Automated analysis of corneal nerve tortuosity in diabetes: implications for neuropathy detection

There is potential benefit in analysing corneal nerve tortuosity as a marker for assessment and progression of systemic diabetic neuropathy.

The aim of this work was to determine whether tortuosity significantly differs in participants with type 1 (T1DM) and type 2 (T2DM) diabetes compared to controls and whether tortuosity differed according to neuropathy status.

Corneal nerves of 164 participants were assessed across T1DM, T2DM and control groups. Images of corneal nerves were captured via in vivo corneal confocal microscopy. Diabetic neuropathy status was examined using the Total Neuropathy Score (TNS). Tortuosity was assessed with Cfibre v0.097. Results were compared between groups with a linear mixed model accounting for location of image and controlling for age, producing Tortuosity Factor (TF), an estimate of the marginal means of each group.

Tortuosity was significantly reduced in the T1DM group compared to controls (TF = 0.241, 95%CI = 0.225-0.257 vs. TF = 0.272, 95%CI = 0.252-0.292; mean difference = -0.031, p = 0.02) and in the T2DM group compared to controls (TF = 0.261, 95%CI = 0.244-0.278 vs. TF = 0.289, 95%CI = 0.270-0.308; mean difference = -0.029, p = 0.03). Tortuosity did not significantly differ between participants with T1DM and T2DM accounting for age and TNS (TF = 0.240, 95%CI = 0.215-0.265 vs. 0.269, 95%CI = 0.244-0.293, mean difference = -0.029, p = 0.11). Tortuosity was significantly reduced in participants with neuropathy (TNS≥2) compared to participants with no neuropathy (TNS< 2) (TF = 0.248, 95%CI = 0.231-0.265 vs. TF = 0.272, 95%CI = 0.260-0.283; mean difference = -0.024, p = 0.03).

Tortuosity is significantly reduced in participants with T1DM and T2DM compared to age matched controls and in participants with neuropathy compared to those without neuropathy.