Automated Fiber Diameter and Porosity Measurements of Plasma Clots in Scanning Electron Microscopy Images

Abstract

Scanning Electron Microscopy (SEM) is a powerful, high-resolution imaging technique widely used to analyze the structure of fibrin networks. Currently, structural features, such as fiber diameter, length, density, and porosity, are mostly analyzed manually, which is tedious and may introduce user bias. A reliable, automated structural image analysis method would mitigate these drawbacks. We evaluated the performance of DiameterJ (an ImageJ plug-in) for analyzing fibrin fiber diameter by comparing automated DiameterJ outputs with manual diameter measurements in four SEM data sets with different imaging parameters. We also investigated correlations between biophysical fibrin clot properties and diameter, and between clot permeability and DiameterJ-determined clot porosity. Several of the 24 DiameterJ algorithms returned diameter values that highly correlated with and closely matched the values of the manual measurements. However, optimal performance was dependent on the pixel size of the images-best results were obtained for images with a pixel size of 8-10 nm (13-16 pixels/fiber). Larger or smaller pixels resulted in an over- or underestimation of diameter values, respectively. The correlation between clot permeability and DiameterJ-determined clot porosity was modest, likely because it is difficult to establish the correct image depth of field in this analysis. In conclusion, several DiameterJ algorithms (M6, M5, T3) perform well for diameter determination from SEM images, given the appropriate imaging conditions (13-16 pixels/fiber). Determining fibrin clot porosity via DiameterJ is challenging.

Keywords: DiameterJ; automated analysis; diameter; fibrin fibers; plasma clots; porosity; scanning electron microscopy; structure.

Figure 1

Manual and DiameterJ analysis of three large data sets. SEM images and patterns for manual analysis are shown for data set 1 (A), set 2 (D), and set 3 (G). The line pattern in each image was used to select 100, 54, and 100 random fibers per image for manual diameter measurements. Representative DiameterJ binary images are shown for set 1 (B), set 2 (E), and set 3 (H). Frequency distributions for manual and DiameterJ diameter measurements are shown for data set 1 (C), set 2 (F), and set 3 (I). The insets in (A,D,G) show the varying pixel size across sets; 8.3 nm in set 1, 4.9 nm in set 2 and 24.3 nm in set 3. Scale bars in (A,D,G) are 2 μm.

Figure 2

DiameterJ measurements (arithmetic mean, algorithms T1, T2, T6, M5, M7) and manual measurements for five healthy control samples and ten postpartum hemorrhage (PPH) patient samples. (A) Manual measurements, and (B) automated DiameterJ measurements of healthy control and PPH patient samples. (C,D) Comparison between manual measurements and automated DiameterJ measurements of healthy control samples (C), and PPH patients (D). No statistically significant difference in fiber diameter between healthy control samples and PPH samples was observed.

Figure 3

SEM analysis of plasma clots from Systematic Lupus Erythematosus (SLE) patients. SEM images of plasma clots from active SLE patients (A), inactive SLE patients (B), and healthy controls (C). (D) Comparison between manual and DiameterJ algorithms measurements of fibrin fiber diameters in active and inactive SLE, and healthy controls. In both manual and the three automated measurements, active SLE samples had a larger diameter than the inactive SLE samples and the healthy controls.