Traveling Subject-Informed Harmonization Increases Reliability of Brain Diffusion Tensor and Neurite Mapping

Area of Science:

• Neuroimaging
• Biomedical Engineering
• Radiology

Background:

• Diffusion-weighted magnetic resonance imaging (dMRI) reveals brain microstructural changes in neurological and psychiatric disorders.
• Inconsistencies across MRI models hinder the clinical use of dMRI metrics.
• Harmonization is crucial for reliable dMRI data analysis in multi-site studies.

Purpose of the Study:

• To evaluate the effectiveness of ComBat and TS-GLM harmonization methods for dMRI metrics.
• To compare the harmonization performance of ComBat and TS-GLM.
• To assess the impact of harmonization on intersubject biological effects and scanner-induced variability.

Main Methods:

• Utilized harmonized dMRI data from 300 scans across 69 traveling subjects (TS).
• Evaluated 13 MRI models and 2 protocols, focusing on diffusion tensor imaging (DTI) and neurite orientation dispersion and density imaging (NODDI) metrics.
• Applied ComBat and TS-GLM harmonization techniques and compared their efficacy.

Main Results:

• Both ComBat and TS-GLM significantly reduced site, model, and protocol effects for diffusion metrics.
• TS-GLM demonstrated superior harmonization power compared to ComBat.
• Harmonization methods effectively preserved intersubject biological effects.

Conclusions:

• ComBat and TS-GLM effectively reduce scanner effects in DTI and NODDI white matter metrics.
• These harmonization approaches retain crucial biological and pathological effects.
• Harmonized dMRI metrics enable accurate detection of subtle pathological changes in large, multi-site studies.