Related Experiment Video
Updated: May 11, 2025
10:26
A Multicenter MRI Protocol for the Evaluation and Quantification of Deep Vein Thrombosis
Published on: June 2, 2015
17.0K
Res-Net-Based Modeling and Morphologic Analysis of Deep Medullary Veins Using Multi-Echo GRE at 7 T MRI
Zhixin Li1,2,3, Li Liang4, Jinyuan Zhang1,3
1State Key Laboratory of Cognitive Science and Mental Health, Institute of Biophysics, Chinese Academy of Sciences, Beijing, China.
NMR in Biomedicine
|April 17, 2025
Summary
This study introduces an automated method for modeling and quantifying deep medullary veins (DMVs) using 7 Tesla MRI. The novel approach accurately measures DMV morphology, revealing significant differences in patients with subcortical vascular dementia (SVaD).
Area of Science:
- Neuroimaging
- Medical Image Analysis
- Vascular Neurology
Background:
- Pathological changes in deep medullary veins (DMVs) are linked to various diseases.
- Accurate modeling and quantification of DMVs present significant challenges in current medical imaging.
- Deep medullary veins play a crucial role in cerebral small vessel diseases.
Purpose of the Study:
- To propose and evaluate an automated approach for modeling and quantifying DMVs using 7 Tesla (7T) MRI.
- To assess the geometric parameters of DMVs in patients with subcortical vascular dementia (SVaD) compared to controls.
- To establish the utility of DMV morphologic parameters in the research and diagnosis of cerebral small vessel diseases.
Main Methods:
- Development of a multi-echo-input Res-Net for precise vascular segmentation of DMVs.
- Implementation of a minimum path loss function for accurate modeling and quantification of DMV geometric parameters.
- Acquisition of gradient echo images with five echoes at 7T from 21 SVaD patients and 20 controls, with manual labeling for comparison.
Main Results:
- The automated method demonstrated high accuracy, with no significant offset in centerline detection compared to manual labeling (p=0.734).
- The length difference between the automated method and manual labeling was less than inter-clinician variability (p<0.001).
- SVaD patients showed significantly fewer DMVs (p=0.011) and higher curvature (p<0.0001), correlating with cognitive decline (VaDAS-Cog and MMSE scores).
Conclusions:
- A novel framework for automated quantification of DMV morphologic parameters has been successfully developed and validated.
- The proposed method offers superior accuracy and consistency compared to manual analysis, reducing inter-observer variability.
- Quantified DMV characteristics show promise as biomarkers for the research and clinical diagnosis of cerebral small vessel diseases, particularly in SVaD.
