Modello spaziale bayesiano con fattori latenti per dati di meta-analisi basate su coordinate

Theory of mind (ToM) is a skill that refers to the ability to understand the mental states of others. In the neuroimaging literature, a large number of studies have been conducted to investigate the underlying neural substrates of ToM. In this work, we propose a Bayesian point process hierarchical model to find specific ToM task-related differences in patterns of neural activation. Specifically, we apply a spatial Bayesian latent factor regression for coordinate-based meta-analysis data to a real-world dataset in order to compare different types of mentalising inference (cognitive versus affective). The foci from each study are modeled as a Cox process, where the logarithm of the intensity function is represented by means of a linear combination of a high-dimensional basis set. A sparse representation of the intensities is guaranteed through latent factor modeling of the basis coefficients. Our methodology can be used to identify brain areas of consistent activation and to perform reverse inference, that is, to build a predictive model of task type for new studies. It is also possible to incorporate covariate information when this is available.