Ultrasound (US) is the recommended image modality for small abdominal aortic aneurysm (AAA) surveillance. Small AAAs are scanned at regular intervals until the US anterior-posterior (AP) diameter reaches the relevant threshold for repair (women/men: 50/55 mm). However, conventional two-dimensional US (2D-US) is limited to the AP direction which is operator dependent. Also, 2D-US relies on the imaging plane being perpendicular to the AAA centerline. A fully-automatic three-dimensional US system (3D-US) allows for on-cart generation of a AAA 3D-mesh and real-time software-assisted diameter estimation perpendicular to the centerline in any direction (3D-SAUS) (AAA model software, Philips). 3D-SAUS is based on a single sweep with a matrix 3D-US transducer (X6-1 Matrix transducer on an Epiq-7 US system, Philips). We hypothesised that 3D-SAUS produces more precise diameter measurements than conventional 2D-US. Thus, this study aims to test the variability and intermodality agreement of AAA diameter measurements by comparing conventional 2D-US with 3D-SAUS in a clinical setting among sonographers with varying experience levels.

Nine patients under AAA surveillance were included. The patients were examined by two groups of US operators: 13 experts and 16 novices. The experts were sonographers and physicians that had performed more than 300 AAA examinations within the last two years. The novices were medical students and junior surgeons with no prior US experience, defined as less than five unsupervised AAA US scans. Prior to the data collection, all novices participated in a focused AAA US course, including 45 minutes of theory lesson and 45 minutes of hands-on training on healthy models. All patients had their AAA AP-diameter assessed by experts and novices using conventional 2D-US and 3D-SAUS. The acquisition process was monitored by members of the research team to ensure that all operators performed, annotated, and saved both types of scans, but without the monitor interfering with any equipment settings or scan methodology. No manual corrections were made to the 3D-meshes generated by the 3D-SAUS technique. Bland-Altman plots for 2D-US vs. 3D-SAUS were created for experts and for novices to assess the level of agreement dependent on experience. AAA diameter means and standard deviations were calculated to compare the diameter variance of novices using 2D-US with novices using 3D-SAUS, and to compare experts using 2D-US with novices using 3D-SAUS.

In total, 460 US examinations were performed, of which 50 pairs of 2D- and 3D-scans were excluded due to missing one or both scans, leaving 410 scans for further analysis. The average difference between 3D-SAUS and 2D-US was 0.4 mm among experts and 1.3 mm among novices (P = 0.018), with limits of agreement of ±4.0 mm for experts and ±6.8 mm for novices (see Figure 1). In the group of novices, the diameter variance was reduced in 6/9 patients with the use of 3D-SAUS compared to 2D-US. In the remaining 3/9 patients, the standard deviations were increased by 0.1-0.4 mm with the use of 3D-SAUS (see Table 1a and Figure 2). When comparing experts using 2D-US with novices using 3D-SAUS, the novices' variations remained higher than the experts' in 7/9 patients. However in 6/9 patients, the novices' standard deviation was less than 1 mm bigger than the experts' standard deviation (see Table 1b and Figure 2).

When novices measure AAA maximal diameters, the 3D-SAUS tool tends to reduce the variances compared to conventional 2D-US. In addition, the novices' AAA diameter variances came closer to "the diagnostic standard" in AAA US diameter (i.e. experts using 2D-US) when using 3D-SAUS. In the hands of experts, the agreement of the maximal AAA diameter when using 3D-SAUS and conventional 2D-US was good, thus supporting that 3D-SAUS can aid in clinical decision-making, even among experts.