Within-subject analysis in event-related functional Magnetic Resonance Imaging (fMRI) first relies on (i) a detection step to localize which parts of the brain are activated by a given stimulus type, and second on (ii) an estimation step to recover the temporal dynamics of the brain response. Recently, a Bayesian detection-estimation approach that jointly addresses (i)-(ii) has been proposed in [1]. This work is based on an independent mixture model (IMM) and provides both a spatial activity map and an estimate of brain dynamics. In [2], we accounted for spatial correlation using a spatial mixture model (SMM) based on a binary Markov random field. Here, we assess the SMM robustness and flexibility on simulations which diverge from the priors and the generative BOLD model and further extend comparison between SMM and IMM on real fMRI data, focusing on a region of interest in the auditory cortex.