The reproducibility of measuring maximum abdominal aortic aneurysm diameter from ultrasound images

Participants were recruited from The Department of Vascular and Endovascular Surgery at Townsville University Hospital [19]. All participants provided written informed consent for inclusion and ethics approval was obtained from The Townsville Hospital and Health Service Human Research Ethics Committee (13/QPCH/16). A convenience sample size of 50 participants was used in concordance with previous similar studies [16, 17, 20]. Participants met the following inclusion criteria: (1) a small infra-renal AAA measuring 30–55 mm in maximal diameter and (2) an ultrasound performed by one of three experienced vascular sonographers within the Townsville University Hospital Vascular Laboratory using a standardised protocol. Patients with a history of previous abdominal aortic surgery were excluded. At recruitment an interview and physical examination was conducted to collect relevant medical history and clinical measurements. Ischaemic heart disease, hypertension and diabetes were defined as a previous diagnosis or treatment of these conditions by a qualified medical physician. Stroke was defined as a documented history of ischaemic or haemorrhagic stroke. The presence of aneurysms at other sites was determined through documented history or relevant medical imaging. Brachial blood pressure, waist circumference and body mass index were measured as previously described [21].

All ultrasound scans were performed using a Phillips iu22 machine (Phillips Medical Systems, United States) with a C5-1 MHz general purpose curvilinear abdominal transducer by one of three experienced sonographers between December 2013 and December 2015. Each sonographer was a formally trained and registered sonographer with specialised experience in vascular ultrasound. Each participant was fasted for 12 h prior to their ultrasound scan to minimise interference from bowel gas. All participants were scanned in the supine position. The abdominal aorta was examined in both the transverse and sagittal planes to identify infra-renal landmarks, vessel tortuosity and obliqueness. Static ultrasound images in the transverse plane acquired in systole were obtained at the point of maximal dilation of the infra-renal abdominal aorta perpendicular to the central vessel line. The images were then centrally stored in accordance with health service policy.

Three observers were trained to measure maximal AAA diameter on static ultrasound images using a predefined protocol. Observers were selected based on pre-existing knowledge of aortic anatomy, cardiovascular physiology and imaging. Observer one was a qualified vascular sonographer with extensive experience in acquiring and interpreting ultrasound imaging of the aorta. Observer two was a clinical medical student with previous experience measuring AAA growth on computerised tomographic angiography scans. Observer three was a research worker and exercise physiologist with extensive experience in interpreting static ultrasound images of the abdominal aorta. A set measurement protocol was developed in consultation between a vascular sonographer, vascular surgeon and researcher. Five measurements were performed to assess AAA diameter. Three in the anterior–posterior (OTO, ITI and LETLE) and two (OTO and ITI) in the transverse plane (Fig. 1). Observer training involved both theoretical discussion and practical demonstration of the measurement protocol. Observers then measured a separate series of ten static ultrasound scan images independently before a consensus discussion was conducted on calliper placement between observers. Static ultrasound images were imported as DICOM images to the OsiriX Lite 32-bit version (Pixmeo, Geneva, Switzerland) software for analysis. To avoid bias, only static images where the sonographer placed measurement callipers had been omitted were included in this study. Prior to each measurement, each observer measured a 10-mm marked interval on a 100-mm scale to ensure accurate calibration of callipers. Each observer independently measured identical static images from the 50 participants and were blinded to the other observers’ measurements. Each observer repeated measurements 1 week later blinded to their earlier results.

Data analyses were performed in SPSS Version 23.0 (IBM SPSS Inc., Chicago, IL, United States) and R (Foundation for Statistical Computing, Vienna, Austria) with assistance from an expert statistician (REJ). Demographic data were reported as count (%) for dichotomous data and median (inter-quartile range 25th to 75th centile) for continuous data.

The intra-observer reproducibility coefficients for each individual observer are shown in Table 2. There was notable variation in reproducibility coefficient between each individual method and each individual observer. Scans obtained in the anterior–posterior plane and with LETLE calliper placement had the lowest overall intra-observer reproducibility coefficient (± 2.2 mm), but this was not statistically significantly different from ITI and OTO calliper placement in the same plane. All three observers consistently measured images in the transverse plane with a poorer repeatability as highlighted by the poorer overall intra-observer reproducibility coefficient for both ITI (± 4.9 mm) and OTO (± 5.6 mm) calliper placement. Measurements in the transverse plane (ITI P = 0.001; OTO P < 0.001) were significantly less reproducible than those measured in the AP plane (Table 2).

The inter-observer reproducibility coefficient for each method is shown in Table 3. The inter-observer reproducibility coefficients were poorer than the corresponding intra-observer reproducibility coefficient for each method. The inter-observer reproducibility was poorest for images measured transversely (Table 3).