|Ph.D Thesis||Department of Biomedical Engineering|
|Supervisors:||Assoc. Prof. Gur Moshe|
|Prof. Pratt Hillel|
Linking neuronal activities to sensory perception is one of the most challenging problems in neuroscience. In visual object recognition tasks, it is of interest to understand the neuronal machinery underlying object and face recognition. Although various attributes of this issue have been extensively studied, very little is known about the correlation and the general relationship between behavioral performance and the physiological elements deriving this specific behavior. Efforts to understand visual perception and object recognition have produced several models. The dominant model can be described as a connectionist model where information flows only upward in a convergent manner. This is the “bottom-up” - hierarchical model where recognition occurs by high-level “expert” cells, located at the pinnacle of the hierarchy (IT /LO area). While there has been a growing amount of experimental evidence for reciprocal and top down effects, none of the experiments have provided results that can not be explained by the bottom-up model. Here, a distinguishing paradigm was design to determine whether feed-forward connectivity or a multi-directional flow of information is the more likely explanation. Comparing monkeys’ electrophysiological results to human performance was, until recently, the main tool to investigate the relationship between perception and neuronal activity. Recent advances in functional imaging techniques, made it possible to study new issues that could not have been explored before. The complementary advantages and disadvantages of available functional neuro-imaging techniques have driven efforts to combine multiple imaging modalities. In this study, we compared fMRI acquisitions with ERP studies and source current density estimation. Correlations between EEG and BOLD signals, even from the same cortical patch, were studied. We looked at the minimum object size that is perceivable and at the relationship between this size and the effective stimulus size of single cell responses. We showed that stimuli smaller than physiological threshold evoke robust responses in ventral stream visual areas. These responses indicate top-down interactions that are incompatible with the classical, feed-forward model. The two imaging techniques were used serially. Spatial relationship between the different measurements was defined and spatial equivalence was achieved. Integrated visualization based on multiple co-registered data sets of the different size face stimuli, implies that the two functional imaging modalities, although reflecting completely different aspects of neural activity, show a high level of correlation. We suggest, however, that under the present spatio-temporal resolution of fMRI and current source localization, special care should be used when analyzing the co-registered data.