Abstract (EN)

This paper introduces a mathematical model of the spatio-temporal patterns of visually evoked activity observed using Voltage-Sensitive Dye Imaging (VSDI) of the visual cortex. The cortical activity is described using a linear superposition of waves traveling with different speeds. This model improves the quality of the wave detection and still respects the previous approaches, as it integrates several biologically plausible constraints: 1) separability of the sources in terms of cortical location; 2) separability of the waves in terms of propagation speed, and 3) additivity of the depolarizing effects of the waves. Under these assumptions, a traveling component analysis algorithm performs a full separation of the set of waves and recovers the locations of the neural sources. Both features could help to better understand the dynamics of evoked activity in cortical sensory networks.