Robustness and stability of volume-based tractography in a multicenter setting
Johannes Forsting1, Robert Rehmann1,2, Marlena Rohm1,3
1Department of Neurology, BG-University Hospital Bergmannsheil, Ruhr-University Bochum, Bochum, Germany.
NMR in Biomedicine
|February 1, 2022
Summary
Volume-based tractography (VBT) demonstrates robust and reproducible muscle imaging results across multiple centers. Semiautomatic segmentation approaches offer excellent agreement with manual methods, enabling data pooling for muscle diffusion tensor imaging (mDTI) studies.
Area of Science:
- Biomedical Engineering
- Medical Imaging
- Musculoskeletal Imaging
Background:
- Muscle diffusion tensor imaging (mDTI)-based tractography aids in detecting subclinical muscle injuries and understanding neuromuscular diseases.
- Traditional region of interest (ROI)-based tractography is labor-intensive and requires significant expertise.
- Volume-based tractography (VBT) offers a potentially more efficient alternative, but its reliability needs validation.
Purpose of the Study:
- To assess the multicenter performance and stability of VBT in muscle imaging.
- To evaluate the comparability of outcome measures using semiautomatic segmentation approaches with VBT data.
- To determine the feasibility of pooling VBT data from different centers.
Main Methods:
- Five volunteers underwent 3-T mDTI scans of calf muscles across six MR sites.
- Tract properties (e.g., tract density, volume) and diffusion metrics were calculated using VBT.
- Multicenter reproducibility was assessed using within-subject coefficients of variance (wsCVs) and intraclass correlation coefficients (ICCs).
- Feasibility of VBT with manual and semiautomatic segmentation was evaluated using 50 pooled datasets.
- Statistical analyses included ANOVA, ICC, and Bland-Altman plots to compare segmentation methods.
Main Results:
- VBT demonstrated good reproducibility for tract density and volume, along with diffusion metrics, across multiple centers.
- No significant differences were found between manual and semiautomatic segmentation approaches (ICCs ≥ 0.992).
- Bland-Altman analysis confirmed no systemic bias between manual and semiautomatic segmentation.
- VBT-derived tract properties and diffusion metrics showed good comparability across centers.
- Semiautomatic segmentation exhibited excellent agreement with manual segmentation.
Conclusions:
- Volume-based tractography (VBT) provides reproducible and comparable results for muscle diffusion tensor imaging (mDTI) across different centers.
- Semiautomatic segmentation methods are reliable and comparable to manual segmentation for VBT data analysis.
- The findings support the feasibility of pooling multicenter mDTI data using VBT and semiautomatic segmentation to build reference databases.
