Harmonization of multi-scanner in vivo magnetic resonance spectroscopy: ENIGMA consortium task group considerations

Abstract

Magnetic resonance spectroscopy is a powerful, non-invasive, quantitative imaging technique that allows for the measurement of brain metabolites that has demonstrated utility in diagnosing and characterizing a broad range of neurological diseases. Its impact, however, has been limited due to small sample sizes and methodological variability in addition to intrinsic limitations of the method itself such as its sensitivity to motion. The lack of standardization from a data acquisition and data processing perspective makes it difficult to pool multiple studies and/or conduct multisite studies that are necessary for supporting clinically relevant findings. Based on the experience of the ENIGMA MRS work group and a review of the literature, this manuscript provides an overview of the current state of MRS data harmonization. Key factors that need to be taken into consideration when conducting both retrospective and prospective studies are described. These include (1) MRS acquisition issues such as pulse sequence, RF and B0 calibrations, echo time, and SNR; (2) data processing issues such as pre-processing steps, modeling, and quantitation; and (3) biological factors such as voxel location, age, sex, and pathology. Various approaches to MRS data harmonization are then described including meta-analysis, mega-analysis, linear modeling, ComBat and artificial intelligence approaches. The goal is to provide both novice and experienced readers with the necessary knowledge for conducting MRS data harmonization studies.

Comparison of different approaches to manage multi-site magnetic resonance spectroscopy clinical data analysis

“…When harmonizing by vendor, a significant effect of site remained, though the effect of vendor was removed. This is perhaps not surprising given the known differences between MRS data collected on different scanners (Harris et al, 2022), and it supports the use of ComBat harmonization at the scanner level, unless the site effect is meaningful for a particular study. Replicating the results of Bell et al (2022) in a different pediatric clinical dataset is important, as it confirms the utility of ComBat harmonization for MRS data, which is in line with recent commentaries on the importance of reproducibility in science (Stoddart, 2016;Kozlov, 2022).…”

“…This work demonstrates that caution is needed when controlling for scanner/site/vendor, as this can have substantial implications on the results. Despite the increasing standardization of imaging and MRS protocols, acquisition schemes can differ by vendor, for example, the pulse shapes and minimum achievable echo time (Harris et al, 2017(Harris et al, , 2022. Furthermore, differences exist between scanners of the same vendor, for example, eddy currents and their impact on MRS data (Harris et al, 2022).…”

“…Despite the increasing standardization of imaging and MRS protocols, acquisition schemes can differ by vendor, for example, the pulse shapes and minimum achievable echo time (Harris et al, 2017(Harris et al, , 2022. Furthermore, differences exist between scanners of the same vendor, for example, eddy currents and their impact on MRS data (Harris et al, 2022). These differences introduce the need to control for site, vendor, and/or scanner.…”

“…As with many imaging modalities, multiple sites and scanners are used to increase sample sizes and to best sample and represent the population in magnetic resonance spectroscopy (MRS) studies. However, scanner effects are substantial; both between and within vendor effects can affect MRS data ( Near et al, 2013 ; Považan et al, 2020 ; Harris et al, 2022 ). Additionally, scanner updates and upgrades occur on different timelines at different sites, further increasing the variability in measures from each scanner/vendor/site.…”

“…It is common to control for site-related variance within statistical models. Another method is to harmonize data, for example, using ComBat ( Harris et al, 2022 ). ComBat is a harmonization approach originally developed for genetic data ( Johnson et al, 2007 ), which has shown promise for removing site and scanner effects in MR imaging data, having been applied to structural imaging ( Fortin et al, 2018 ), diffusion imaging ( Fortin et al, 2017 ), and functional MRI (fMRI) ( Yu et al, 2018 ) data.…”