FIGURE 1.

An input for a developed tool is time-lapse sequences collected from four fields of view of each well of a multi-well plate (A). The software processes consequently each input grayscale image and applies image correction procedure where noise filtering and histogram reassignment are performed (B). Then, the script applies a threshold on the filtered grayscale image to convert it into a binary form for correction and gap detection procedure (C). The obtained binary masks are then collected into a sequence of wound contours and area counts (D). The results of the contour segmentation accuracy were compared with manually drawn overlays. The difference between manually outlined and segmented by algorithm gaps was, on average, ∼2.09% and was set as error measure (E). Detailed examination of the obtained gap contour overlays reveals that there are groups of cells moving faster than the neighbor lattice for several cell lines. This event occurs after 6–12 h of scratch (F, red arrows). Such cells start to detach from the cell lattice and may reenter back. It was observed more frequently for HT1080 (enlargement E) and 3T3 cells. This process is relatively rare in the case of A549, HaCaT, and MCF-7 cells. However, scripted segmentation includes the area of such cells to the calculated result (F, white arrows).