Comparison of traveling‐subject and ComBat harmonization methods for assessing structural brain characteristics

Abstract

Multisite magnetic resonance imaging (MRI) is increasingly used in clinical research and development. Measurement biases—caused by site differences in scanner/image‐acquisition protocols—negatively influence the reliability and reproducibility of image‐analysis methods. Harmonization can reduce bias and improve the reproducibility of multisite datasets. Herein, a traveling‐subject (TS) dataset including 56 T1‐weighted MRI scans of 20 healthy participants in three different MRI procedures—20, 19, and 17 subjects in Procedures 1, 2, and 3, respectively—was considered to compare the reproducibility of TS‐GLM, ComBat, and TS‐ComBat harmonization methods. The minimum participant count required for harmonization was determined, and the Cohen's d between different MRI procedures was evaluated as a measurement‐bias indicator. The measurement‐bias reduction realized with different methods was evaluated by comparing test–retest scans for 20 healthy participants. Moreover, the minimum subject count for harmonization was determined by comparing test–retest datasets. The results revealed that TS‐GLM and TS‐ComBat reduced measurement bias by up to 85 and 81.3%, respectively. Meanwhile, ComBat showed a reduction of only 59.0%. At least 6 TSs were required to harmonize data obtained from different MRI scanners, complying with the imaging protocol predetermined for multisite investigations and operated with similar scan parameters. The results indicate that TS‐based harmonization outperforms ComBat for measurement‐bias reduction and is optimal for MRI data in well‐prepared multisite investigations. One drawback is the small sample size used, potentially limiting the applicability of ComBat. Investigation on the number of subjects needed for a large‐scale study is an interesting future problem.

Effects of upgrading acquisition-techniques and harmonization methods: A multi-modal MRI study with implications for longitudinal designs

“…For the TS and TS-Combat methods, classifiers could not distinguish participants from both protocols when data from 19 to 26 participants were available for estimating the protocol bias. We also observed the limited effectiveness of the Combat method, which is consistent with a prior study [Maikusa et al, 2021]. The limited efficacy might be attributed to the specificity of the dataset used in this study.…”

“…The TS-Combat method is a recently-developed method for adjusting the site effects using TS data in the framework of the ComBat method [Maikusa et al, 2021]. The TS-Combat method replaces the row vector, zim, with the coding vector for the participant factor, xip, such that…”

“…Several harmonization methods have been proposed [Beer et al, 2020;Fortin et al, 2017;Fortin et al, 2018;Maikusa et al, 2021;Yamashita et al, 2019]. These harmonization methods might be plausible to resolve discontinuities with existing MRI data, yet no prior studies have investigated their efficacy on MRI data before and after upgrading acquisition techniques.…”

“…We also applied classification analyses to ask whether brain parameters could provide information about protocols in a multivariate manner. We then used the three most frequently utilized harmonization methods, including TS [Yamashita et al, 2019], TS-ComBat [Maikusa et al, 2021], and ComBat [Fortin et al, 2017;Fortin et al, 2018;Yu et al, 2018], to investigate whether these methods alleviate the effects of protocol upgrades and how many participants are necessary for estimating the impacts. We selected these methods because of their simplicity and applicability.…”

Neuroimaging brain growth charts: A road to mental health

“…ComBat, a technique adopted from batch-effect correction in genomics, performs great in removing unwanted inter-site variability in multi-modal MRI data, including dMRI maps (Fortin et al ., 2017 ), cortical thickness (Fortin et al ., 2018 ), as well as functional connectivity measurements (Yu et al ., 2018 ). Other attempts at reducing site effects include the improved ComBat method (Yamashita et al ., 2019 ; Maikusa et al ., 2021 ) and meta-analytic approach (Koshiyama et al ., 2020 ). With the development of new study frameworks and methodology, we can imagine a future in which highly reliable laboratory neuroimaging results are used to construct brain growth charts and guide clinical practice.…”