Building Three-Dimensional Intracranial Aneurysm Models from 3D-TOF MRA: a Validation Study

Abstract

To create realistic three-dimensional (3D) vascular models from 3D time-of-flight magnetic resonance angiography (3D-TOF MRA) of an intracranial aneurysm (IA). Thirty-two IAs in 31 patients were printed using 3D-TOF MRA source images from polylactic acid (PLA) raw material. Two observers measured the maximum IA diameter at the longest width twice separately. A total mean of four measurements as well as each observer's individual average MRA lengths were calculated. After printing, 3D-printed anatomic models (PAM) underwent computed tomography (CT) acquisition and each observer measured them using the same algorithm as applied to MRA. Inter- and intra-observer consistency for the MRA and CT measurements were analyzed using the intraclass correlation coefficient (ICC) and a Bland-Altman plot. The mean maximum aneurysm diameter obtained from four MRA evaluations was 8.49 mm, whereas it was 8.83 mm according to the CT 3D PAM measurement. The Wilcoxon test revealed slightly larger mean CT 3D PAM diameters than the MRA measurements. The Spearman's correlation test yielded a positive correlation between MRA and CT lengths of 3D PAMs. Inter and intra-observer consistency were high in consecutive MRA and CT measurements. According to Bland-Altman analyses, the aneurysmal dimensions obtained from CT were higher for observer 1 and observer 2 (a mean of 0.32 mm and 0.35 mm, respectively) compared to the MRA measurements. CT dimensions were slightly overestimated compared to MRA measurements of the created models. We believe the discrepancy may be related to the Laplacian algorithm applied for surface smoothing and the high slice thickness selection that was used. However, ICC provided high consistency and reproducibility in our cohort. Therefore, it is technically possible to produce 3D intracranial aneurysm models from 3D-TOF MRA images.

Keywords: 3D-TOF magnetic resonance angiography; Intracranial aneurysm; Three-dimensional printing.

Fig. 1

a–c 3D TOF MRI of a 74-year-old-man with saccular intracerebral aneurysm arising from terminal ICA is depicted. Axial (a), sagittal (b), and coronal (c) images (arrows) show aneurysmal outpouching

Fig. 2

a, b Image shows segmentation and data extraction process (pre-smoothing) from 3D TOF MRA data set with 3D Slicer software (a). Image depicts surface extraction and model processing (post-smoothing) of digital model via Z Brush software (b)

Fig. 3

Photographic image of 3D-printed anatomic model (arrow) created from polylactic acid raw material

Fig. 4

a–d CT scan of the model is shown in axial (a), sagittal (b), coronal (c), and 3D reconstruction (d) images

Fig. 5

a, b Graphic demonstrates intracranial aneurysm MRA averages of observers 1 and 2 for two consecutive measurements (a). Graphic reveals CT diameter means of 3D-printed anatomical model averaged by observers 1 and 2 for two serial measurements (b)

Fig. 6

a, b Bland-Altman plots of the mean differences between CT and MRA diameter means of 3D-printed anatomical models averaged by observer 1 (a) and observer 2 (b)