Integration of cardiac and respiratory motion into MRI roadmaps fused with x-ray

Abstract

Purpose: Volumetric roadmaps overlaid on live x-ray fluoroscopy may be used to enhance image guidance during interventional procedures. These roadmaps are often static and do not reflect cardiac or respiratory motion. In this work, the authors present a method for integrating cardiac and respiratory motion into magnetic resonance imaging (MRI)-derived roadmaps to fuse with live x-ray fluoroscopy images, and this method was tested in large animals.

Methods: Real-time MR images were used to capture cardiac and respiratory motion. Nonrigid registration was used to calculate motion fields to deform a reference end-expiration, end-diastolic image to different cardiac and respiratory phases. These motion fields were fit to separate affine motion models for the aorta and proximal right coronary artery. Under x-ray fluoroscopy, an image-based navigator and ECG signal were used as inputs to deform the roadmap for live overlay. The in vivo accuracy of motion correction was measured in four swine as the ventilator tidal volume was varied.

Results: Motion correction reduced the root-mean-square error between the roadmaps and manually drawn centerlines, even under high tidal volume conditions. For the aorta, the error was reduced from 2.4 ± 1.5 mm to 2.2 ± 1.5 mm (p < 0.05). For the proximal right coronary artery, the error was reduced from 8.8 ± 16.2 mm to 4.3 ± 5.2 mm (p < 0.001). Using real-time MRI and an affine motion model it is feasible to incorporate physiological cardiac and respiratory motion into MRI-derived roadmaps to provide enhanced image guidance for interventional procedures.

Conclusions: A method has been presented for creating dynamic 3D roadmaps that incorporate cardiac and respiratory motion. These roadmaps can be overlaid on live X-ray fluoroscopy to enhance image guidance for cardiac interventions.

Figures

Figure 1

Flow-chart showing processing steps for (top) image acquisition and retrospective sorting according to cardio-respiratory phase; and (bottom) image warping and fit to affine model.

Figure 2

MR image showing three-chamber view of the left ventricle. Arrows point to the RCA, aorta, and diaphragm. Additional slices acquired for motion tracking are depicted as intersecting lines. The dotted line is perpendicular to the long axis of the aorta, and the dashed line intersects the short axis of the proximal RCA. All three slices intersect the diaphragm, allowing tracking of respiratory motion.

Figure 3

Respiratory navigator on real-time MR images. (a) Position of respiratory navigator (filled white circle) on real-time images. Open circles represent the sampled grid used for warping. (b) Head-foot position of navigator, with extracted inspiratory and expiratory phases. (c) Interpolation of respiratory and cardiac phases. Acquired images (dots) are sorted according to cardiac phase and respiratory position. They are linearly interpolated to integer mm diaphragm positions, and across cardiac phases (grid). End-expiration and end-inspiration are defined as the maximum and minimum respiratory positions for which all cardiac phases have also been acquired (or interpolated). In this example, end-expiration = 0 mm and end-inspiration = −7 mm.

Figure 4

Contours on MR images for (left) aorta and (right) RCA. These contours were drawn on breath held cine SSFP images and were used for the overlaid roadmaps on live x-ray.

Figure 5

Respiratory navigator on angiography images. Aorta (surface) and proximal RCA (line) are also shown. Navigator (plus sign) tracks edge of diaphragm. End-inspiration (a) and end-expiration (b) positions (triangles) are estimated from training phase.

Figure 6

Affine transform components for aorta (left) and proximal RCA (right) for one subject. From top to bottom, plots are of scale, shear, rotation, and translation. For respiratory motion (first and third columns), cardiac phase is fixed to end-diastole, and for cardiac motion (second and fourth columns) respiratory phase is fixed to end-expiration. For both aorta and RCA, the magnitude of cardiac motion is larger than respiratory motion.

Figure 7

Centerline tracking in the aorta and proximal RCA. Top row shows regions of magnification for context. (a)–(c) Aortagram indicating regions of interest. (d)–(g) show overlaid centerlines. The respiratory phase (R) and cardiac phase (C) are shown at the top of each frame. Respiratory phases are from 0 (end-expiration) to ±1 (end inspiration), with (−/+) indicating inspiratory/expiratory phases, respectively. Cardiac phases are from 0 to 1. The manually drawn centerline is drawn as a solid white line, the motion-corrected centerline with filled white circles, and the static centerline with open white circles. During systole, (d) and (f), the motion-corrected centerline is closer to the manual contour than the static centerline. At end diastole, (e) and (g), the centerlines overlie one another, as expected. The differences between the static and motion-corrected contours for the aorta were less compared with differences for the RCA.

Figure 8

Root-mean-square error for (a) aorta and (b) right coronary artery. Values are shown for BH, LOW VT, and HIGH VT settings. Bar height is median value, and error bar is interquartile range. Static errors (black bars) are compared with motion-corrected errors (white bars). Asterisks (*) indicate statistically significant difference between static and motion-corrected errors