Comparison of 2D fiber network orientation measurement methods

Abstract

The mechanical properties of tissues, tissue analogs, and biomaterials are dependent on their underlying microstructure. As such, many mechanical models incorporate some aspect of microstructure, but a robust protocol for characterizing fiber architecture remains a challenge. A number of image-based methods, including mean intercept length (MIL), line fraction deviation (LFD), and Fourier transform methods (FTM), have been applied to microstructural images to describe material heterogeneity and orientation, but a performance comparison, particularly for fiber networks, has not been conducted. In this study, we constructed 40 two-dimensional test images composed of simulated fiber networks varying in fiber number, orientation, and anisotropy index. We assessed the accuracy of each method in measuring principal direction (theta) and anisotropy index (alpha). FTM proved to be the superior method because it was more reliable in measurement accuracy (Deltatheta = 2.95 degrees +/- 6.72 degrees , Deltaalpha = 0.03 +/- 0.02), faster in execution time, and flexible in its application. MIL (Deltatheta = 6.23 degrees +/- 10.68 degrees , Deltaalpha = 0.08 +/- 0.06) was not significantly less accurate than FTM but was much slower. LFD (Deltatheta = 9.97 degrees +/- 11.82 degrees , Deltaalpha = 0.24 +/- 0.13) consistently underperformed. FTM results agreed qualitatively with fibrin gel SEM micrographs, suggesting that FTM can be used to obtain image-based statistical measurements of microstructure.