Multi-contrast submillimetric 3 Tesla hippocampal subfield segmentation protocol and dataset

Area of Science:

• Neuroimaging
• Neuroanatomy
• Magnetic Resonance Imaging (MRI)

Background:

• The hippocampus has distinct subregions crucial for cognition.
• Neurological and psychiatric conditions often affect these subregions.
• Current MRI methods for hippocampal substructure analysis require time-consuming manual segmentation.

Purpose of the Study:

• To share manual labels and high-resolution MRI data of hippocampal substructures.
• To facilitate the development of automated segmentation algorithms.
• To support neuroimaging assessments in basic and clinical neurosciences.

Main Methods:

• Acquired submillimetric T1- and T2-weighted MRI data from 25 healthy subjects using a 3 Tesla MRI system.
• Defined three hippocampal subregions: subicular complex, CA1-3, and CA4-DG.
• Employed segmentation guided by molecular layer characteristics and geometry-based boundaries.

Main Results:

• Generated a dataset (MNI-HISUB25) including MRI data, sequence information, and probabilistic maps.
• Achieved excellent intra-rater (Dice ≥90%) and inter-rater (Dice ≥87%) reliability for manual segmentation.
• The segmentation protocol demonstrated robustness and high reliability.

Conclusions:

• The provided dataset and manual labels can advance non-invasive identification of hippocampal substructures.
• This resource is valuable for developing automated segmentation algorithms for the hippocampus.
• The data will aid in neuroimaging studies of the mesiotemporal lobe and related disorders.