Background:

• Current hippocampus segmentation relies on pre-segmented templates (atlases) for image registration.
• Using a single standard atlas struggles with anatomical variations, particularly in older brains.
• Multiple atlases improve precision but increase computational costs.

Purpose of the Study:

• To develop an automated method for selecting optimal brain atlases for hippocampus segmentation.
• To improve segmentation accuracy without manual intervention or high computational overhead.
• To evaluate the efficacy of meta-information parameters in atlas selection.

Main Methods:

• Defined 24 atlases based on combinations of medical condition, age range, and gender.
• Segmented 352 brain volumes from the Alzheimer's Disease Neuroimaging Initiative (ADNI) database using these atlases.
• Compared segmentation results against reference segmentations to assess accuracy.

Main Results:

• Atlas selection based on meta-information significantly improved Dice similarity coefficient (DSC) to 0.68 ± 0.11.
• This outperformed the standard MNI152 atlas, which yielded a DSC of 0.62 ± 0.12.
• Statistical analysis confirmed significant accuracy gains from using the three meta-information parameters.

Conclusions:

• Meta-information-driven atlas selection is an effective strategy for enhancing hippocampus segmentation.
• This approach offers a computationally efficient way to improve segmentation precision across diverse anatomies.
• The findings support the use of demographic and clinical data for optimizing neuroimaging analysis.