Traveling subject (TS) approach is a good harmonization method for multi-site brain image data (Yamashita et al., 2019), which can differentiate measurement bias (mainly derived from MRI equipment and protocols) from sampling bias (mainly derived from case and control recruitment).
Based on the previous study, we are also obtaining TS data for the project and CRHD and HARP protocols. In some sites, we intend to combine with existing large-scale datasets and include the former scan procedure in the TS project. Previous data harmonization using the TS dataset was based on a general linear model (GLM; Yamashita et al., 2019). However, the present TS project may provide under a general linear mixed model (GLMM), this is because the new protocols require a longer scan time (60 min) compared to previous protocols (~20 min only for T1W imaging and rsfMRI). Therefore, we applied a hub-and-spoke model to arrange the scans. Each participant undergoes CRHD and HARP scans using the Prisma (~2 hours) at one or more hub sites (UTK, UTI, and ATR) to harmonize the data (See the connections in red in the figure below). The other visiting sites were pre-determined considering the site locations, machine differences, and target disease similarities (within Group) between the sites. Each participant receives 2 or more scans at the recruitment site to see a test-retest reliability. The duration between test-retest scans may vary due to scanner time restrictions and therefore we intend to test whether the gap would be associated with the reliability.
A strength of the GLMM approach is the flexibility of TS scans, the number of scans per participant and length of scan time per protocol, which is reasonable for a big project. This approach allows to add another site, scanner, and protocol to an existing TS network easily. In this project, the scanners at two sites were upgraded another machine after the project had started and we can customize the TS scans for two sites to ensure that the data are properly connected before and after the upgrades. Therefore, the project welcomes other sites to participate in the TS network.
For the TS project, each site recruits five or more participants and total 75 healthy adults planned to be recruited. Each participant undergoes 6 to 8 scans at three or more sites within 6 months. The total scans and the spokes between the sites are expected to be 455 and 465, respectively. As of March 2020, 74 participants were registered and 405 scans (89.0 %) were completed which provided 368 spokes (76.1 %).
- Yamashita, A., Yahata, N., Itahashi, T., Lisi, G., Yamada, T., Ichikawa, N., . . . Imamizu, H. (2019). Harmonization of resting-state functional MRI data across multiple imaging sites via the separation of site differences into sampling bias and measurement bias. PLoS Biol, 17(4), e3000042. doi:10.1371/journal.pbio.3000042