Fig. 4

Validation of quickPIV on non-segmentable data sets. a PIV analyses were performed on the actin signal of a double hemizygous transgenic embryo before (top) and during (bottom) gastrulation. For each time point, the two consecutive volumes analyzed with PIV are shown in red and blue, next to the computed vector fields after similarity-selective spatial averaging. b PIV was also performed for the same time points on the nuclear signal, and the resulting similarity-selective averaged vector fields are shown next to the actin vector fields. c The orientation similarity between each pair of vectors in the two channels is computed through their normalized dot product. The Euclidean error between each pair of vectors is computed as well to measure the combined magnitude and direction differences between the vectors. The scatter plot of these two quantities shows that most vectors are clustered around a region of high normalized dot product and low euclidean error, indicating good agreement between the vector fields in (a) and (b). d Three patterns of cell migration can be distinguished in the T. castaneum data set (i): Segmentable and trackable (S/T), segmentable and non-trackable (S/NT) and non-segmentable and non-trackable (NS/NT) nuclei. The serosa consists of segmentable nuclei. While some regions are easily trackable, in others it is difficult to establish unambiguous correspondences of the nuclei between the two time points. High cell densities render nuclei in the gastrulating embryo non-segmentable, and therefore non-trackable. e Three-dimensional mapping of the height of the maximum peak of NSQECC at each interrogation area during the PIV analysis of the two volumes in (d). High values are achieved both in the segmentable and trackable and non-segmentable regions of the embryo, indicating that the interrogation and search patterns in these regions are well approximated by a translation and high PIV accuracies are expected