Abstract
Brain imaging studies for attention-deficit/hyperactivity disorder (ADHD) have not always yielded consistent findings, potentially owing to measurement bias in magnetic resonance imaging (MRI) scanners. This study aimed to elucidate the structural brain characteristics in children with ADHD by addressing measurement bias in multi-site MRI data using the harmonization method, traveling-subject (TS) approach. The MRI data of 14 traveling subjects, 178 typically developing (TD) children, and 116 children with ADHD were collected from multiple sites. The TS method and ComBat were used to correct for measurement bias. Gray matter volumes were estimated using FreeSurfer, and the ADHD and TD groups were compared using mixed-effect models. Compared to raw data, the TS method significantly reduced measurement bias while maintaining sampling bias. In contrast, ComBat effectively reduced measurement bias but also significantly decreased sampling bias. TS-corrected data showed decreased brain volumes in the frontotemporal regions in the ADHD group compared to the TD group. Specifically, significant volumetric reductions were found in the right middle temporal gyrus in children with ADHD (TS-corrected data: β = -0.255, FDR [family discovery rate] p = 0.001). These results demonstrate that the TS method effectively reduces measurement bias across MRI scanners, ensuring reliable findings in multi-site studies. The observed frontotemporal volume reductions in ADHD, especially in the right middle temporal gyrus, highlight the reliability of findings obtained with TS correction.