Using temporal and structural data to reconstruct 3D cerebral vasculature from a pair of 2D digital subtraction angiography sequences

Abstract

Purpose: The purpose of this work is to present a new method for reconstructing patient-specific three-dimensional (3D) vasculature of the brain from a pair of digital subtraction angiography (DSA) image sequences from different viewpoints, e.g., from bi-plane angiography. Our long-term goal is to provide high resolution visualization of 3D vasculature with dynamic flow of contrast agent from limited data that is readily available during surgical procedures. The proposed method is the second of a three-stage process composed of 1) augmenting vessel segmentation with vessel radii and timing of the arrival of a bolus of contrast agent, 2) reconstructing a volumetric representation of the augmented vessel data from the augmented 2D segmentations, and 3) generating a 3D model of vessels and flow of contrast agent from the volumetric reconstruction. Unlike previous methods, which are either limited to relatively simple vessel structures or rely on multiple views and/or prior models of the vasculature, our method requires only a single pair of 2D DSA sequences taken from different view directions.

Methods: We developed a new mathematical algorithm that augments vessel centerlines with vessel radii and bolus arrival times derived directly from the 2D DSA sequences to constrain the 3D reconstruction. We validated this method on digital phantoms derived from clinical data and from fractal models of branching tree structures.

Results: In standard reconstruction methods, reconstruction by projection of two views into 3D space results in 'ghosting' artifacts, i.e., false 3D structure that occurs where vessels or vessel segments overlap in the 2D images. For the complex vascular of the brain, this ghosting is severe and is a major hurdle for methods that attempt to generate 3D structure from 2D images. We show that our approach reduces ghosting by up to 99% in digital phantoms derived from clinical data.

Conclusion: Our dramatic reduction in ghosting artifacts in 3D reconstructions from a pair of 2D image sequences is an important step towards generating high resolution 3D vasculature with dynamic flow information from a single DSA sequence acquired using bi-plane angiography.

Keywords: 3D reconstruction; Brain blood vessels; Cerebral vasculature; Digital subtraction angiography.

Figure 1.

Cerebral vasculature imaging of a healthy subject. Left: 2D DSA near the end of the arterial phase. Center: maximum intensity projection from MRA. Right: 3D model generated from CTA. MRA and CTA provide lower spatial resolution than DSA.

Figure 2.

Cone beam projection geometry typical in DSA angiography systems. Top: X-rays are projected from source S located at a distance L from the detector D about a central axis that intersects the detector at right angles at origin O. Image $\left(\text{x},\text{y}\right)$ coordinates are centered at O with y defined the up vector. Bottom: In biplane angiography, two sources (S₁ and S₂) and two detectors (D₁ and D₂), typically at right angles, can be used to produce two DSA sequences simultaneously. The sources and detectors are mounted so they can be rotated on a circular arc about the patient to obtain multiple sequences (e.g., for rotational angiography).

Figure 3.

The three stages of our full pipeline: 1) process each 2D DSA sequence to generate an annotated 2D image or an annotated set of 2D model elements; 2) reconstruct a volume representation from the annotated images/elements; and 3) generate a 3D model of the vasculature and simulate dynamic flow of contrast agent. This paper provides and evaluates a new algorithm for stage 2.

Figure 4.

Fractal trees used for testing. Fractal trees have the advantage that tree shape, branching pattern, average branch length and radius and number of branching level scan be controlled. Left: two orthogonal views of a fractal tree phantom with one branching level. Center: the same views of the phantom with two branching levels. Right: the same tree with three branching levels.

Figure 5.

Examples of 3D vessel phantoms of brain arterial trees generated from The Brain Vasculature (BraVa) database [60] which stores centerlines and radii of the left and right anterior, middle and posterior cerebral arteries from MRA of 61 subjects.

Figure 6.

A) Simple fractal tree phantom with one level of branching. B), C) Projections of the phantom onto two detector planes at 90 degrees to each other with B) illustrating overlapping of the smaller branches. D) A projection at 45 degrees between B and C. E) 3D reconstruction using only B) and C) shows ghosting (transparent blue ghost branches in the expanded circle). F) Ghosting is reduced when the third projection of C) is also used in the 3D reconstruction.

Figure 7.

Left to Right: Fractal tree phantom with two levels of branching, projections of the phantom onto two detector planes at 90 degrees to each other, and the 3D reconstruction viewed from the one of the projection directions and from an oblique direction. When the reconstruction is viewed from one of the two view directions, it looks perfect (blue centerlines are well aligned with the centerlines of the original tree model). However, as shown in the oblique view of the far right, overlapping vessels in the two views lead to large amounts of ghosting in the reconstructed model.

Figure 8.

Annotations of a fractal tree phantom. Top row: lateral view. Bottom row: front-to-back view of the fractal tree with two branching levels. The two left images show the 3D model and the projected centerlines. The three right images, left to right, show annotations of the vessel radii, bolus arrival time, and $v_y$ encoded as grey scale values.

Figure 9.

Annotations of a BraVa vascular model. Top row: lateral view. Bottom row: front-to-back view of a phantom of the middle cerebral artery from a BraVa model. The two left images show the 3D model and the projected centerlines. The three right images, left to right, show annotations of the vessel radii, bolus arrival time, and $v_y$ encoded as grey scale values.

Figure 10.

Fractal tree phantom (left two columns) and a phantom of right middle carotid artery (right two columns) from oblique views showing ghosting. Each row (top to bottom) shows: a 3D model of the phantom; volume reconstruction with no constraints; volume reconstruction constrained by vessel radii; volume reconstruction constrained by bolus arrival times; volume reconstruction constrained by $v_y$; and volume reconstruction constrained by all three constraints.