A 1.5-T MRI scanner was customized for interventional use, with rapid imaging, independent color highlighting of catheter channels, multiple-slice 3D rendering, catheter-only viewing mode, and infarct-enhanced imaging. MRI receiver coils were incorporated into guiding catheters and injection needles. These devices were tested for heating and used for targeted MSC delivery. In infarcted pigs, myocardium was targeted by MR fluoroscopy. Infarct-enhanced imaging included both saturation preparation MRI after intravenous gadolinium and wall motion. Porcine MSCs were MRI-labeled with iron-fluorescent particles. Catheter navigation and multiple cell injections were performed entirely with MR fluoroscopy at 8 frames/s with 1.7×3.3×8-mm voxels. Infarct-enhanced MR fluoroscopy permitted excellent delineation of infarct borders. All injections were safely and successfully delivered to their preselected targets, including infarct borders. Iron-fluorescent particle–labeled MSCs were readily visible on delivery in vivo and post mortem.

We obtained serial representative 8-μm-thick sections of injection regions identified by tissue fast dye. Immunohistochemistry specimens were stained for desmin, with DAPI nuclear counterstain, and imaged under fluorescence, differential-phase, or light microscopy.

All injections were guided by steady-state free precession sequences (RT-SSFP),⁹ run in fluoroscopic (real-time) mode, with either rectilinear or radial k-space trajectories. For rectilinear trajectories, image-acceleration techniques were applied, such as echo sharing, in which acquisitions were alternated between even and odd phase-encoding lines with temporal filtering similar to that in UNFOLD,¹⁰ generating 8 frames/s and an acquisition-to-display latency of ≈250 ms. Typical rectilinear imaging parameters were repetition time (TR), 3.8 ms; time to echo (TE), 1.9 ms; flip angle, 60°; bandwidth, ±62.5 kHz; 192×96 matrix; 32×24-cm field of view; ¾ partial-phase acquisition generating 1.7×2.5×8-mm voxels. For radial trajectories, image acceleration was achieved by use of undersampling of projection views.⁷ The user interactively adjusted the compromise between temporal resolution and image quality by varying the amount of undersampling.⁷ Typical radial imaging parameters were TR, 4.0 ms; TE, 1.9 ms; flip angle, 60°; bandwidth, ±62.5 kHz; 160×64 projections; 32-cm field of view; generating 2×2×8-mm voxels. MSC injection sites were also inspected by use of segmented, ECG-gated, breath-held fast gradient echo: TR, 6.5 ms; TE, 2.9; flip angle, 15°; bandwidth, ±31.5 kHz; 256×128 matrix; 34×34-cm field of view; ¾ phase acquisition generating 1.3×2.7×8-mm voxels; and SSFP MRI: TR, 4.8 ms; TE, 1.9; flip angle, 60°; bandwidth, ±62.5 kHz; 256×128 matrix; 34×34-cm field of view; ¾ phase acquisition generating 1.3×2.7×8 mm voxels.

Infarcted targets were identified by impaired myocardial thickening during RT-SSFP MR fluoroscopy and by contrast enhancement. Although infarcted tissue is visible in RT-SSFP imaging without contrast agent, blood-to-infarct contrast is poor. To improve this contrast, delayed hyperenhancement (DHE) was used in conjunction with a 90° nonselective saturation pulse played before each echo-shared image (sat-RT-SSFP). DHE of the infarcted region was created by injecting Gd-DTPA 0.2 mmol/kg via a peripheral intravenous line 20 to 60 minutes before imaging. This non–ECG-gated, free-breathing technique was compared with RT-SSFP and with conventional segmented, ECG-gated, breath-held inversion recovery gradient echo (IR-GRE) DHE as the reference standard for identification of infarcted tissue.¹¹

Intravascular MRI receiver coils can potentially heat during scanning.¹² Catheter heating was tested¹³ with a 4-channel fluoroptic thermometer (model 3100, Luxtron). Probes were attached at the tips of the 9F left ventricular (LV) sheath, coaxial LV steering guide, and Stiletto, respectively. A reference probe measured the bath (in vitro) or core body (in vivo) temperature. Static phantoms, intended to exaggerate heating, included a 8×110-cm tube (small heat sink) and a 10-L tub (large heat sink), each filled with normal saline, with the catheter positioned in the isocenter or closer to the RF transmitter (20 cm off center of the 60-cm diameter bore, to increase susceptibility to heating). Recordings were also obtained in the iliac artery, aortic root, and LV cavity of a naive pig. Temperature fluctuations were recorded during continuous SSFP TR, 3.8 ms;α, 60° to 90°; with and without saturation preparation, for 120 seconds after steady state was achieved, typically 150 to 300 seconds. Worst-case scenarios for heating were also simulated by disconnecting one or both of the catheter coils from the MRI receivers.

Guiding catheters were introduced into the LV under RT-SSFP guidance using color highlighting (green) of signal from the built-in coil. Preselected normal, infarct, or border targets were identified by regional wall motion and DHE. Catheter manipulations were visualized with RT-SSFP and were facilitated by toggling catheter-only mode (Figure 2) when part of the catheter moved out of the selected imaging slice and by 3D rendering of multislice acquisitions to portray complex anatomic relationships. Next, the Stiletto was inserted and extended beyond the tip of the guiding catheter to engage the targeted myocardium. Needle purchase was confirmed by 50- to 150-μL test injections of 30 mmol/L Gd-DTPA, after toggling saturation preparation, looking for localized hyperenhancement, which was brighter than the surrounding normal or hyperenhanced myocardial infarction. A representative injection sequence is shown in Figure 3. For portions of the active guiding catheter in the imaging plane, the lumen was visible as a dark stripe with color-highlighted regions (green) on either side. Portions of the catheter that were near but not in the imaging plane were still visible as color-highlighted regions. Thus, the sensitivity profile of the active guiding catheter was used interactively to optimize the imaging plane position and orientation. Satisfactory myocardial purchase was demonstrated reproducibly by apposition of the deployed needle coil with myocardium and by mechanical feedback to the operator. To test the precision of infarct border targeting, a very small infarct (12×8×8 mm) was created. Multislice MR fluoroscopy permitted cells to be positioned precisely along the apical and basal borders of this infarct (Figure 4 and online-only Data Supplement).

Transcatheter intramyocardial injection will most likely be useful for clinical delivery of MSCs and other emerging cellular therapies. Depositing cells into the myocardial inter-stitium may circumvent the requirement that intravenously or intra-arterially administered cells home to target tissue on the basis of poorly understood signaling processes. Precise targeting of myocardial injections based on tissue viability, eg, the border between healthy and infarcted myocardium, could be important for efficacy. Even after direct injection, a large fraction of injected material may immediately exit the myocardium.^14,15 Although iron labeling most likely exaggerates the volume of delivered cells, we demonstrate that magnetic labeling of stem cells permits immediate confirmation of successful delivery and precise localization in vivo. However, the size of signal voids in vivo is probably not proportional to cell quantity; therefore, this technique is limited in ability to visualize early injectate loss. IFP and other labeling techniques¹⁶ may prove useful to trace the distribution of injected cells in the beating heart.

Online regional functional myocardial assessment was performed in these pig experiments with real-time SSFP, and the findings were satisfactory to target therapy delivery. This was possible without breath holding or ECG gating. The imaging rate of 8 frames/s was adequate to provide a useful qualitative assessment of myocardial contractility. In addition, saturation preparation after systemic Gd-DTPA administration provided additional contrast between infarcted myocardium and blood that was crucial for targeting delivery. Further study will be necessary to determine the usefulness of this technique in determining infarct size.

The “active” catheters we used, so designated because they serve as radiofrequency MRI antennas and receive signal from nearby blood and tissue, proved essential in these experiments. For example, color highlighting of the catheter and needle tip signals markedly improved operator confidence in needle position and purchase. Interactive catheter-only display was helpful to visualize the catheter when outside the selected imaging slice. “Passive” catheters, such as ferromagnetic coronary stents¹⁷ and atrial septal occluder devices,^18,19 are visualized as signal voids (black spots) and are not amenable to such interventional imaging enhancements.

Several alternative technologies allow endomyocardial injections. Transesophageal and intracardiac ultrasound systems have been used to guide direct transcatheter myocardial gene transfer.²³ Ultrasound imaging is limited by “black-out” regions created by device shadows. Electromechanical navigation and injection systems, which overlay a position sensor onto a previously acquired endocardial roadmap, can guide myocardial drug delivery.²⁴ However, the spatial precision of this approach is controversial, given the coarse “road mapping” over previously acquired moving images. Moreover, injection success and intramyocardial distribution cannot be assessed with electromechanical mapping.