Related Experiment Video
Updated: Apr 11, 2026
10:31
In Vitro Wedge Slice Preparation for Mimicking In Vivo Neuronal Circuit Connectivity
Published on: August 18, 2020
6.1K
Improving Neuropathological Reconstruction Fidelity via AI Slice Imputation
Marina Crespo Aguirre1,2, Jonathan Williams-Ramirez1, Dina Zemlyanker1
1Martinos Center for Biomedical Imaging, Massachusetts General Hospital and Harvard Medical School, United States.
Research Square
|April 10, 2026
Summary
This study introduces a super-resolution method to improve 3D brain reconstructions from dissection photographs, enhancing anatomical accuracy for neuropathology and neuroimaging research.
Area of Science:
- Neuroimaging
- Computational Anatomy
- Neuropathology
Background:
- Accurate 3D brain reconstructions are crucial for neuropathological analysis and morphometric accuracy.
- Current methods reconstructing 3D volumes from 2D dissection photographs often produce overly smooth and coarse results, especially with anisotropic data.
- High anisotropy, resulting from thick slab dissections, poses a significant challenge for detailed anatomical delineation.
Purpose of the Study:
- To develop a computationally efficient super-resolution technique for generating anatomically consistent isotropic volumes from anisotropic 3D reconstructions of dissection photographs.
- To enhance the resolution and anatomical fidelity of 3D brain models derived from 2D dissection images.
- To improve automated segmentation and atlas registration accuracy for neuropathological studies.
Main Methods:
- Introduced a super-resolution step to impute missing slices in anisotropic 3D reconstructions.
- Trained the model on domain-randomized synthetic data to ensure generalization across protocols and robustness to artifacts.
- Validated the method on surface reconstruction and atlas registration tasks.
Main Results:
- The super-resolution method generates anatomically consistent isotropic volumes from anisotropic 3D reconstructions.
- Imputed volumes led to improved automated segmentations with higher Dice scores, particularly in cortical and white matter regions.
- Demonstrated more accurate cortical surfaces and better alignment with target templates in atlas registration tasks.
Conclusions:
- The developed super-resolution approach significantly enhances the resolution and anatomical fidelity of 3D reconstructions from dissection photographs.
- This method strengthens the integration of neuropathology and neuroimaging by improving morphometric accuracy and anatomical delineation.
- The publicly available method facilitates advanced neuropathological analyses using enhanced 3D brain models.
