Jerod M Rasmussen1, Alice M Graham2, Pathik D Wadhwa1, Sonja Entringer3, Martin Styner4, Beatriz Luna5, Thomas G O'Connor6, Damien A Fair7, and Claudia Buss3
1UC Irvine, Irvine, CA, United States, 2Oregon Health Sciences University, Portland, OR, United States, 3Charité – Universitätsmedizin Berlin, Berlin, Germany, 4University of North Carolina, Chapel Hill, NC, United States, 5University of Pittsburgh, Pittsburgh, PA, United States, 6University of Rochester, Rochester, NY, United States, 7University of Minnesota, Minneapolis, MN, United States
1UC Irvine, Irvine, CA, United States, 2Oregon Health Sciences University, Portland, OR, United States, 3Charité – Universitätsmedizin Berlin, Berlin, Germany, 4University of North Carolina, Chapel Hill, NC, United States, 5University of Pittsburgh, Pittsburgh, PA, United States, 6University of Rochester, Rochester, NY, United States, 7University of Minnesota, Minneapolis, MN, United States
Whether within or across sites, including QC measures greatly increases FAWM variance explained by PMA. Thus, adjustment results in an improved capability for identifying reproducible small-effect size relationships in large multi-site infant imaging studies.
Figure 1. Multi-Site White Matter FA Model Without and With Consideration of Quality Control Measures. QC measures accounted for 47% of the remaining variance in FAWM after adjusting for PMA. Note the decrease in MSE and increase in the remaining variance in FAWM explained by PMA after adjusting for QC measures.
Table 2. Single Site Association Between Whole White Matter FA and Postmenstrual Age at Scan Modeled Without and With Controlling for Quality Control Measures. All sites demonstrated marked improvement in model performance when controlling for simple QC measures (mean FD, SNR, CNR). Improved performance is indicated by increased partial R-square values, t-scores, and increased consistency in age-FA slope estimates across sites. All age-FA associations were significant at a p<10-3 threshold for all models and sites.