Different FreeSurfer versions might generate different statistical outcomes in case-control comparison studies

Area of Science:

• Neuroimaging
• Computational Neuroscience
• Medical Image Analysis

Background:

• Neuroimaging pipelines, including FreeSurfer, are known to produce variable results due to factors like software version.
• The impact of these variations on inter-group comparisons, particularly in clinical research, remains under-explored.

Purpose of the Study:

• To investigate how different FreeSurfer versions affect statistical inferences in comparing brain volumes between healthy controls and type 1 diabetes patients.
• To assess the impact of software version on detecting significant volumetric differences in a case-control study model.

Main Methods:

• Structural MRI scans (T1- and T2-weighted) from healthy controls and type 1 diabetes patients were processed using FreeSurfer versions 5.3, 5.3 HCP, 6.0, and 7.1.
• Inter-group statistical comparisons of volumetric parameters were performed for each FreeSurfer version's output.

Main Results:

• FreeSurfer 5.3 identified significant cortical and subcortical volume differences.
• FreeSurfer 5.3 HCP and 6.0 reported less pronounced subcortical differences.
• FreeSurfer 7.1 failed to detect any statistically significant inter-group volume differences.

Conclusions:

• Discrepancies in findings are attributed to increased within-group variability in the type 1 diabetes cohort, not group average differences.
• Validation of neuroimaging software in clinical scenarios like case-control studies is crucial during development.
• Confirmatory analyses with analogous software may be necessary alongside quality control for new FreeSurfer versions.