Hostile aortic neck characteristics, such as short length (<1 cm), severe suprarenal (>45˚) and infrarenal (>60˚) angulation, conicity, and large diameter (>30 mm) have been associated with increased risk for a type Ia endoleak (T1aEL). However, the actual aortic neck is a complex 3D shape that is oversimplified by these two-dimensional neck characteristics. Statistical shape models (SSM) are composed of principal components and enable the description of a neck's true 3D morphology. This study aimed to investigate the ability of the SSM besides the conventional measurements to differentiate between patients with and without T1aEL.

The dataset was composed of EVAR patients facing a T1aEL during follow-up and a control group without T1aEL, each consisting of 63 patients. Patients in the control group were not matched to the T1aEL group on characteristics, but 63 patients with the longest follow-up, either computed tomography angiography (CTA) or duplex ultrasound, were included. In the pre-operative CTA scan of each patient, the center lumen line (CLL) and a triangular surface segmentation of the aortic lumen were semi-automatically obtained. Moreover, 3D coordinates were placed at the orifice of the lowest renal artery (LRA) and at the distal end of the aortic neck (10% increase in diameter compared to the diameter at the orifice of the LRA). The data was pre-processed for a principal component analysis (PCA) by means of a parametrization method including rigid registration. For each patient, 360 contour points along 10 CLL points were created, ensuring anatomical point-to-point correspondence. The contour points were used as input for a PCA from which the SSM was created. Differences in neck length, neck diameter, supra- and infrarenal angulation and principal component (PC) scores were assessed with a Mann-Whitney U Test. Two logistic regression models, the first model SSM-based and the second model based on conventional aortic neck measurements, were created to differentiate between the two groups. Sensitivity, specificity and area under the curve (AUC) values were calculated for these two logistic regression models.

Of the 126 patients, 106 patients were male (84%), and the mean age was 74 ± 7 years. Patient demographics and total imaging follow-up duration, which consisted of CTA for the T1aEL group and either CTA or duplex for the control group, showed not to be statistically different between the two groups. Neck length was significantly smaller in the T1aEL group, whereas neck diameter and suprarenal angulation were significantly larger in the T1aEL group. Nine principal components (PCs) were needed to describe 98% of the total aortic neck shape variation. The mean shape of the neck in the T1aEL group (red) versus the mean shape of the neck in the control group (brown) are shown in Figure 1. Figure 2 shows the mean shape, -3 standard deviation (SD) shape, and +3SD shape for the first three PCs, describing 54%, 27% and 9% of the total shape variation, respectively. Scores for PC 1, 8 and 9 were significantly different between the T1aEl group and the control group. Sensitivity, specificity and AUC values for the SSM-based and conventional measurements-based logistic regression models were 79%,70% and 0.82 versus 70% and 71% and 0.79, respectively.

This is the first study that created an SSM of the infrarenal AAA neck of a substantial group of patients with and without a T1aEL. With this semi-automatic SSM, shape variations can be analysed that are otherwise impossible to obtain from single conventional neck characteristics. The SSM was able to differentiate between with hostile and non-hostile aortic necks with comparable accuracy as conventional measurements.