Variability of vascular CT measurement techniques used in the assessment abdominal aortic aneurysms

Purpose: The aim of this project is to assess the variability of six CT measurement techniques for sizing abdominal aortic aneurysms (AAAs). Method: 37 CT scans with known AAAs were loaded on to a departmental picture archiving and communication system (PACS). A team of three observers, with experience in aortic CT measurements and the PACS performed a series of 2D and 3D measurements on the abdominal aorta. Each observer was asked to measure 3 quantities; anterior-posterior AAA diameter, maximum oblique AAA diameter, maximum aneurysm area using both 2D and 3D techniques. In order to test intra-observer variability each observer was asked to repeat their measurements. All measurements were taken using electronic callipers, under standardised viewing conditions using previously calibrated equipment. 3D measurements were conducted using a computer generated central luminal line (CLL). All measurements for this group were taken perpendicular to the CLL. Results: A total of 972 independent measurements were recorded by three observers. Mean intra-observer variability was lower for 2D diameter measurements (AP 1.3 ± 1.6 mm; 2D Oblique 1.2 ± 1.3 mm) and 2D areas (0.7 ± 1.3 cm²) when compared to inter-observer variability (AP 1.7 ± 1.9 mm; Oblique 1.6 ± 1.7 mm; area 1.1 ± 1.5 cm²). When comparing 2D with 3D measurements, differences were comparable except for 3D AP diameter and area which had lower inter-observer variability than their 2D counterparts (AP 2D 1.7 ± 1.9 mm, 3D 1.3 ± 1.3 mm; area 2D 1.1 ± 1.5 cm², 3D 0.7 ± 0.7 cm²). 3D area measurement was the only technique which had equal variability for intra- and inter-observer measurements. Overall observer variability for the study was good with 94-100% of all paired measurements within 5.00 mm/cm² or less. Using Pitman's test it can be confirmed that area measurements in the 3D plane have the least variability (r = 0.031) and 3D oblique measurements have the highest variability (r = 0.255). Conclusion: 3D cross-sectional area measurement techniques have the lowest variability and should be preferred for repeatable measurements of AAAs where possible. Results confirm that both inter- and intra-observer variability exists for all measurement techniques.