A multi-center study: intra-scan and inter-scan variability of diffusion spectrum imaging.

Area of Science:

• Neuroimaging
• Medical Physics
• Data Science

Background:

• Diffusion Spectrum Imaging (DSI) enables detailed brain white matter tract analysis.
• Standardizing DSI data acquisition across multiple sites and scanners is crucial for large-scale studies.
• Variability in DSI data can arise from different scanner hardware and configurations.

Purpose of the Study:

• To assess the feasibility of pooling Diffusion Spectrum Imaging (DSI) data acquired from four scanners at three different sites.
• To quantify the reproducibility and variability of DSI scalar maps (ADC, FA, GFA) across scanners and sites.
• To evaluate the impact of identical versus different scanner configurations on data consistency.

Main Methods:

• Extracted scalar maps (ADC, FA, GFA) from DSI data.
• Performed region-based and tract-based analyses to measure variability using coefficient of variation (CV) and Intraclass Correlation Coefficient (ICC).
• Conducted intra-scanner reproducibility, twin-scanner (identical settings), and cross-configuration reproducibility experiments, including a phantom study.

Main Results:

• Region-based CV ranged from 1%-4.2% and tract-based CV was below 3% for most bundles in intra-scanner and twin-scanner experiments.
• The uncinate fasciculus showed the highest variability (CV 3.7-6%) for FA and GFA values.
• High similarity in outcomes was observed between scanners with identical settings, but not between those with different configurations. ICC values of 0.7+ were common for FA and GFA maps.

Conclusions:

• Cross-site pooling of DSI data is feasible, particularly when utilizing identical scanner configurations.
• Scanner configuration is a significant factor influencing DSI data reproducibility across sites.
• The findings support the standardization and harmonization of DSI data acquisition protocols for multi-site research.