Figure 1

Binarization of the spatiotemporal data. The aim is to achieve a format suitable for computing the mutual information. (A) shows the time course x _ij (t) of a single spatial point, together with the corresponding time series of temporal differences x _ij (t + 1) - x _ij (t) for 300 images from the early-stage patterns of D. discoideum. The temporal differences are inserted into a binarization filter δ _ij (t) = Sig(x _ij (t + 1) - x _ij (t)). If δ _ij (t) is zero, the value at the previous time step is used instead. For convenience one then passes to values 1 and 0, rather than 1 and -1. The lower row in (A) shows the resulting binarizations as bar code diagram. As described in the main text, variant (a) of the mutual information can be computed from this sequence of zeroes and ones. For variant (b) the binary patterns of two adjacent spatial points are considered (B). Variant (c) compares each spatial point with an arbitrarily selected reference point (C).