A fast and robust hippocampal subfields segmentation: HSF revealing lifespan volumetric dynamics

Area of Science:

• Neuroimaging
• Neuroanatomy
• Developmental Neuroscience

Background:

• Hippocampal subfields are critical for episodic memory and exhibit distinct structural characteristics.
• Understanding hippocampal subfield volumes across the lifespan is essential for studying memory development and decline.
• Current MRI segmentation methods for small hippocampal subfields lack standardization and present challenges.

Purpose of the Study:

• To introduce and validate a novel, automated deep learning tool, the hippocampal segmentation factory (HSF), for segmenting hippocampal subfields.
• To assess the impact of age and sex on hippocampal subfield volumes using the HSF tool.
• To reconcile conflicting findings regarding lifespan trajectories of hippocampal subfields.

Main Methods:

• Developed an end-to-end deep learning pipeline (HSF) for automated segmentation of hippocampal subfields.
• Validated HSF against established tools (ASHS, HIPS, HippUnfold) using Dice Coefficient, Hausdorff Distance, and Volumetric Similarity.
• Applied HSF to 3,750 subjects from HCP datasets (development, young adults, aging) to analyze age and sex effects.

Main Results:

• HSF demonstrated superior accuracy compared to existing tools, closely matching manual segmentation.
• Significant differential maturation and aging effects were observed across hippocampal subfields, with the dentate gyrus most affected by age.
• Most hippocampal subfields showed accelerated growth and decay in males compared to females.

Conclusions:

• HSF provides a fast, robust, and accurate solution for hippocampal subfield segmentation, overcoming limitations of current methods.
• The study elucidates distinct developmental and aging trajectories for hippocampal subfields, influenced by age and sex.
• Findings contribute to a more unified understanding of memory-related brain structure changes throughout the human lifespan.