guest ::
| Home > Publications database > Deep texture features for surface-based characterization of fiber architecture in the human hippocampus (v1) |
| Home > Publications database > Deep texture features for surface-based characterization of fiber architecture in the human hippocampus (v1) |
| Dataset | FZJ-2024-07230 |
; ; ; ; ; ; ;
2024
EBRAINS
This record in other databases:
Please use a persistent id in citations: doi:10.25493/T1N3-KCX
Abstract: This dataset contains deep texture features characterizing architectonic patterns of the pyramidal layer in the hippocampal cornu Ammonis region and the subicular complex, as captured by three-dimensional polarized light imaging (3D-PLI). The 3D-PLI measurements were obtained from 547 individual brain sections of the right hippocampus of an 87-year-old male. Texture features were extracted from 3D-PLI images using a width-reduced ResNet-50 encoder trained on a 3D-Context Contrastive Learning (CL-3D) objective. The features were aggregated along the full depth of the pyramidal layer and transformed into an unfolded coordinate space using HippUnfold. In addition, we provide projections of the feature vectors onto 52 principal components with largest explained variance. These components reflect the general regional organization and reveal an expected functional rostro-caudal heterogeneity of the pyramidal layer. By transferring features to an unfolded coordinate system using HippUnfold, the data are placed in a canonical reference space, enabling comparisons with other unfolded datasets. Moreover, we provide unfolded surfaces for both the MNI ICBM 152 Nonlinear Asymmetric 2009c and the BigBrain histological space, allowing access to the features within these volumetric reference spaces as well.
Keyword(s): Neuroscience
|
The record appears in these collections: |