Interpreting scan data acquired from multiple scanners: a study with Alzheimer's disease

Area of Science:

• Neuroimaging
• Medical Physics
• Neurology

Background:

• Multi-site studies using Magnetic Resonance Imaging (MRI) data offer research advantages but face potential confounds from scanner variability.
• Existing literature lacks comprehensive analysis of multi-scanner data interactions with research outcomes, particularly in neurological disease studies.

Purpose of the Study:

• To investigate the impact of scanner variability on neuroimaging data within a single-center Alzheimer's disease (AD) study.
• To determine if scanner differences confound the detection of disease-specific brain changes and if scanner-by-disease interactions exist.

Main Methods:

• Utilized a dataset of 136 subjects (62 AD patients, 74 controls) with MRI scans acquired over 10 years on 6 different scanners.
• Employed whole-brain voxel-wise analysis to assess effects of scanner, disease, and their interaction.
• Analyzed grey matter differences in the medial temporal lobe, a region typically affected in AD.

Main Results:

• Significant grey matter reduction was observed in AD patients, consistent with disease progression.
• Scanner-related differences were minimal and only significant in the thalamus, substantially less than disease effects.
• No significant interaction between scanner type and disease group was detected.

Conclusions:

• Scanner variations in this study did not confound the detection of Alzheimer's disease-related brain changes.
• The findings support the validity of pooling multi-scanner MRI data for research, especially for rare diseases or multi-center collaborations.
• This analytical approach can justify data aggregation in future large-scale neuroimaging studies.