Related Experiment Video
Updated: Aug 6, 2026
07:45
Quantifying Intermembrane Distances with Serial Image Dilations
Published on: September 28, 2018
A comparative evaluation of multiple enlarged perivascular space segmentation tools
James D LeFevre1, W Hudson Robb2, Dandan Liu3
1Vanderbilt Memory and Alzheimer's Center, Vanderbilt University School of Medicine, Nashville, TN, USA; Vanderbilt Alzheimer's Disease Research Center, Nashville, TN, USA; Vanderbilt Brain Institute, Vanderbilt University, Nashville, TN, USA.
Magnetic Resonance Imaging
|July 22, 2026
Summary
We developed DORES, an automated deep learning tool for segmenting enlarged perivascular spaces (ePVS) in brain MRI scans. DORES shows strong performance in older adults, aiding in the assessment of cerebral small vessel disease.
Area of Science:
- Neuroimaging
- Artificial Intelligence
- Cerebrovascular Diseases
Background:
- Enlarged perivascular spaces (ePVS) are indicators of cerebral small vessel disease and may signify impaired brain waste clearance.
- Manual quantification of ePVS is impractical for large-scale studies.
Purpose of the Study:
- To develop and validate an automated deep learning tool for segmenting enlarged perivascular spaces (ePVS) in brain MRI.
- To assess the performance and generalizability of the automated tool.
Main Methods:
- Developed DORES, a 3D nnU-Net deep learning algorithm for ePVS segmentation using T1-weighted and FLAIR MRI.
- Trained DORES on manually segmented scans and large pseudo-labeled datasets from the Vanderbilt Memory and Aging Project (VMAP).
- Evaluated DORES against other tools using Dice scores, volume differences, and correlation analyses, with external validation on the Alzheimer's Disease Neuroimaging Initiative (ADNI3) dataset.
Main Results:
- DORES achieved robust performance within the VMAP cohort, with Dice scores of 0.61 (white matter) and 0.72 (basal ganglia).
- Strong correlations and agreement were observed for ePVS count and volume.
- Performance showed modest decline in the external ADNI3 validation set, with scanner-dependent variations noted.
Conclusions:
- DORES offers a reliable, multimodal nnU-Net-based pipeline for ePVS segmentation in older adults.
- The tool demonstrates strong within-cohort performance and acceptable external validity.
- Scanner-related effects may influence cross-site consistency, highlighting a limitation for multi-center applications.
