Undersampled projection reconstruction for active catheter imaging with adaptable temporal resolution and catheter-only views

Abstract

In this study undersampled projection reconstruction (PR) was used for rapid catheter imaging in the heart, employing steady-state free precession (SSFP) contrast. Active catheters and phased-array coils were used for combined imaging of anatomy and catheter position in swine. Real-time imaging of catheter position was performed with relatively high spatial and temporal resolution, providing 2 x 2 x 8 mm spatial resolution and four to eight frames per second. Two interactive features were introduced. The number of projections (Np) was adjusted interactively to trade off imaging speed and artifact reduction, allowing acquisition of high-quality or high-frame-rate images. Thin-slice imaging was performed, with interactive requests for thick-slab projection images of the signal received solely from the active catheter. Briefly toggling on catheter-only projection images was valuable for verifying that the catheter tip was contained within the selected slice, or for locating the catheter when part of it was outside the selected slice.

Figures

FIG. 1

Real-time interactive adjustment of temporal resolution is shown schematically. Imaging can be performed using any Np, during uninterrupted scanning. The Np is increased or decreased using a simple keyboard command. Here 32 projections and a 4-ms TR were used to acquire a number of short-axis cardiac images rapidly; Np was then increased to 40, and a 64-Np frame was obtained. During the 40-ms interscan delays, all of the necessary updates to gradient waveforms, and receiver phase and frequency tables were performed for acquisition with an adjusted Np.

FIG. 2

A zoomed view of the left ventricular cavity and ascending aorta. The catheter (arrow) was visualized in three consecutive frames, as the catheter device was navigated through the aortic arch into the left ventricular cavity using 196-ms temporal resolution, 48 projections, and 2 × 2 × 8 mm spatial resolution. Note the sharp visualization of the entire length of the active device, and the endocardial wall. Motion artifact is observed in the central image, due to the rapidly moving catheter (see Discussion).

FIG. 3

Catheter-only mode. a: A thin slice reveals part of the catheter (bright signal marked by the arrow indicates the apparent tip of the catheter) with respect to the anatomy; however, it was unclear whether the distal tip shown in this slice is the actual tip of the catheter, or the catheter tip was outside of the selected slice. b: Less than a half second later, the slice thickness was increased by a factor of 4, and the catheter channel was reconstructed alone, so that the entire active device was visualized by interactive request. The thick arrows point to identical positions in the FOV. The thin lower arrow in b points roughly toward the true catheter tip. A total of 64 projections were used.

FIG. 4

Comparison of catheter tracking with (a) PR and (b) Cartesian acquisitions, showing a single frame from real-time imaging in the same pig with each trajectory, at the same slice orientation. Imaging parameters for PR were: FOV = 32 cm, slice thickness = 8 mm, 160 Nr × 48 Np, receiver bandwidth = ±62.5 kHz, TR/θ = 4.1 ms/50°, resulting in 2 × 2 × 8 mm spatial resolution and 5.1 fps. For Cartesian: FOV = 36 cm, slice thickness = 8 mm, 192 Nx × 96 Ny, FOV = ¾ (72 views), receiver bandwidth = ±125 kHz, TR/θ = 3.7 ms/60°, resulting in 1.8 × 3.7 × 8 mm spatial resolution and 3.8 fps. SSFP was employed for both acquisitions.

FIG. 5

A series of long-axis images acquired about 5–10 s apart during continuous scanning, with 160 Nr and (a) 32 Np (7.6 fps), (b) 48 Np (5.1 fps), and (c) 64 Np (3.8 fps). Note the minor changes in artifact (especially radiating from the catheter) and SNR.

FIG. 6

a: A catheter (arrow indicates tip) was positioned toward the apex of the heart, imaged with 40 Np (164 ms temporal resolution). b: An injection of gadolinium (arrow) was made into the endocardium and observed with T1-weighted imaging, imaged with 40 Np (164 ms). c: The injection was imaged interactively with 64 Np (262 ms), providing higher SNR and fewer artifacts.