Coordinate based meta-analysis: new clustering algorithm, and inclusion of region of interest studies

Abstract

There are many methods of conducting coordinate based meta-analysis (CBMA) of neuroimaging studies that have tested a common hypothesis. Results are always clusters indicating anatomical regions that are significantly related to the hypothesis. There are limitations such as most methods necessitating the use of conservative family wise error control scheme and the inability to analyse region of interest (ROI) studies, which can be overcome by cluster-wise, rather than voxel-wise, analysis. The false discovery rate error control scheme is a less conservative option suitable for cluster-wise analysis and has the advantage that an easily interpretable error rate is estimated. Furthermore, cluster-wise analysis makes it possible to analyse ROI studies, expanding the pool of data sources. Here a new clustering algorithm for coordinate based analyses is detailed, along with implementation details for ROI studies.