Related Experiment Video
Updated: Mar 26, 2026
11:03
High-resolution In Vivo Manual Segmentation Protocol for Human Hippocampal Subfields Using 3T Magnetic Resonance Imaging
Published on: November 10, 2015
10.1K
Automated Hippocampal Subfield Segmentation at 7T MRI
L E M Wisse1, H J Kuijf2, A M Honingh3
1From the Penn Image Computing and Science Laboratory, Department of Radiology (L.E.M.W., J.B.P., S.R.D., P.A.Y.) Laura.Wisse@uphs.upenn.edu.
AJNR. American Journal of Neuroradiology
|February 6, 2016
Summary
An automated technique accurately segments hippocampal subfields and the entorhinal cortex using 7T MRI. This computational method offers a viable, efficient alternative to manual segmentation for research.
Area of Science:
- Neuroimaging
- Computational anatomy
Background:
- High-resolution 7T MRI enables detailed in vivo investigation of hippocampal subfields.
- Manual segmentation of these structures is time-consuming and labor-intensive.
Purpose of the Study:
- To evaluate an automated computational technique for segmenting hippocampal subfields and the entorhinal cortex.
- To compare the accuracy of automated segmentation against manual segmentation in 7T MRI data.
Main Methods:
- Manual segmentation of hippocampal subfields (CA1-CA3, dentate gyrus, subiculum) and entorhinal cortex on 0.70-mm³ T2-weighted 7T MRI images from 26 participants.
- Application and evaluation of an automated segmentation approach using leave-one-out cross-validation.
Main Results:
- Automated segmentation achieved Dice similarity coefficients >0.75 for CA1, dentate gyrus, subiculum, and entorhinal cortex.
- Intraclass correlation coefficients were >0.74 for CA1, dentate gyrus, and subiculum.
- Accuracy of automated segmentation was comparable to or slightly exceeded that of manual raters with extensive training.
Conclusions:
- The developed computational technique automates hippocampal subfield and entorhinal cortex labeling at 7T MRI with high accuracy.
- This automated method is a feasible and competitive alternative to manual segmentation.
- The software and atlas are publicly available for broader research use.
