What is the correct distance measurement metric when measuring carotid ultrasound intimamedia thickness automatically?

The aims of this study are: (a) analyze the inter-observer carotid intima-media thickness (CIMT) variability using three different measurement metrics on large multi-institutional databases; (b) evaluate the three kinds of metrics when comparing completely automated CIMT measurement (Auto Edge CIMT) to two manually derived CIMT (manual CIMT). Two expert sonographers manually analyzed 665 carotid B-Mode ultrasound images collected from five Institutions and using four different scanners. The two readers traced the lumen-intima (LI) and media-adventitia (MA) interfaces. The manual CIMT was computed from the LI/MA tracings by using three different distance measurement metrics: the Hausdorff, the PolyLine, and the Centerline distance metrics. The LI/MA tracings of a completely automated method we previously developed were then compared to manual CIMT. The average CIMT values of Readers 1 and 2 were 1.904 ± 0.650 mm and 1.421 ± 0.394 using Hausdorff, 0.808 ± 0.269 mm and 0.790 ± 0.227 mm using Polyline, and 0.762 ± 0.266 mm and 0.782 ± 0.228 mm using Centerline, respectively. The correlation coefficients were 0.14 (0.07-0.22) for Hausdorff, 0.77 (0.74-0.80) for Polyline, and 0.82 (0.79-0.84) for Centerline. The variation coefficients (CV) were equal to 46.4% (Hausdorff), 2.6 % (Polyline), and 14.1% (Centerline). The Auto Edge CIMT values were: 1.655 ± 0.676 mm using Hausdorff, 0.808 ± 0.282 mm using Polyline, and 0.776 ± 0.275 mm using Centerline. In conclusion, Centerline and Polyline yield very close results and are clinically suitable distance measurement techniques for computing the CIMT from LI/MA profiles.