High-resolution ultrasound of peripheral nerves in systemic sclerosis: a pilot study of computer-aided quantitative assessment of nerve density.

Abstract

To quantitatively evaluate and compare nerve density in patients with limited cutaneous systemic sclerosis (lcSSc) and control subjects using high-resolution ultrasound (US) with a computer-aided assessment. Forty patients and 40 age- and sex-matched control subjects were prospectively enrolled. Ultrasound (US) examination (17-5 MHz probe) of the median nerve at the elbow was performed bilaterally by one radiologist. A software quantified the ratio between the hypoechoic and hyperechoic areas of peripheral nerves on ultrasound. Two observers set the threshold in the images acquired, and three observers performed the digital analysis of nerve density. Statistical analysis included Mann-Whitney U-test of patients versus control subjects and subgroup analysis of symptomatic and non-symptomatic patients. Intra and inter-observer agreement of the three observers were assessed with the kappa statistic. In all, 160 median nerves were evaluated. According to the US, nerve density was significantly reduced in lcSSc patients compared to control subjects (mean and standard deviation: 41 ± 3 vs 56 ± 4, p < 0.01). Subgroup analysis showed that symptomatic patients (n = 15) had reduced nerve density compared to non-symptomatic (n = 25) patients (39 ± 5 vs 43 ± 4, p < 0.01). Intra-observer agreement was very good (K = 0.82). Inter-observer agreements were good: reader 1 vs reader 2: k = 0.78 (95% confidence interval 0.65 to 0.91); reader 2 vs reader 3: k = 0.72 (95% confidence interval 0.65 to 0.82); reader 3 vs reader 1: k = 0.71 (95% confidence interval 0.64-0.81). In lcSSc patients, nerve density was reduced, especially in the symptomatic group, compared to control subjects.