Sjobeck Gustav R, Torbati Mahbaneh Eshaghzadeh, Minhas Davneet S, DeCarli Charles S, Wilson James D, Tudorascu Dana L
University of Pittsburgh, Department of Psychiatry, United States of America.
University of Pittsburgh, Department of Psychiatry, United States of America.
Neuroimage. 2025 Aug 15;317:121317. doi: 10.1016/j.neuroimage.2025.121317. Epub 2025 Jun 23.
As the practice of aggregating multi-site neuroimaging data has become more common, the field of neuroscience has increasingly recognized the importance of harmonization , or the removal of scanner effects from brain imaging data. While many harmonization methods exist, like ComBat and CovBat, few explicitly incorporate the network structure of the brain. Researchers studying structural connectivity are therefore not guaranteed to model the true underlying brain network. This study offers a new harmonization method, called NetBat, which was designed to incorporate network parameters from the weighted stochastic block model (WSBM) as covariates in the popular ComBat harmonization method. NetBat is demonstrated through analysis of eighteen neurotypical individuals each scanned on four MRI scanners. Results suggest that under tested circumstances NetBat provides more accurate overall harmonization and better retention of network structure than competing methods.
随着多站点神经影像数据聚合的实践变得越来越普遍，神经科学领域越来越认识到标准化（即从脑成像数据中消除扫描仪效应）的重要性。虽然存在许多标准化方法，如ComBat和CovBat，但很少有方法明确纳入大脑的网络结构。因此，研究结构连通性的研究人员无法保证对真正的潜在脑网络进行建模。本研究提供了一种新的标准化方法，称为NetBat，它旨在将加权随机块模型（WSBM）的网络参数作为协变量纳入流行的ComBat标准化方法中。通过对18名神经典型个体进行分析来验证NetBat，这些个体每人都在四台MRI扫描仪上进行了扫描。结果表明，在测试条件下，NetBat比其他竞争方法提供了更准确的整体标准化和更好的网络结构保留。
