Changepoint estimation in multi-subject fMRI studies

Most statistical analyses of fMRI data assume that the nature, timing and duration of the psychological processes being studied are known. However, in many areas of psychological inquiry, it is hard to specify this information a priori. Examples include studies of drug uptake, emotional states or experiments with a sustained stimulus. In this paper we assume that the timing of a subject's activation onset and duration are random variables drawn from unknown population distributions. We propose a technique for estimating these distributions assuming no functional form, and allowing for the possibility that some subjects may show no response. We illustrate how these distributions can be used to approximate the probability that a voxel/region is activated as a function of time. Further a procedure is discussed that uses a hidden Markov random field model to cluster voxels based on characteristics of their onset, duration, and anatomical location. These methods are applied to an fMRI study (n=24) of state anxiety, and are well suited for investigating individual differences in state-related changes in fMRI activity and other measures.