Abstract: PURPOSE: Distinguishing velocity is critical for animals’ survival and human life. The visual dorsal pathway involves the perception of velocity. Although various studies about velocity have been reported, we know a rough speed range for each area respectively. How each brain area processes various velocities from low to high speed is still unclear. Specifically, how the higher-level brain areas process velocity when the lower-level areas lost direction selectivity at high speed. Our main hypothesis was that the lower-level areas could provide sequential retinotopic activations to the higher-level areas to process velocity at high speed.
METHODS: To get a whole picture of velocity processing for visual brain areas across speeds, we recorded single units in V1, MT, and MST in alert macaques for dots, gratings, and plaids. Then we built a receptive field model based on retinotopic information processing to simulate neuronal responses.
RESULTS: 1) The optimal speed and the cutoff speed of direction selectivity increased hierarchically along V1, MT, and MST, no matter the cell subtype (component, pattern, or unclassified), across increasing speed. 2) The differences in latencies among V1, MT, and MST suggested the hierarchically bottom-up processing for velocity. 3) Model results were consistent with the physiological observations.
CONCLUSIONS: These results supported our hypotheses that low-level areas provide retinotopic signals after losing direction selectivity for assisting high-level areas detect motion direction. This study may promote the understanding of velocity processing in the brain and the development of computer vision in artificial intelligence (AI).

Key words: primate, dorsal pathway, velocity processing, receptive field model