Autistic vs. Control Differences in MRI Scan Quality Across ABIDE-II Sites

Area of Science:

• Neuroimaging
• Autism Spectrum Disorder Research
• Data Quality Assessment

Background:

• Head motion and scan quality variability are significant challenges in autism neuroimaging.
• Large multi-site datasets like ABIDE-II enable systematic analysis of MRI data quality and site heterogeneity.
• Understanding these factors is crucial for accurate diagnostic differences in autism research.

Purpose of the Study:

• To quantify diagnostic differences in MRI data quality between autistic (ASD) and typically developing (TD) individuals.
• To assess the influence of site-level heterogeneity on neuroimaging data quality.
• To evaluate the impact of quality-control thresholds on sample composition in multi-site autism studies.

Main Methods:

• Utilized Functional MRI Quality Assessment Protocol (QAP) metrics and phenotypic data from the ABIDE-II dataset.
• Analyzed key quality metrics including mean framewise displacement (mFD), signal-to-noise ratio (SNR), and entropy focus criterion (EFC).
• Employed non-parametric tests and linear mixed-effects models, accounting for site as a random factor, to compare ASD and TD groups.

Main Results:

• ASD participants showed significantly greater head motion (mFD) compared to TD individuals across most sites.
• Signal quality differences (SNR, EFC) were modest and largely attributable to site effects.
• Simulated quality-control procedures disproportionately excluded ASD participants (up to 31%) versus TD participants (18%).

Conclusions:

• Autistic individuals consistently exhibit higher head motion in neuroimaging studies.
• Signal quality variations are minimal and primarily driven by site-specific factors.
• Current quality-control methods unfairly exclude ASD individuals, underscoring the need for improved motion handling and balanced QC strategies in multi-site research.