Fig. 9

(a) Contour plots of neurons that were detected to be active in all six imaging sessions overlaid on the correlation image of the sixth imaging session. Each color corresponds to a different session. (b) Stability of multiday registration method. Comparisons of forward and backward registrations in terms of F1 scores for all possible subsets of sessions. The comparisons agree to a very high level, indicating the stability of the proposed approach. (c) Comparison (in terms of F1 score) of pair-wise alignments using readouts from the union vs direct alignment. The comparison is performed for both the forward and the backwards alignment. For all pairs of sessions the alignment using the proposed method gives very similar results compared to direct pairwise alignment. (d) Magnified version of the tracked neurons corresponding to the squares marked in panel (a). Neurons in different parts of the FOV exhibit different shift patterns over the course of multiple days, but can nevertheless be tracked accurately by the proposed multiday registration method.