You are here
Vittorio Iacovella: Published MLMI-paper on Brain Mapping for Longitudinal Studies
In longitudinal studies, such as those about neurocognitive rehabilitation or brain plasticity, functional MRI sessions are subsequently recorded to detect how the patterns of activation differs through time. Usually, consecutive recordings are interleaved by a period of treatment or training. The general purpose of fMRI data analysis is to compute a brain map accounting for plasticity effects on BOLD response. Specifically, the challenge of longitudinal studies analysis is to disentangle plain session effect from the plasticity one. To date, a commonly used practice is to separately compute these two effects and then combine results through a comparison. The brain map for session effect is obtained with a univariate analysis of the contrast between two subsequent recordings. We argue that such a brain map includes also the bias of the coregistration error related to the alignment of two sessions. Here we do not propose a method to reduce the coregistration error but a method to reduce the effect of this bias when estimating the session effect. We show how, combining functional hyperalignment and multivariate pattern analysis, it is possible to summarize a single brain map that accounts for plasticity effect without the session effect.