Multisite fMRI reproducibility of a motor task using identical MR systems

Abstract

Purpose:To estimate between-scanner functional MRI (fMRI) reproducibility in a multisite study. Materials and Methods:A total of five identical 1.5T MR systems were used to repeatedly scan five subjects while performing a finger tapping task. A two-way (scanners, subjects) random effects analysis of variance (ANOVA) was used to estimate between-scanner and between-subject variability on two outcome variables: task-related mean blood oxygenation level dependent (BOLD) signal change and volume of activation within a motor region-of-interest (ROI). Results:Between-scanner variability of fMRI data accounted for a small proportion of the total variation in the BOLD signal change (8.34%, P ϭ 0.114) and volume of activation (5.46%, P ϭ 0.203). Between-subject variation accounted for more than half of the total variation for both measurements (57.17% and 54.46%, respectively, P Ͻ 0.01). Conclusion:These results support the feasibility of multisite studies using identical scanner systems. MULTISITE STUDIES may remediate some of the limitations in current functional magnetic resonance imaging (fMRI) research, in particular the difficulties in assembling large sample sizes in clinical populations. Sufficient reproducibility of fMRI measurements across scanners is necessary for multisite investigations. In an extreme situation, if most of the variance in the measurements were attributable to differences between scanners rather than the factors of interest (i.e., subjects, tasks), fMRI results would be largely site-dependent and the validity of pooling results over sites would be in doubt. Estimating variance components is thus important to assess the validity of multisite measurements.Previous quantitative estimates of the effect of between-scanner variability on functional measurements are, to our knowledge, limited to the results of an experiment performed by the Biomedical Informatics Research Network (1,2). This study identified two scanner-related variables (field strength and type of k-space) in addition to the particular subject under study as significant predictors of result accuracy. However, no estimate of the variance attributable to each of these factors was reported. Other multiscanner investigations have assessed reproducibility using qualitative comparison of the activation maps (3,4).In the present work, we report the results of a pilot investigation of between-scanner reproducibility. Five subjects were scanned in five magnetic resonance systems on the same day, while performing a blocked design finger-tapping paradigm. Our analysis focuses on a region of interest (ROI) in the motor strip. Reproducibility was estimated using random effects variance components such that the total variance was partitioned into between-scanner, between-subject, and residual components (5). MATERIALS AND METHODS Participants and TaskParticipants were five healthy, right-handed volunteers (four female), aged 25 to 30 years. The experimental condition was self-paced finger tapping with the right index finger, and the control cond...

The neuro/PsyGRID calibration experiment

“…Analysis of variance modeled within-and between-center factors, with within-center reliability being relatively much greater, and enumerated the between-center differences in a number of key metrics. Other calibration studies have been reported [Costafreda et al, 2007;Deoni et al, 2008;Gountouna et al, 2009;Moorhead et al, 2009;Suckling et al, 2008], with similar outcomes from comparable analyses.…”

“…Justification for pooling of imaging data has also been sought for reproducibility of task-induced fMRI activation patterns, dependent on the applied statistical threshold and the details of the task stimuli Vlieger et al 2003] and comparison within activated regions of within-and between-center variance components [Costafreda et al 2007;Gountouna et al, 2009). Similar regional analyses have also been conducted on tissue segmentations of sMRI van Haren et al, 2003].…”

Progress toward openness, transparency, and reproducibility in cognitive neuroscience

“…They reported that the reliability of group-level results, using individual participants tested at different times, varied depending on the temporal gap between the tests, the specific tasks employed (sensory/motor vs. cognitive), design factors (block vs. event-related), the magnitude of activations, and other inter-individual factors. Importantly, Miller 59 and others 60 found that, like the difference between within-and between site variability found by the fBIRN team 49 , variability within participants across testing sessions was lower than variability between participants. Differences in tasks, the degree of selectivity of active voxels to those tasks, and subject motion appeared to be the biggest contributors to intersubject variability 61 .…”

Estudo da variabilidade do efeito BOLD em ressonância magnética funcional em sistemas de 3Teslas

“…Estas máquinas trazem vantagens pela maior relação sinal ruído (RSR), aumentando a sensibilidade do efeito BOLD à atividade neural, reduzindo artefatos descritos de baixa sensibilidade mais frequentes em estudos anteriores realizados com aparelhos de 1,5T (Beisteiner et al, 2000, Kruger et al, 2001, Krings et al, 2001. Além disso, é possível utilizar o aumento da RSR para adquirir imagens com voxels isotrópicos de menores dimensões e consequente maior resolução espacial (Roessler et al,2005 (Costafreda et al, 2007). …”

“…Estas variações estão associadas a diferentes tipos de equipamentos de RM, indivíduos e função testada, entre outros (Le e Hu 1997;Genovese et al, 1997;Miezin et al, 2000;Brannen et al, 2001;Maitra el al 2002;Wei et al, 2002). E, apesar de qualquer estudo de RMf ser influenciado, em maior ou menor grau, pela variabilidade da técnica, estes fatores estão presentes em particular em estudos multicêntricos (Casey et al, 1998;Macgonigle et al, 2000;Duann et al, 2002 (Costafreda et al, 2007).…”