Image processing toolbox

Time, number of cycles, axial load and displacement were recorded using the machine transducers with a frequency of 64 Hz. The axial stiffness was calculated from the load–displacement curves of cycles 10 to 19. Therefore, minimal and maximal force, as well as displacement values have been used. For further evaluation, the mean value was used. Every 250 cycles, an x-ray in anterior–posterior view, including a reference sphere for scaling, was taken at the base load of 100 N. X-ray evaluation was performed using a custom-made software routine (Matlab 7.9 R2009b, Image processing Toolbox, The MathWorks GmbH, Ismaning, Germany). The number of cycles until 5°varus collapse compared with the initial x-ray was identified for all specimens and defined as number of cycles to failure [8 (link)]. Specimens were tested until catastrophic implant or construct failure.
After testing normal distribution of the data (Shapiro–Wilk test), the Wilcoxon signed-rank test was carried out to identify differences between study groups regarding axial stiffness and cycles to failure. The software package SPSS 24.0 (IBMN, IBM SPSS Statistics, Version 24.0, New York, US) was used for all statistical evaluations. The level of significance was set to alpha = 0.05.

Z Stack images of fibrin matrices were visualized and quantified using multiple feature detection techniques of the Image Processing Toolbox in MATLAB (MathWorks).⁵¹ Geometrical tubular elements were identified utilizing the fibermetric function, which employed multiscale second-order local structures (a vesselness measure) of eigenvalues from an Hessian pixilated matrix to detect tubular structures.⁵² This function was used to identify tubular structures and was also used to filter out areas of dense fibrin fiber overlap. A mean filter function was used to identify densely populated fiber areas within each original z stack slice. These 2 preprocessing techniques (Figure 1) were combined into a single preprocessed image and used in subsequent segmentation and image-processing methods. Specifically, houghpeaks (Hough transform) and houghlines (Hough line) functions were applied to generate Cartesian line segments through the center of tubular structures to detect fibrin fibers.^{46 (link),52 ,53 (link)} Overall, fibrin fiber overlap and fibrin fiber length were calculated after Hough line generation, while fiber matrix porosity and fractal dimension were computed from preprocessed 2-D slices prior to segmentation.