メタアナリシスと脳画像解析

抄録

Meta–analysis plays an important role in systematic review. It allows strong evidence to be obtained through combining the results (the differences between groups) from several related research studies. The method by which results are combined in meta–analysis has been developed in the statistics framework. On the other hand, imaging analysis has been applied in many studies of brain diseases. Meta–analysis of brain imaging studies is necessary mainly because of the small sample size typically used in individual brain imaging studies because of the acquisition cost. Meta–analyses have already been published for studies of neurological diseases, such as Alzheimer's disease and Parkinson's disease. Many methods have been developed and are reviewed in this article. One of the most popular methods is coordinate–based meta–analysis (CBMA), which combines coordinates obtained as the statistical results of each brain imaging study on the standard brain template. Since the coordinate represents only one of the regions (the peak), the original spatial map needs to be reconstructed for each study. The reconstructed spatial maps represented by multiple neighborhood coordinates are combined for several studies. Inferential statistical analysis is then applied to these maps to produce the final output of the meta–analysis, using permutation tests ; this is a similar approach to that used to create the statistical parametric map (SPM) in individual studies. This procedure is called an activation likelihood estimate (ALE) and there exist similar approaches such as kernel density analysis (KDA) and signed differential mapping (SDM). This article introduces the software that allows the implementation of these techniques, and also the database for the output of the brain imaging data analysis. Further development of the CMBA methodology, such as with the bias correction used in traditional meta–analysis, would be desirable. In conclusion, meta–analysis of brain imaging analysis would be helpful to obtain more credible evidence. The methodology is already established and available for implementation.