Effect of Analyst Segmentation Variability on Computed AAA Stress Distributions

Abdominal Aortic Aneurysms (AAA) are a symptomless condition and are common in the geriatric population (over 65 years old). It is imperative for clinicians to identify regions of the AAA with greatest risk of rupture as well as overall rupture risk. A more dependable, patient-specific approach based on computational biomechanics which could prove useful. This research paper aims to identify and evaluate the extent of analyst variability during semi-automatic segmentation providing the geometry for a model. We used patient data obtained from Fiona Stanley Hospital in Perth, Western Australia, in the form of CT images at 80% through the cardiac cycle. Individual segmentations were performed by four analysts of similar experience with a repeat segmentation conducted four weeks later. Segmentations by different analysts produced different stress results and therefore affected the rupture potential prediction. Segmentations by the same analyst were not exactly reproducible. We recommend more stringent guidelines to guide analysts while utilising an automated segmentation process by edge detection may also prove to be feasible.