A longitudinal human phantom reliability study of multi-center T1-weighted, DTI, and resting state fMRI data

Abstract

Multi-center MRI studies can enhance power, generalizability, and discovery for clinical neuroimaging research in brain disorders. Here, we sought to establish the utility of a clustering algorithm as an alternative to more traditional intra-class correlation coefficient approaches in a longitudinal multi-center human phantom study. We completed annual reliability scans on 'travelling human phantoms'. Acquisitions across sites were harmonized prospectively. Twenty-seven MRI sessions were available across four participants, scanned on five scanners, across three years. For each scan, three metrics were extracted: cortical thickness (CT), white matter fractional anisotropy (FA), and resting state functional connectivity (FC). For each metric, hierarchical clustering (Ward's method) was performed. The cluster solutions were compared to participant and scanner using the adjusted Rand index (ARI). For all metrics, data clustered by participant rather than by scanner (ARI > 0.8 comparing clusters to participants, ARI < 0.2 comparing clusters to scanners). These results demonstrate that hierarchical clustering can reliably identify structural and functional scans from different participants imaged on different scanners across time. With increasing interest in data-driven approaches in psychiatric and neurologic brain imaging studies, our findings provide a framework for multi-center analytic approaches aiming to identify subgroups of participants based on brain structure or function.

Keywords: Cortical thickness; DTI; Fractional anisotropy; Functional connectivity; Hierarchical clustering; Human phantoms; Multi-site; Scanner reliability; Scanner variability; T1; rsfMRI.

Fig. 1.

Mean cortical thickness (CT; left), fractional anisotropy (FA; center), and functional connectivity (FC; right) values for every scan, separated by color/shape (for participant ID) and scanner (columns). Significant scanner based differences in the means were present in both CT and FA, but not FC.

Fig. 2.

Results of the hierarchical clustering analysis for: A) cortical thickness (CT); B) fractional anisotropy (FA); C) functional connectivity (FC). The distance matrix shows Euclidean distances between scans (defined as the sum of the squared difference between each ROI for each pair of scans, such that lower distances between scans mean they are more similar). The cluster tree (dendrogram) is shown on the left. Color coding on the dendrogram represents participant ID. Participant ID, scanner, and year for each scan in the distance matrix is shown on the right.

Fig. 3.

Cluster solutions for cortical thickness (CT; top), fractional anisotropy (FA; middle) and functional connectivity (FC; bottom) for cluster solutions ranging from two to 20 clusters. An analysis was conducted to establish if a given cluster solution was related to the participant ID or scanner. Year was included as a ‘control’ variable. An adjusted Rand index (ARI) was calculated for each cluster solution and scan-related variables, namely participant ID (4 labels; P1, P2, P3, P4), scanner (5 labels; CMH, MRC, MRP, ZHH, ZHP), year (3 labels; Year1, year2, year3), a combination of ID by scan site (16 labels; e.g. P1 at CMH, P1 at MRC, etc.), ID by year (9 labels; e.g. P1 during Year1, P1 during year2, etc.), and year by scanner (9 labels, e.g. CMH at Year1, CMH at year2, etc.). Circles/dots indicate ARI values which are above chance as determined via a null distribution created using a permutation test, suggesting greater than chance overlap between that label and the cluster solution.