Extended Data Fig. 4: Flowchart for gap-closing scheme.

a, A matrix of differences between all new tracks’ first frame number and all lost tracks’ last frame number. b, The frame difference matrix is thresholded based a maximum search time threshold of 15 seconds to create a mask. c, The mask is applied to the matrix of the intensity weighted centroid distances between the new tracks’ first centroid position and last tracks’ last centroid position. d, The distance matrix is thresholded based on the maximum velocity threshold of 1 μm/s. e, Any gap closure resulting in a travel angle coefficient of variation under 0.2 is removed. f, The final masked gap closing matrix is produced, which is then globally minimized to assign new tracks with possible lost track candidates. g, The gap-closing scheme is repeated in its entirety until the total number of tracks stabilizes. In this example, it took 4 iterations to stabilize to 469 tracks.