Complex motion processing mechanisms in the posterior parietal lobe: Psychophysics and computational modeling.

摘要

In the work presented here we combine human psychophysical performance with physiologically based neural models to elucidate the neural structures and computational processes underlying flow-based motion pattern processing of the visual scene. Our aim is to link human visual motion perception to putative neural structures in the medial superior temporal (MST) cortex of non-human primates that are sufficient to encode and process ecologically relevant motion patterns encountered in the visual environment.; First we quantify psychophysical thresholds to perturbations in the motion pattern structure, speed, and center-of-motion (COM) across a broad range of experimental conditions. We discuss these results within the context of local versus global motion processing mechanisms and examine the trends in performance for perceptual correlates to the motion pattern properties reported in cortex. To account for the observed performance in a subset of the psychophysical tasks, we develop physiologically constrained computational models to identify neural structures within MST sufficient to encode the visual motion tasks. Within simulated populations of MST-like units we identify a robust set of anti-preferred inhibitory structures whose computational effects on equivalent measures of perceptual performance are consistent with those of human observers. We interpret these results as suggesting that the robust processing of motion patterns associated with self-motion and optic flow is mediated by similar neural structures in the human homologue to MST.; In a second set of experiments we quantify object trajectory discrimination in the presence of planar background motions. We show that discrimination thresholds are consistently lower when objects move opposite a wide-field planar motion than when they move in the same direction. We discuss these results in the context of a specialized set of relative motion detectors within the cortex and propose additional experimental tasks to quantify interactions between the motion pattern and relative motion mechanisms proposed here.; Given the purported role of flow-based motion processing in tasks of visually guided navigation, the systematic examination of the underlying perceptual and neural motion mechanisms provides additional insights into the cortical and computational structures linking perception to action in the human visual system.