Fractal Analysis of MCF10A and CA1d Cell Lines. (A) Two simulated test patterns of known Df, a "perfect" 2-D circle and the "Koch snowflake", were used to validate the FracLac program, and to put our data into context within these controls. These patterns were normalized by pre-processing in the same manner as the contours of nests prior to fractal analysis, and resulted in measured values of 1.08 and 1.28, respectively. (B) Representative time-lapse images (0, 10, 22, 36 h) of MCF10A and CA1d cells taken in the absence or presence of 25% or 50% Matrigel overlay were thresholded to normalize background levels, and outlines were systematically applied to borders using a "define contours" function. Df measurements were then generated using the FracLac plugin. (C) All values are presented as the mean ± standard deviation of the fractal value for each nest captured at each time point of interest. Since NEA is highly standardized, all nests across both cell lines, and all treatments, had similar fractal measurements at 0 h. However, at all other time points measured after 0 h, the aggressive, invasive CA1d nests exhibited significantly greater Df measurements than the noninvasive MCF10A cell line (N = 8 per group; P < 0.01 in all cases), regardless of the microenvironment. Slopes calculated for lines fit to time-course data for CA1d Df values were approximately 2-fold greater than slopes for MCF10A nests, for all experimental conditions measured.