Functional anatomy of stereoscopic visual process assessed using functional magnetic resonance imaging and structural equation modelling
The purpose of this thesis is to study the functional anatomy of stereoscopic vision. Although many studies have investigated the physiological mechanisms by which the brain transforms the retinal disparities into three-dimensional representations, the invasive nature of the techniques available have restricted them to studies in non-human primates, whilst the research on humans has been limited to psychophysical studies. Modem non-invasive neuroimaging techniques now allow the investigation of the functional organisation of the human brain. Although PET and fMRI studies have been widely used, few researchers have explored the functional anatomy of stereoscopic vision. Most of these studies appear to be pilot work, showing inconsistency, not only in the areas sensitive to stereo disparities, but also in the specific role that each of these possesses in the perception of depth. In order to investigate the cortical regions involved in stereoscopic vision, four fMRI studies were performed using anaglyph random dot stereo grams. Our results suggest that the stereo disparity processing is widespread over a network of cortical regions which include VI, V3A, V3B and B7. Functionally, the V3A region seems to be the main processing centre of pure stereo disparities and the V3B region to be engaged in motion defined purely by spatio-temporal changes of local horizontal disparities (stereoscopic -cyclopean- motion). Interregional connectivity was investigated with two approaches. Structural Equation Modelling (SEM), as the classical technique for the analysis of effective connectivity, was used to assess one connectivity model proposed to· explain the cortical interaction observed in the first experiment. The implementation and application of this technique permitted us to identify some of its weaknesses in representing fMRI data. An extension of the SEM technique was introduced as a Non-linear Auto-Regressive Moving Average with eXogenous variables (NARMAX) approach. This can be thought of as an attempt to bring SEM towards a non-linear dynamic system modelling technique which permits a more appropriate representation of effective connectivity models using fMRI time series.