Semi-automatic border detection software for the quantification of arterial lumen, intima-media and adventitia layer...

The aim was to evaluate the accuracy, precision and feasibility of semi-automatic border detection software (AMS) in comparison to manual electronic calipers (EC) in the analysis of arterial images obtained with transcutaneous very-high resolution vascular ultrasound (VHRU, 25-55 MHz).

100 images from central elastic and peripheral muscular arteries were obtained on two separate imaging occasions from 10 healthy subjects, and independently measured with AMS and EC.

No bias between AMS and EC was found. The intraobserver coefficients of variation (CV) for carotid lumen dimension (mean dimension 5.60 mm) was lower with AMS compared with EC (0.4 vs. 1.9%, p = 0.033; N = 20). No consistently significant differences in intra, inter or test-retest CVs were observed overall for muscular artery dimensions between AMS and EC. The intra CV for adventitial thickness (AT, mean 0.111 mm; 15.6 vs 24.8%, p = 0.011; N = 41) and inter CV for intima-media thickness (IMT, mean 0.219 mm; 14.3 vs. 21.2%, p = 0.001; N = 58) obtained with AMS in higher quality thin muscular artery images was lower compared with EC. The mean reading time was significantly lower with AMS compared with EC (71.5 s vs. 156.6 s, p < 0.001).

AMS is accurate, precise, and feasible in the analysis of arterial images obtained with VHRU. Minor, although statistically significant, differences in the precision of AMS and EC-systems were found. The precision of AMS was superior for AT and IMT in higher quality images likely related to a decrease in the technical variability imposed by the observer.