|Home | About | Journals | Submit | Contact Us | Français|
This study was designed to evaluate the characteristics of gadolinium-enhanced imaging of radiofrequency ablations.
Gadolinium-enhanced magnetic resonance imaging (MRI) has been used successfully to evaluate tissue necrosis after myocardial infarction. In electrophysiology, radiofrequency energy is used to create a targeted myocardial necrosis for the treatment of various arrhythmias.
Using a power-controlled, cooled-tip 7-F catheter system, radiofrequency lesions (10 to 40 W for 30 s) were created on the epicardium of the right ventricle in eight mongrel dogs. After injection of 0.225 mmol/kg gadolinium, T1-weighted fast gradient echo images were obtained during a follow-up of 10 h using an intrathoracic high-resolution coil. Radiofrequency ablations were analyzed on the MR images and compared with gross anatomy and histopathology.
Four distinct phases of signal enhancement were observed. After gadolinium injection, radiofrequency lesions were delineated clearly as contrast-free areas of low signal intensity (contrast-to-noise ratio [CNR] = −21.1 ± 19.8). Signal enhancement in the lesion periphery started 4.0 ± 1.8 min after injection and progressively extended toward the lesion center at a rate of 0.02 mm/min. Full delayed enhancement was observed after 98 ± 21 min (CNR = +17.8 ± 9.0). During the follow-up period, CNR started to decrease, but the lesions were detectable for as long as 10 h of follow-up. During the first three phases of enhancement, MRI correlated well with the pathological findings (r = 0.88, r = 0.88, and r = 0.86 [p < 0.001], respectively).
Radiofrequency ablation can be evaluated accurately by using gadolinium-enhanced MRI, which may allow the noninvasive assessment of procedural success. The dissimilar wash-in and wash-out kinetics compared with myocardial infarction suggest a different pathophysiological process with complete loss of microvasculature.
Radiofrequency ablation (RFA) allows the catheter-based creation of well-defined, limited myocardial necrosis and has significantly changed the treatment of arrhythmias since its introduction to electrophysiology (1). However, newer ablation strategies increasingly rely on anatomic rather than electrophysiological approaches (2,3), which makes the visualization of the lesions with respect to their extent and anatomic location increasingly important. Yet, the current catheter guiding techniques mostly rely on fluoroscopy and three-dimensional mapping devices, which are limited in their ability to confirm the presence, exact location, and extent of ablation lesions, parameters that serve as indicators of the procedural success (3,4).
This anatomic information could potentially be provided by gadolinium-enhanced magnetic resonance imaging (MRI), as it has been shown to reliably visualize necrotic tissue (after myocardial infarction) with a reproducible enhancement pattern (5). First-pass images performed immediately after a gadolinium injection frequently demonstrates areas of hypoenhancement in the endocardial core of the infarct (6,7), followed by complete delayed enhancement of the myocardial necrosis 5 to 25 min after contrast injection with a subsequent wash-out period (6,8).
Correspondingly, MRI can detect the heat-related changes of T1 and T2 parameters in the necrotic myocardium after RFA and is able to confirm the existence of RFAs in the very acute setting (9). However, many clinical relevant questions, such as the characterization of radiofrequency lesions during the different stages of enhancement, the effect of radiofrequency energy on lesion visualization, and the ability to assess anatomic correlates to ablation success such as transmural extent and lesion gaps, will have to be clarified before this technology can be applied for prolonged and frequently complicated clinical procedures. Therefore, this study sought to systematically evaluate the characteristics of MRI for radiofrequency lesions.
Eight mongrel dogs weighting 30 to 35 kg were premedicated with 10 mg of ketamine intramuscularly and maintained on 80% oxygen and 1.5% isoflurane gas using a Narkomed Anesthesia ventilator (Draegar, Telford, Pennsylvania). Surface electrocardiogram leads I, II, and III were monitored continuously. Using the cutdown technique, an 8-F introducer sheath was placed in the right jugular vein for the administration of fluids and medication. Using clinical-grade surgical equipment, a thoracotomy was performed in the midsternal line and the pericardium removed. Using a standard 7-F ablation catheter with a 4-mm electrode single, radiofrequency lesions were created on the epicardial surface of the right ventricle with a clinical-grade radiofrequency generator (Atakr, Medtronic, Minneapolis, Minnesota) while the ablation site was constantly irrigated with 0.9% normal saline (21° C) at a rate of 5 ml/min. Ablation lesions were created in a linear interrupted fashion using a power-controlled mode at 10, 20, 30, and 40 W for 30 s each. After the ablation, the animals were monitored hemodynamically for 10 min, surgical tools and paramagnetic materials were removed, and the live animals were transported immediately with an unclosed chest preparation to the MRI suite. All animal protocols were reviewed and approved by the Animal Care and Use Committee at the Johns Hopkins University and conformed to the guidelines published in the “Position of the American Heart Association on Research Animal Use.”
The animal was placed in a 1.5-T MR scanner (Signa XL, GE, Waukesha, Wisconsin). Using a 3-inch high-resolution coil, which was placed intrathoracically on the right ventricle, the correct imaging plane of the ablation lesions was identified with standard T1- and T2-weighted scout images. Magnevist (0.225 mmol/kg; gadopentetate dimeglumine, Berlex, Monteville, New Jersey) was injected intravenously. Continuous long-axis images were obtained throughout the area of RFA with a 3-mm slice thickness every 5 to 15min during the first 3 h and every 30 to 60 min up to a duration of 15 h using a T1-weighted, inversion prepared fast gradient recall echo sequence (TR = 3.6 ms, TE = 1.7 ms, TIadjusted = 175 to 275 ms [to “null” normal myocardium], field of view = 18 cm, flip angle = 20°, readout bandwidth = 32 kHz, matrix 256 × 192, number of signal averages (NEX) 1, slice thickness = 3 mm). All images were visually inspected to rule out the presence of susceptibility artifacts caused by the air-tissue interface of the open chest preparation. Animals were monitored and alive during all imaging sequences.
Immediately after the imaging protocol, the animals were injected with 20 ml of a 20% solution of 2,3,5-triphenyltetrazolium chloride and sacrificed with an intravenous injection of high-molar potassium chloride solution. The hearts were excised, fixated, and sectioned through the right ventricular lesion corresponding to the tomographic MRI slices. Lesion location, morphology, width, length, and transmural extent were determined and recorded at gross examination and photographed for later comparison with the MR images. Sections from thermally damaged tissues were submitted for histological staining (Masson’s trichome and hematoxylin-eosin) and analyzed under light microscopy to characterize global morphological changes.
Lesion character and signal intensity (SI) were measured directly from MR images and analyzed with an off-line quantitative analysis package (version 1.53; e-film, Merge Technology, Milwaukee, Wisconsin). Lesion size was determined as the maximal lesion diameter at the outer margin of the contrast void (Phase 1) or the enhanced lesion (Phase 2 to 4).
Signal-to-noise ratio (SNR) was measured as the SI of the lesion over the standard deviation (SD) of the background noise. Contrast-to-noise ratio (CNR) was determined as the SI of lesion minus the SI of the adjacent myocardium, which was then divided by the SD of the background noise. Lesion parameters at gross examination were measured independently of magnetic resonance images and compared.
Changes in lesion size and SI were analyzed during the time-course and between the various energy groups and were considered significant at a level of p < 0.05 in a paired t-test (analysis of variance [ANOVA]). Comparisons of ablation lesions in MRI and in pathological specimen was performed by a native analysis using linear regression and additionally by an adjusted analysis using a generalized estimating equation approach (GEE) to control for a potential cluster effect. Unless noted otherwise, results measurements are reported as arithmetic mean ± SD. We considered p values at a level of <0.05 as statistically significant.
Independent of the radiofrequency energy and the size of the ablation lesion, all radiofrequency lesions could be visualized after enhancement using gadolinium. We injected the gadolinium 2.1 ± 0.6 h after the creation of the radiofrequency lesions.
In all experiments, four characteristic phases of enhancement were observed after the administration of gadolinium. In the first phase immediately after injection, radiofrequency lesions were visible as areas of contrast void (Figs. 1A and 1B) in comparison with the enhanced, nonablated myocardium (CNR = −21.1 ± 19.8) (Fig. 2A) and blood pool (CNR = −49.8 ± 38.7). Measured SNRs were 8.1 ± 3.9 vs. 29.3 ± 17.1 (p < 0.01) (Fig. 2B) and 8.1 ± 3.9 vs. 59.1 ± 43.2 (p < 0.01), respectively. Lesion size by MRI was 7.8 ± 2.8 mm compared with in the pathological specimen 6.4 ± 2.7 mm (p > 0.05 for difference). Magnetic resonance imaging correlated well with pathological results (r = 0.88, p < 0.001; GEE: regression coefficient 0.87, p < 0.0001) (Fig. 3A) and slightly overestimated the lesion size by 1.4 ± 1.7 mm.
In the second phase, a high SI enhancement was observed starting from the lesion periphery 4.0 ± 1.8 min after injection (Figs. 1C and 1D). This ring-like enhancement increased in thickness at a rate of approximately 0.02 mm/min toward the lesion center (r = 0.88, p < 0.001; GEE correlation coefficient 0.78, p < 0.0001) (Fig. 3D). Peripheral hyperenhancement was well delineated against the nonenhanced lesion center (CNR = 11.5 ± 3.3; SNR 30.2 ± 14.6 vs. 18.7 ± 13.4; p < 0.001) and the surrounding myocardium (CNR = 18.0 ± 11.7; SNR 30.2 ± 14.6 vs. 12.2 ± 11.3; p < 0.001) as shown (Figs. 2A and 2B). Lesion size did not differ significantly from measurement of the contrast void (Phase 1; 7.7 ± 2.6 mm vs. 7.8 ± 2.8 mm; ANOVA: p > 0.05) and correlated well with the pathological measurements (r = 0.88, p < 0.001; GEE regression coefficient 0.89, p < 0.0001) (Fig. 3B). Magnetic resonance imaging overestimated the lesion size by 1.3 ± 1.9 mm, which was not significantly different from the magnetic resonance measurements in Phase 1.
During the third phase, the complete lesions showed homogenous gadolinium enhancement (Fig. 1E). Time to complete enhancement was longer with larger lesion size and occurred in average after 98 ± 21 min (Fig. 3D). The lesion was easily detected as a result of a CNR of 17.8 ± 9.0 (SNR 24.1 ± 10.3 vs. 6.3 ± 2.1; p < 0.0001) (Figs. 2A and 2B) and correlated well with the pathological specimen (r = 0.86, p < 0.001; GEE regression coefficient 0.85, p < 0.0001) (Fig. 3C). The lesion size by MRI was 8.0 ± 3.0mm, which was not different from the lesion size estimated during the Phase 1 or 2. The magnetic resonance image overestimated the pathological lesion size by 1.6 ± 2.1mm, which was not statistically significant from Phase 1 or 2 (ANOVA).
During further follow-up, the imaged lesion size started to decrease from the periphery (Phase 4), which was matched by a beginning decrease in overall SI (Fig. 1F). After 288 ± 33 min, the CNR was 15.6 ± 6.2 (SNR 20.0 ± 6.5 vs. 4.4 ± 1.5; p < 0.001), which allowed the lesions to remain visible for the complete follow-up of up to 10 h (Figs. 2A and 2B). As expected, the lesion size increased with the amount of ablation energy (p < 0.05), but the lesion-enhancement pattern and the overall signal intensities were independent of the radiofrequency energy applied (p > 0.05 by ANOVA).
Trans-mural extent was assessed during the first phase of a contrast void as well as during the third phase of complete enhancement in 25 radiofrequency lesions with an average depth of 4.9 ± 1.8 mm (range, 2.0 to 10.6 mm corresponding to 26% to 100% of the ventricular wall thickness). Seven of the 25 lesions were transmural. Assessment of the transmural extension by MRI correlated well with the pathological specimen (r = 0.87 and r = 0.85, respectively, p < 0.01 for both; GEE regression coefficient 0.83 and 0.73, p < 0.0001 for both) and did not show a significant difference when analyzed for increasing degrees of transmural extent (26% to 49%, 50% to 75%, or 76% to 100%, respectively). Measurements during the phase of contrast void overestimated the transmural extent by 1.1 ± 1.3 mm and classified three nontransmural lesions incorrectly (sensitivity: 100%, specificity: 83%). During “very” delayed enhancement, the lesions appeared 1.7 ± 1.5 mm larger. One lesion was incorrectly classified as transmural and one as nontransmural (sensitivity: 86%, specificity: 95%).
Twenty-one discontinuous lesions (“gaps”) with an average size of 6.0 ± 4.7 mm [range, 0 mm to 23.9 mm] were similarly assessed during the early phase of contrast void and the delayed phase of complete gadolinium enhancement (Figs. 2 and and4).4). Early (contrast-void) and late (complete enhancement) images after gadolinium injection allowed excellent delineation of the radiofrequency gaps with a CNR of −21.2 ± 19.6 and 17.9 ± 10.2, respectively (p < 0.001). The gap size was correlated well with the pathological measurements (r = 0.86 for both; p < 0.01, GEE regression coefficient 0.77 and 0.80, p < 0.0001 for both) and underestimated the gap by 1.3 ± 1.1 mm and 0.8 ± 1.2 mm during contrast void and complete enhancement, respectively.
Radiofrequency lesions were characterized by a core of severe coagulation necrosis with complete loss of cellular and vascular architecture, which covered >80% of the total lesion area. In the lesion periphery, a rim of contraction band necrosis was observed with partially intact myocardial cells and areas of intralesional hematoma. Intact vascular structures obstructed with neutrophiles, erythrocytes, and cellular fragments were observed. This area was surrounded by a perilesional rim of intact myocardial cells exhibiting edema with occasional intralesional hemorrhage, but mostly intact vasculature (Fig. 5).
The main findings of this study are: 1) gadolinium-enhanced MRI of radiofrequency lesions displays four distinct phases of enhancement: signal void, peripheral enhancement, “very” delayed complete enhancement, and loss of enhancement; 2) radiofrequency lesions size, transmural extent, and interlesional gap can be adequately assessed throughout the different stages of gadolinium enhancement; 3) MRI appearance of radiofrequency lesions is similar over a wide range of applied energy; and 4) wash-in kinetics are different from myocardial infarction and may suggest a diffusion process.
Radiofrequency lesions of all energy levels were visible after gadolinium injection as a signal void. A lack of enhancement in the endocardial core has been described in acute myocardial infarction during the first transit of the gadolinium in only about 50% of ischemically injured segments (10) and is thought to represent occlusion of microvasculature with erythrocytes, neutrophils and cellular debris, which is also known as “no-reflow” phenomenon (6,7). However, all these segments displayed some gadolinium enhancement after about 7 min, corresponding to the near-equilibrium (pseudoequilibrium) phase with complete enhancement after about 20 min (10). Similarly, in patients undergoing ethanol-induced septal ablation for hypertrophic obstructive cardiomyopathy complete delayed enhancement could be demonstrated after 20 min (11). Differently, in RF lesions, the contrast void was a predominant characteristic after gadolinium injection in all lesions and was not replaced by complete enhancement until about 1.5 h after gadolinium was administered. This observation might be explained by the histology of the RF lesions, which showed severe coagulation and contraction band necrosis with complete loss of cellular and vascular architecture up to a small transition zone at the periphery (12). Thus, RF lesions may exhibit an extreme and irreversible “no-reflow” phenomenon, responsible for the different wash-in kinetics observed.
These histological findings also may explain the second phase of a slowly progressive, ring-like, peripheral enhancement of RF lesions, which is different from infarcted myocardial tissue. Gadolinium with a molecular weight of approximately 800 Daltons is thought to be biologically inert and to passively diffuse throughout the extracellular space (13). Although after myocardial infarction reperfusion or collateral blood supply achieves complete delayed hyper-enhancement after 10 min (6–8), enhancement of the devascularized RF lesion seems to rely mostly on diffusion from the lesion periphery. Because the diffusion time increases with the distance the lack of functional capillaries might explain the prolonged time course when compared with infarcted myocardium (14).
Thus, we suggest the term “very delayed enhancement” to describe the prolonged time course of RF lesion enhancement compared with the observations in myocardial infarction that occurs significantly earlier (5,6). Increase in volume of distribution and regional differences in wash-in/wash-out time constants have been proposed as possible mechanisms of hyperenhancement in infarcted tissue (5,15–17). An accumulation of gadolinium within the area of hyperenhancement was recently shown by Rehwald et al. (17), who demonstrated an increased gadolinium concentration of 235 ± 24% in acute infarct and of 472 ± 78% in chronic infarct measured directly with electron probe x-ray microanalysis. On the other hand, Kim et al. (5) were able to show that in 13 rabbits the differences in regional enhancement were mostly due to changes in regional microsphere-determined flow.
In our experiments, gadolinium imaging slightly overestimated the pathological lesion size (and therefore underestimated the gap size) during the different phases of delayed enhancement. Similarly, several MR studies overestimated the size of myocardial infarction by 10% to 20%, which was contributed to an increased accumulation in the border zone (7,16,18,19). However, detailed ex vivo studies using ultra-thin MRI (500 μm) have demonstrated a good correlation of delayed hyperenhancement with the area of myocardial necrosis (8). Additionally, direct measurements of gadolinium in the adjacent myocardium of risk that failed to show a significant increase in gadolinium contents (17). Therefore, overestimation in lesion size of myocardial infarction and RF lesions is possibly the result of partial-volume effects caused by slice thickness requirements inherent to in vivo imaging (8).
During further follow-up, the peripheral enhancement of the RF lesions was increasingly lost (“shrinking lesion”) as well as the overall signal decreased (“less bright lesion”). However, lesions remained visible for up to 10 h. Wash-out studies in myocardial infarction similarly described a peripheral decrease in enhancement. However, a significant loss of hyperenhanced regions except for centrally located patches has been described as early as 25 min after gadolinium injection (5). The differences in the time course can likely be attributed to the histologic changes in RF lesions, resulting in a slower diffusion-only process out of the ablated area.
Experiences with gadolinium-enhanced imaging of RF lesions have been very limited. In the only animal study Lardo et al. (9) reported a 12 min follow-up of RF lesions in the right ventricle of six mongrel dogs. He described hyperenhancement of the lesions after 1 min with an SI of 1.55 ± 0.16 times the preinjection values and gradual signal decrease during the 11 min of follow-up. These data differ from our observations and might be the result of a partial-volume effect and a difference in resolution. Although our study used a high-resolution coil with a slice thickness of 3 mm for an RF lesions size of 6.4 ± 2.7 mm, Lardo et al. (9) used a cardiac phase array with a slice thickness of 7 mm for an average lesion of only 3.4 ± 2.1 mm. This may have prohibited a more detailed evaluation of regional differences within the lesion. Additionally, a smaller lesion size may have facilitated a more rapid diffusion of gadolinium into the lesion.
The only other report is a case study showing delayed enhancement in an area of previous RFA for idiopathic left ventricular tachycardia in an 45-year-old patient (20).
Although our findings differ substantially from the observations in gadolinium-enhancement after myocardial infarction several similarities can be found with MRI of RFA in noncardiac tissue. After RFA of prostate cancer in 10 patients, lesions appeared as hypointense foci immediately after gadolinium injection (21). After ablation of liver metastasis, Goldberg and Dupuy described a lack of enhancement immediately after gadolinium injection, but a rim-like enhancement in delayed images (22). In multiple animal studies, RF lesions in liver, lung, brain, and pancreas displayed a characteristic pattern of nonenhancing center with peripheral rim enhancement (23–26). Lesion size correlated well with the pathological specimen (r = 0.91 to 0.93) (23) and overestimated the lesion size by 0.8 ± 1.2 mm (23) to 2.2 ± 1.0 mm (27), which may suggest that the histology changes rather than the type of tissue determine the MRI characteristics.
Several limitations have to be considered related to this study. Although most of the clinical RF lesions are created endocardially, epicardial lesions were chosen in this study to allow controlled lesion creation and guarantee lesion creation over a wide variety of energy as its evaluation was on of the goals of the study. While the macroscopic and histological appearance of the ablation lesions in our study was similar to those described for endocardial ablations (12), we cannot rule out a possible effect on imaging characteristics.
Gadolinium injection and MRI was performed 2.1 ± 0.6 h after RFA. In our study, this was due to the transfer time and the necessary setup in the MRI scanner. However, clinical procedure times for complicated ablations frequently are several hours; thus, if MRI is used for postprocedural assessment, an RF lesion age of 2 h is realistic.
Histological changes occurring during the days and weeks after ablation (12) and may influence the wash-in and wash-out kinetics. In our experiments, no differences in the enhancement pattern were observed between different experiments, nor did other RF studies of longer follow-up (24). However, future studies will have to address this issue.
Recent advances in ablation strategies advocate an anatomic rather than electrophysiological approach toward ablation of, e.g., isthmus-dependent flutter, atrial fibrillation, and nonidiopathic ventricular tachycardia (2,3). Although current catheter guiding techniques have only a limited ability to demonstrate the location and size of an ablation lesion, gadolinium-enhanced MRI could assess the ablation results after complex RFA and potentially provide an end point for these anatomically guided procedures. In interventional MRI/fluoroscopy units, MR images could be used to assess incremental progress during the procedure. The new generation of electrophysiologic mapping systems with image integration capability could provide images of failed ablations (e.g., gaps in ablation lines) and direct repeat procedures. Finally, with a potential real-time MRI system continuous information about the cardiac anatomy and the ablation lesions could be obtained.
This study showed that some significant differences exist between the pattern of delayed enhancement of necrotic tissue in myocardial infarction and after radiofrequency ablation. However, the knowledge about these dissimilarities may allow—as in this study—to reliably evaluate the ablation lesions throughout the different phases of enhancement. Additional studies are now warranted to further define its clinical applicability in electrophysiology and possibly open another application of MRI in the field of cardiology.
This work has been supported by a grant of the AHA and the NHLBI (HL64795).