Activation Detection and Characterisation in Brain fMRI Sequences. Application to the study of monkey vision.

In this report, we propose a number of new ways of detecting activations in fMRI sequences that require a minimum of hypotheses and avoid any pre-model- ling of the expected signal. In particular, we try to avoid as much as possible linear models. The sensitivity of the methods with respect to signal autocorrelation is investigated, in order to reduce or control it. Considering a experimental block design, a key point is the ability of taking into account transitions between different signal levels, but still without the use of predefined impulse response. The methods that we propose are based on well-known Anova and information theoretical models. The problem of statistical test validation is also studied and partly solved. The power of these methods seems high enough to avoid any smoothing, spatial or temporal, of the data. Once an activation map is obtained, we attempt to characterise activations of the block experiment by studying the pre- and post- activation transitions. This more descriptive part of our work can be continued by searching for brain areas with homogenous characteristics, for example similar impulse responses. Quite naturally, this problem can be formulated as a clustering problem, which we solve through the use of a \textitfuzzy C-means algorithm. This part of the analysis is performed without spatial or anatomical constraints, in order to allow for the observation of unexpected phenomena. A first application is presented on a sequence of visual tasks obtained at Leuven University in order to characterise monkey motion perception. We propose activation maps, and, as a first step towards a spatio-temporal model of the brain, a map of impulse response patterns.