Comparison of Automated vs Manual Analysis Programs for Quantification of Corneal Nerve Morphology in Patients with or without Type 2 Diabetes Mellitus

OBJECTIVES: To assess the agreement between the automated analysis program and a manual program for quantification of corneal nerve morphology.
METHODS: Twenty-seven non-diabetic controls (mean age: 48.6±5.9 years) and 60 subjects with diabetes (mean age: 52.1±6.5 years) were enrolled. Corneal nerve fiber density (CNFD), branch density (CNBD), and fiber length (CNFL) were quantified by the manual (CCMetrics software, University of Manchester, UK) and automated program (ACCMetrics software, University of Manchester, UK). Bland-Altman plots were generated to assess agreement between the two methods.
RESULTS: There were no differences in gender, age, total cholesterol, and LDL between the two groups, whereas BMI, HbA1c, and triglyceride were significantly higher and HDL was significantly lower in the T2DM group. CNFL was overestimated in the diabetic group and CNFD was underestimated in both groups with ACCMetrics (p=0.001, p<0.001, respectively). The Bland-Altman plots for both groups demonstrated good agreement for CNFL, with a wide limits of agreement (LoA) of 95% for CNFD and CNBD.
DISCUSSION AND CONCLUSION: Manual and fully automated protocols for sub-basal nerve evaluation had lower agreement in the measurement of CNFD and CNBD than CNFL in healthy controls and subjects with diabetes.