A Bayesian hierarchical spatial point process model for multi-type neuroimaging meta-analysis
Citation: Kang, J., Nichols, T. E., Wager, T. D., & Johnson, T. D. (2014). A Bayesian hierarchical spatial point process model for multi-type neuroimaging meta-analysis. The annals of applied statistics, 8(3), 1800.
There is growing interest in neuroimaging meta analysis, an important tool for synthesizing the ever-expanding brain mapping literature that is largely based on samples of 20 or fewer subjects. Neuroimaging meta analysis identifies consistent activation regions by using peak activation coordinates (foci) that are collected from different independently performed studies. Kang et al. (2011) proposed a fully parametric spatial Bayesian model that provides richer results
than other methods by, for example, modeling inter-study variation in activation location. However, that method only models one population of studies with a single-type point process and is sensitive to prior specifications that are based on expert opinion. To address these limitations, in this work, we adopt a non-parametric Bayesian approach for meta analysis data from multiple classes or types of studies. In particular, foci from each type of study are modeled as a
cluster process driven by a random intensity function that is modeled as a kernel convolution of a gamma random field. The type-specific gamma random fields are linked and modeled as a realization of a common gamma random field shared by all types, inducing correlation between study types and mimicking the behavior of a univariate mixed effects model. We illustrate our model on simulation studies and a meta analysis of five emotions from 219 studies. In addition,
we show how to use the model to predict the study type for a newly presented study. We evaluate the performance of our methods via leave-one-out cross validation, which are efficiently implemented using importance sampling techniques.