Histomorphometry in Peripheral Nerve Regeneration: Experimental Comparison of Different Axon Counting Methods

Abstract

PURPOSE: Histomorphometry is a common tool to quantitatively evaluate outcomes of peripheral nerve regeneration by measuring axonal parameters such as count, diameter, and myelination. Manual measurement across the entire nerve cross-section is the most accurate technique of histomorphometric analysis but is extremely labor intensive. Thus, most researchers have opted for analysis of just a sample of the cross-section. Currently, automated sampled analysis is the most common method. However, no study has been performed to compare the accuracy of these techniques, making it difficult to compare results across the literature. METHODS: Rat sciatic nerves were transected and repaired with a 20 mm nerve isograft. A total of 34 sections were stained with toluidine blue and digitized: 24 from native nerve distal to the graft, 5 from nerve proximal to the graft, and 5 from the graft itself. Three blinded researchers manually counted total myelinated fibers in each full cross-section. Total myelinated fiber counts were also extrapolated from sampled fields representing 20% of the cross-sectional area. Sampled counts were performed both manually and automatically with the software ImageJ (National Institutes of Health). Myelinated fiber diameter (FD), axon diameter (AD), and myelin sheath thickness (MTh) were measured manually in the full cross-sections and the sampled fields. Repeated measures MANOVA, Spearman’s correlation, and Wilcoxon signed-rank tests were performed. RESULTS: Results are expressed in mean ± standard deviation. Full manual axon count was 12,504 ± 4,195 overall and 13,506 ± 4,217 distally. Sampled axon counts were significantly higher than full manual, especially distally (sampled manual: 15,316 ± 4,613, P < 0.001; sampled automated: 16,297 ± 7,733, P = 0.037). All three methods showed strong, significant correlation with each other, especially for distal sections (full manual and sampled manual: rs = 0.912, P < 0.001; full manual and sampled automated: rs = 0.599, P = 0.002; sampled manual and sampled automated: rs = 0.708, P < 0.001). Overall full manual fiber diameter, axon diameter, and myelin sheath thickness were 5.42 ± 1.18 µm, 3.61 ± 0.78 µm, and 0.93 ± 0.28 µm, respectively. They did not differ from sampled measurements (P = 0.144, P = 0.059, and P = 0.817, respectively). CONCLUSIONS: Manual count of sampled nerve sections produces highly correlated, reliable results when using standardized and systematic sampling methods. Automated data should be regarded with more caution as correlation was not as high. However, the significant correlation indicates it is an acceptable technique if manual sampled analysis is not possible. Although the methods explored in this study correlated highly with one another, the accuracy of both sampled techniques differed significantly from full manual analysis. Therefore, comparison of numerical results between papers reporting these parameters (ie, for a metanalysis) may not be possible. Researchers must be cognizant of the wide variety of techniques reported in the literature and exercise caution when comparing data between studies.