Harmonization of multi-site diffusion tensor imaging data for cervical and thoracic spinal cord at 1.5 T and 3 T...

Area of Science:

• Neuroimaging
• Biomedical Engineering
• Radiology

Background:

• Diffusion tensor imaging (DTI) is crucial for spinal cord studies.
• MRI scanner hardware, field strengths, and sequence parameters introduce variability in DTI metrics.
• Reliability of DTI across different scanners, especially for the thoracic cord, is not well established.

Purpose of the Study:

• To assess the reliability of DTI metrics across different MRI scanners and field strengths for the entire spinal cord.
• To evaluate the effectiveness of longitudinal ComBat (longComBat) harmonization in reducing scanner-induced variability.
• To determine the potential for multi-site and longitudinal spinal cord DTI research.

Main Methods:

• DTI data collected from 30 healthy adults using four scanners (Siemens 3T/1.5T, Philips 3T/1.5T).
• DTI metrics (FA, MD, AD, RD) extracted and averaged per vertebral level.
• Data harmonization performed using longitudinal ComBat (longComBat) to correct for scanner variability.

Main Results:

• Fractional Anisotropy (FA) metrics showed relative consistency across scanners.
• Significant variability observed in diffusivity measures (MD, AD, RD) across scanners.
• longComBat harmonization substantially improved agreement of all DTI metrics between scanners.

Conclusions:

• Scanner hardware and sequence differences are major sources of DTI variability in the spinal cord.
• longComBat harmonization is effective in improving inter-scanner agreement for spinal cord DTI.
• Harmonized DTI data facilitates multi-site clinical research and trials for spinal cord studies.