Fig 1.
Schematic illustration of the image processing protocol.
‘1- ‘ indicates the pre-first coiling treatment time point; ‘1+’, post-first coiling treatment time point; ‘2-’, pre-second coiling treatment time point; ‘2+’, post-second coiling treatment time point; 3DRA, 3D rotational angiogram. The first column depicts the image processing protocol for the 1- time point, beginning with generation of the aneurysm and vessel model from the subtracted 3DRA scan. The aneurysm sac is then automatically isolated from the vasculature. A representative 1- aneurysm sac model is shown at the bottom of column 1. The adjacent column details the workflow for analyzing the data from the 1+ time point. At this time point the coil mass model is generated from the baseline (or bone) 3DRA scan, while the vessel and residual blood model is generated from the subtracted angiographic scan. The coil mass, vessel and residual blood models are then added together by Boolean union. The aneurysm sac is the combination of the coil mass and any outlying residual blood, of which the aneurysm neck surface is automatically determined. A representative 1+ aneurysm sac model is shown at the bottom of column 2. The remaining two columns outline a similar workflow for analyzing data from the 2- and 2+ time points respectively.
Table 1.
Study subject procedural and demographic information
Fig 2.
Comparison of the computed sac and coil volumes by two investigators blinded from each other’s results.
Fig 3.
Aneurysm sac growth (VSG) and Coil mass growth (VCG) in the recurrence and control cohorts.
The box and whisker plots show quartiles and the p-values are from paired one-tail Wilcoxon tests for hypothesizing that the aneurysm sac will grow (null hypothesis, H0: VSG ≤ 0; alternative hypothesis, HA: VSG > 0) and the coil mass will compact (H0: VCG ≥ 0; HA: VCG < 0).
Fig 4.
(A) Coil mass center translation, δ, in the recurrence (N = 9) and control cohorts (N = 9). The box and whisker plots show quartiles and the p-values are from one-tail Mann-Whitney U test for hypothesizing that δ will be higher in the recurrence cohort than in control (null hypothesis, H0: δRECR ≤ δCTRL; alternative hypothesis, HA: δRECR > δCTRL). The triangles indicate raw data values. (B) Receiver operator curve showing the predictability of δ in recurrence (N = 9) and control (N = 9) aneurysms. δ is a better predictor of recurrence than clinically measured sac size, as it has a larger area under the curve or AUC (AUC = 0.74). Optimal sensitivity and specificity in differentiating recurrence aneurysms from control was found at δ = 1.1 mm.
Table 2.
Outcome summary of the paired one-tailed Wilcoxon tests for the recurrence and control cohorts.