Author + information
- Received August 28, 2007
- Revision received October 3, 2007
- Accepted October 4, 2007
- Published online January 1, 2008.
- Timm Dickfeld, MD, PhD⁎,†,1,⁎ (, )
- Peng Lei, MA⁎,
- Vasken Dilsizian, MD⁎,
- Jean Jeudy, MD⁎,
- Jun Dong, MD, PhD†,1,
- Apostolos Voudouris, MD⁎,
- Robert Peters, MD⁎,
- Magdi Saba, MD⁎,
- Raj Shekhar, PhD⁎ and
- Stephen Shorofsky, MD, PhD⁎
- ↵⁎Reprint requests and correspondence:
Dr. Timm Dickfeld, Department of Cardiology, University of Maryland, 22 South Greene Street, Room N3W77, Baltimore, Maryland 21201.
Objectives This study sought to assess the feasibility of deriving 3-dimensional (3D) scar maps from positron emission tomography (PET)/computed tomography (CT) hybrid imaging and to integrate those into clinical mapping systems to assist in ventricular tachycardia (VT) ablations.
Background Ablation strategies for nonidiopathic VT are increasingly based on the anatomic information of the scar and its border zone. However, the current “gold standard” of voltage mapping is limited by its inability to accurately describe a complex 3D scar morphology, its imperfect spatial resolution, and prolonged procedure times.
Methods Fourteen patients underwent PET/CT multimodality imaging before the VT ablation. We used PET/CT-derived scar maps to characterize myocardial scar using a 17-segment analysis and surface reconstruction. In 10 patients, reconstructed 3D metabolic scar maps were integrated into a clinical mapping system and compared with high-resolution voltage maps.
Results A good correlation was found between the voltage maps and PET/CT-derived scar maps (r = 0.89; r < 0.05). In addition, 3D metabolic scar maps accurately displayed endocardial and epicardial surface and could be successfully integrated with a registration error of 3.7 ± 0.7 mm. A combination of visual alignment and surface registration was most accurate for myocardial scar accounting for ≤15% of the left ventricular surface. Scar size, location, and border zone accurately predicted high-resolution voltage map findings (r = 0.87; p < 0.05). Integrated scar maps revealed metabolically active channels within the myocardial scar not detected by voltage mapping and correctly predicted non-transmural scar despite normal endocardial voltage recordings. Areas of low voltage within wall segments displaying preserved metabolic activity were shown to be due to suboptimal catheter contact and prevented unnecessary ablation lesions.
Conclusions We found that PET/CT fusion imaging is able to accurately assess left ventricular scar and its border zone. The integration of a 3D scar map into a clinical mapping system is feasible and may allow supplementary scar characterization that is not available from voltage maps. This technique could significantly facilitate substrate-based VT ablations.
Patients with internal cardiac defibrillators present increasingly with frequent and appropriate shocks for ventricular tachycardia (VT) (1,2). Because the use of antiarrhythmic medications frequently is limited by side effects and decreasing long-term efficacy, radiofrequency ablation of VT is required in many of these patients (3–5).
Three-dimensional (3D) mapping systems have facilitated “substrate modification” approaches, which are required in up to 90% of patients because of hemodynamically unstable VT (3). These systems consist of placing tangential or radial ablation lines along or across the scar border to interrupt “exit sites” or “slowly conducting channels” (3,6,7). Therefore, an exact intraprocedural knowledge of the scar geometry and scar border is extremely important for the correct placement of these curative lesions.
However, the current “gold standard” in electrophysiology to define scar relies solely on bipolar endocardial voltage measurements of <0.5 mV (voltage mapping) to define the myocardial scar (3,6,7). Unfortunately, voltage mapping has several important limitations. First, extensive mapping is time-intensive. Second, a single endocardial voltage measurement cannot accurately describe a complex 3D scar morphology and differentiate between endocardial and epicardial scar components. Third, suboptimal catheter contact can result in falsely low-voltage measurements, leading to an overestimation of the ventricular scar and unnecessary ablation lesions. Fourth, the spatial resolution of the voltage map is limited to an ∼5 mm catheter mapping area (catheter tip = 3.5 mm, interelectrode distance = 1 mm, and ring electrode = 1 mm), which can make the detection of small, localized scar components difficult. Additionally, time constraints in clinical practice can result in significantly lower mapping density for large portions of the left ventricle (LV).
Hybrid positron emission tomography (PET)/computed tomography (CT) imaging has the potential to address these limitations by obtaining spatially aligned corresponding datasets, which enable complimentary anatomical and metabolic characterization of myocardium in a single imaging session. The integration of PET/CT datasets into clinical mapping systems would allow the intraprocedural display of anatomically and metabolically defined scar anatomy and scar borders to guide substrate-based VT ablations.
Fourteen consecutive patients with internal cardiac defibrillators who were scheduled to undergo VT ablation were enrolled into this study. All protocols were reviewed and approved by the Institutional Review Board at the University of Maryland.
Studies were performed on a Philips Gemini system (Phillips, Best, the Netherlands), and PET images were obtained with the whole-body camera perpendicular to the long axis to create transaxial tomograms. Approximately 1 h before the injection of 18-fluorodeoxyglucose (FDG), fasting patients received an oral load of approximately 25 to 50 g of glucose. A 3- to 4-min attenuation correction scan was performed using a rotating Cesium-137 line source, which allowed averaging the PET emission data over many respiratory cycles, thereby minimizing the potential inaccuracies of attenuation correction due to respiratory motion. Twelve millicuries of FDG was injected, and data were acquired over the course of 25 to 30 min. Quantitative and qualitative myocardial metabolic activity was evaluated using the PET dataset to define myocardial scar. Regional blood flow and sarcolemmal integrity (as reflected by the Na-K-ATPase transport system) was assessed by a preprocedural rest-redistribution thallium scan or rubidium PET. Those images were reconstructed analogously to the FDG-PET images, to define myocardial regions as scarred (FDG to blood flow match pattern), diseased/hibernating/stunned (FDG to blood flow mismatch pattern), and normal.
The CT was performed with the integrated 16-slice multidetector CT system. The scans were contrast-enhanced with 100 ml of 350 Omnipaque (GE Healthcare, Fairfield, Connecticut) using retrospective ECG gating. Scans were 120 kV, 500 to 800 mA, and 0.4-s rotation speed. The acquired data were reconstructed into 10 cardiac phases from 0% to 90%. The CT datasets were processed at a Philips Extended Brilliance Workspace to assess anatomic (wall thinning, intramural calcification, aneurysm) abnormalities consistent with myocardial scar. Then, PET/CT datasets were fused using the internal Phillips Gemini or external custom-made software (8).
Voltage maps were created in all 14 patients by an electrophysiologist who was blinded to the PET/CT datasets. Voltage maps were obtained using a clinical 3D-mapping system using either a 4-mm tip (RPM, Boston Scientific, Natick, Massachusetts; 4 patients) or 3.5-mm tip catheter (CartoMERGE, Biosense, Johnson & Johnson, New York, New York; 10 patients) using a filling threshold of 15 mm. Bipolar electrograms were filtered at 10 to 30 to 400 to 500 Hz. Standard clinical voltage settings of <0.5 mV, 0.5 to 1.5 mV, and >1.5 mV were used for scar, abnormal, and normal myocardium, respectively (3).
Comparison of voltage map and PET/CT
Voltage map and PET/CT images of all 14 patients were sectioned into standardized 17-segment maps, which is consistent with the current American Heart Association guidelines for quantitative comparison (Fig. 1) (9). For additional analysis, the PET images underwent a 3D reconstruction (Emory Toolbox, GE Healthcare), which displayed the metabolic signal intensities analogously to the voltage map and allowed their correlation with the corresponding CT images (Fig. 2).
3D scar map reconstruction and integration
The mapping system used in the first 4 patients did not allow image integration (RPM, Boston Scientific). In the other 10 patients, the axial PET dataset underwent post-procedural DICOM3 formatting analogous to the CT or magnetic resonance imaging (MRI) datasets to allow recognition by the proprietary CartoMERGE software (Biosense Webster, Diamond Bar, California). The PET dataset was transferred from a CD using the CartoMERGE Image Importation algorithm. Using the CartoMERGE Image Processing tool, we adjusted the upper and lower thresholds to color-code LV myocardium, which was determined as viable by a board-certified nuclear cardiologist on the raw dataset using a 50% metabolic activity cut-off. After “seed” placement, the 3D cardiac model was extracted using the CartoMERGE software (10). Right ventricular and noncardiac structures were removed with the CartoMERGE clipping tools. Using the CartoMERGE export function, the 3D model was saved for the clinical study.
After creating an LV electroanatomical map, the exported study was uploaded with the regular CartoMERGE algorithm. “Registration,” which describes the process of superimposing the extracted PET scar map with the electroanatomical map, was performed either by obtaining 3 matching landmark pairs (at a 6- and 12-o’clock position of the mitral valve and apex) or by using visual alignment after approximating both maps (using 12-o’clock position of the mitral valve). Several different registration strategies were evaluated by performing either landmark point registration alone, landmark point registration with surface registration (SURF), visual alignment, and visual alignment with SURF. The registration quality was evaluated by visual assessment of the voltage map/3D scar map alignment, internal summation statistics, and individual point-to-shell distances. Area and distance measurements on the voltage and 3D scar maps were performed using the internal CartoMERGE software.
Statistical analysis was performed using SPSS for Windows (release 10.07, SPSS Inc., Chicago, Illinois). Continuous variables are expressed as mean ±1 SD unless noted otherwise. Comparisons were performed using a 2-tailed, 2-sample t test (analysis of variance). The Pearson correlation coefficient was performed to assess possible correlations. Differences were considered significant at a level of p < 0.05.
Patient characteristics are shown in Table 1. All 12 patients with ischemic cardiomyopathy had a transmural infarction. The 2 patients with nonischemic cardiomyopathy had an inferior, transmural scar and a nontransmural lateral scar, respectively.
Endocardial maps were created in all 14 patients with 194 ± 32 mapping points. Endocardial voltage mapping in the 12 patients with ischemic cardiomyopathy-delineated homogenous myocardial scar in the anterior (n = 4), anteroseptal (n = 2), inferior (n = 4), and lateral wall (n = 2), whereas nonischemic patients showed inferior (n = 1) or no endocardial scar (n = 1). In the last patient additional epicardial mapping was performed (421 points), which demonstrated an epicardial lateral scar.
We successfully performed PET/CT imaging and image fusion in all 14 patients; the use of PET/CT allowed the preprocedural identification of the LV scar. Scar, as defined with PET, correlated well with wall thinning on the CT images (r = 0.87; p < 0.01). Aneurysm and intramyocardial calcification were each observed in 2 patients, respectively. Fused PET/CT images revealed metabolically viable myocardium within areas of wall thinning in 3 patients (Fig. 1). The presence of intramyocardial calcium detected by CT predicted a lack of metabolic activity by PET and scar by voltage map in all segments.
Comparison voltage map versus PET/CT
Areas defined as scar (<0.5 mV) by voltage map corresponded on PET/CT images to a metabolic signal count of 42 ± 7% (range 31% to 56%) and demonstrated myocardial wall thinning or lack of contractility on CT in 92% (Fig. 2). Abnormal myocardium with a voltage of 0.5 to 1.5 mV exhibited a metabolic count of 67 ± 15% (range 48% to 92%) and displayed a variable degree of wall thinning. Healthy myocardium (>1.5 mV) demonstrated a metabolic count of 86 ± 12% (range 69% to 100%) and showed normal wall thickness in 97%. A threshold of 50% metabolic activity had a sensitivity of 89% and a specificity of 93% to predict myocardial scar defined by voltage.
The scar surface area, location, and geometric scar shape correlated well between the voltage map and PET data (Fig. 2). Voltage mapping demonstrated a myocardial scar area of 32.6 ± 22.1 cm2 with an LV scar burden of 11.7 ± 23.4%, whereas the PET-defined scar measured 29.1 ± 24.0 cm2 with a scar burden of 10.0 ± 21.4% (r = 0.89, p < 0.05). Maximum and minimum distance across scar area was 7.9 ± 3.1 cm versus 6.8 ± 2.7 cm and 3.8 ± 1.7 cm versus 3.6 ± 1.5 cm for the voltage map and metabolic scar map, respectively (r = 0.89 and r = 0.87).
The CT criteria (wall thinning/contractility) correlated less well with scar area and diameter (r = 0.74 and r = 0.77, respectively). In 2 of the patients with an inferior myocardial infarction, a channel of metabolically alive tissue was found along the mitral valve ring, which had not been demonstrated by the voltage map (Fig. 3).
Integration of 3D scar map (n = 10)
Transfer, extraction, and segmentation of the LV from the PET datasets were successfully performed in all 10 cases using the CartoMERGE mapping system. The resulting 3D scar map displayed the endocardial and epicardial surface and defined the myocardial wall and thickness (Fig. 3). Because the current version of CartoMERGE only allows a binary display (absent or present voxel), myocardial scar had to be displayed as absent myocardium (i.e., “hole in the wall”). However, this approach accurately reflected the location and size of the myocardial scar and border zone as assessed by the reading of a nuclear cardiologist.
In 4 of 10 patients, 4.1 ± 2.3 abnormal voltage measurements (0.6 ± 0.3 mV) within metabolically normal myocardium were shown on repeat measurement to have voltages >1.5 mV despite initial radiological appearance suggesting good catheter contact. Consistently good registration results could be achieved with visual alignment resulting in a position error of 3.7 ± 0.7 mm (Table 2). The only individual point-shell distances ≥10 mm were observed within the myocardial scar as they were measured to the scar map border zone (Fig. 4). The addition of SURF improved the registration accuracy to 3.3 ± 0.3 mm in patients with <15% of scar burden. However, the large infarcts in Patients #5, #7, and #8 resulted in significant wall defects in the integrated scar maps, which impaired the accuracy of the surface registration algorithm (overall position error of 4.4 ± 2.1 mm).
Landmark point registration alone demonstrated a position error of 11.2 ± 3.9 mm using apical and mitral valve points. The addition of SURF improved the position error to 8.0 ± 3.5 mm (Table 2). Tilting errors of up to 15 mm, long-axis shifts of up to 11 mm, and rotational errors of up 13 mm could be minimized by adding visual alignment as part of the registration method.
After registration the voltage map-defined scar size and borderzone demonstrated an overall good correlation with 3D scar map in patients with endocardially detected scar (r = 0.86, r = 0.87, respectively; p < 0.05) (Fig. 5). In these patients, scar voltage map points (<0.5 mV) were within the metabolic scar area in 89%, with an additional 8% being within an adjacent 1- to 1.5-cm border zone. Voltage measurements within the PET-defined scar area demonstrated a voltage amplitude of only 0.3 ± 0.12 mV compared with 7.9 ± 5.4 mV in the area with normal metabolic activity (p < 0.001).
Supplementary scar characterization
Metabolically viable myocardium was observed crossing an area of lateral myocardial scar as defined by voltage mapping. Despite repeat attempts by 2 electrophysiologists and transseptal as well as retrograde aortic access, no endocardial amplitude of >0.5 mV could be recorded (Fig. 6). Additionally, the 3D metabolic map diagnosed myocardial scar despite a normal endocardial voltage map. Epicardial mapping confirmed a large, nontransmural inferobasal scar, which was consistent with the exit site of the observed clinical VT (Fig. 7).
The novel findings of this study are that: 1) PET/CT fusion imaging can accurately predict location and extent of LV myocardial scar when compared with the current “gold standard” of voltage mapping; 2) current image integration systems can be modified to accept, register, and display 3D volume datasets; and 3) integration of PET scar maps can provide complimentary scar characterization not available from conventional, endocardial voltage mapping.
Cardiac imaging of myocardial scar
We used PET or PET/CT imaging for scar imaging in this study because they are not affected by the same restrictions as MRI or CT alone. Although delayed-enhanced MRI has been well-established to visualize scar tissue (11,12), defibrillators, metal artifacts, and the association with nephrogenic systemic sclerosis limit its use in the VT patient population (13–15). Similarly, although scar imaging with delayed-enhanced CT has been described, its applicability in chronic human infarcts is still evolving (16,17).
Positron emission tomography has been extensively validated for the detection of myocardial scar (18,19), and its prognostic value for cardiac interventions and cardiac surgery has been shown repeatedly (20,21). This study extends this observation and showed that PET and PET/CT imaging can accurately predict the location of nonviable (scar) versus viable LV myocardium. The combination of PET and CT is especially attractive in this context because the submillimeter resolution of the CT images can further improve the spatial resolution of the combined imaging data set.
Image integration of a 3D scar map
This study demonstrates the ability to integrate 3D volume datasets into a clinical mapping system. Current image integration systems only extract a “paper-thin” surface reconstruction representing the blood-endocardium interface (10). However, a clinically useful 3D scar map requires a “volume display” of the LV wall showing the endocardial/epicardial/myocardial extent of the LV scar. As the current CartoMERGE software allows only a binary display, scar had to be represented as absent voxels resulting in a “hole.”
The available registration tools, especially visual alignment and surface registration, were sufficient to provide a registration accuracy of 3 to 4 mm, which is in the range of previous image integration studies (10,22–27). Surface registration was not useful in patients with a large myocardial scar, which resulted in large LV wall “holes” for which the current SURF algorithm cannot compensate.
After successful registration, this study found a good correlation between 3D scar map and voltage map, which has the potential to guide VT mapping and ablation. In our experience, a significant advantage of the scar map integration was the immediate recognition of falsely low-voltage recordings in areas of normal metabolic activity. In all circumstances, they were shown to have resulted from suboptimal catheter contact and thus prevented unnecessary and possibly proarrhythmic ablation lesions. Additionally, the scar map was able to provide information about the location and extent of the myocardial scar with a current spatial resolution of approximately 5 to 6 mm. This accuracy could facilitate the detection of myocardial scar, which might be close to the highest possible resolution of a 3.5-mm catheter tip and allowed the electrophysiologist to concentrate on scar areas and border zones, which were thought to participate in the VT.
With the use of a metabolic definition of cell survival, PET may provide a supplementary and possibly more sensitive way of assessing viable myocardium than the electrically based voltage mapping. Indeed, in this study, 2 patients with inferior infarct on the 3D scar map demonstrated channels of viable myocardium along the mitral valve annulus, which were not detected with voltage mapping and can present successful ablation sites (28).
In another patient, PET displayed a channel of viable myocardium crossing a lateral scar, which appeared homogeneously dense by voltage mapping. This discrepancy may represent endocardial scar, which results in low-voltage recordings, but still has significant epicardial or mid-myocardial surviving tissue that can be detected with PET. In addition, the use of PET enabled us to detect nontransmural scarring in an area that appeared normal during endocardial voltage mapping and could only be confirmed with an epicardial mapping approach. Therefore, this technique may assist in planning the ablation strategies.
Two different clinical systems were used for the acquisition of the voltage maps. Although the use of 2 systems may have resulted in some technical variation, no obvious differences were appreciated. Second, the current CartoMERGE software does not contain quantitative tools to standardize the threshold selection, which defines myocardial scar. This limitation may lead to significant interobserver and intraobserver variability. Currently, the CartoMERGE software is limited to a binary display presenting scar as a “hole” in the wall. Additionally, it is not possible to process fused datasets or to combine 2 different imaging modalities. Appropriate software solutions are currently being developed. Finally, this study evaluated the correlation and feasibility of 3D scar map integration. Although some immediate clinical benefits were observed, this study did not systematically investigate the clinical outcomes of PET/CT integration.
The findings of this study suggest that the use of PET/CT fusion imaging can enable accurate assessment of LV scar and its border zone. The integration of a 3D scar map into a clinical mapping system is feasible and allows supplementary scar characterization that is not available from endocardial voltage maps. This difference could significantly facilitate substrate-based VT ablations.
↵1 Drs. Dickfeld and Dong are receiving research funding from Biosense Webster.
This study was supported by a Scientist Development Grant (0635304N) by the American Heart Association.
Jeroen J. Bax, MD, PhD, served as Guest Editor for this article.
- Abbreviations and acronyms
- computed tomography
- left ventricle/ventricular
- magnetic resonance imaging
- positron emission tomography
- surface shell registration
- ventricular tachycardia
- Received August 28, 2007.
- Revision received October 3, 2007.
- Accepted October 4, 2007.
- American College of Cardiology Foundation
- Marchlinski F.E.,
- Callans D.J.,
- Gottlieb C.D.,
- Zado E.
- Soejima K.,
- Suzuki M.,
- Maisel W.H.,
- et al.
- Stevenson W.G.,
- Friedman P.L.,
- Kocovic D.,
- Sager P.T.,
- Saxon L.A.,
- Pavri B.
- Shekhar R.,
- Walimbe V.,
- Raja S.,
- et al.
- Cerqueira M.D.,
- Weissman N.J.,
- Dilsizian V.,
- et al.
- Dong J.,
- Calkins H.,
- Solomon S.B.,
- et al.
- Kim R.J.,
- Fieno D.S.,
- Parrish T.B.,
- et al.
- Lima J.A.,
- Judd R.M.,
- Bazille A.,
- Schulman S.P.,
- Atalar E.,
- Zerhouni E.A.
- Faris O.P.,
- Shein M.
- Nazarian S.,
- Roguin A.,
- Zviman M.M.,
- et al.
- Marckmann P.,
- Skov L.,
- Rossen K.,
- et al.
- Gerber B.L.,
- Belge B.,
- Legros G.J.,
- et al.
- Czernin J.,
- Porenta G.,
- Brunken R.,
- et al.
- Yoshida K.,
- Gould K.L.
- de Silva R.,
- Yamamoto Y.,
- Rhodes C.G.,
- et al.
- Perrone-Filardi P.,
- Bacharach S.L.,
- Dilsizian V.,
- et al.
- Dickfeld T.,
- Calkins H.,
- Zviman M.,
- et al.
- Reddy V.Y.,
- Malchano Z.J.,
- Holmvang G.,
- et al.
- Wilber D.J.,
- Kopp D.E.,
- Glascock D.N.,
- Kinder C.A.,
- Kall J.G.