Non-Echocardiographic Imaging in Evaluation for Cardiac Resynchronization Therapy

In patients with heart failure and prolonged QRS duration, randomized clinical trials have shown that cardiac resynchronization (CRT) is associated with improvement in quality of life, left ventricular remodeling, and survival ^1–3. The improvements are believed to be mediated by more effective synchronized contraction in the presence of a wide QRS, but mechanical and electrical dyssynchrony are not equivalent ^4,5. Although the concept of CRT response remains problematic ⁶, 20–40% of patients who receive CRT based on electrical dyssynchrony criteria (i.e. QRS duration) do not derive symptom improvement or demonstrate reverse remodeling ^7–10. Scar burden ^11–13 and failure to place the left ventricular (LV) pacing lead at the site of latest onset of contraction^14–17 have been linked to a poor response. Thus, optimal clinical decision making for CRT must include a comprehensive evaluation of all these factors to identify heart failure patients who will benefit.

The standard echocardiographic parameters of LV mechanical dyssynchrony have been extensively reviewed,¹⁸ with more than 600 published papers. Most of this work has been done using tissue Doppler imaging, with more recent work using speckle tracking ^19,20, three-dimensional echocardiography ²¹, echo-contrast imaging ²², and intra-cardiac echocardiography ²³. Despite the important benefits of high temporal resolution, success in individual centers, and ability to assess the impact of scar burden and concordance of LV lead with latest activation site,¹⁴ fundamental limitations of tissue Doppler imaging include the inability to measure over a sufficient number of cardiac cycles to overcome beat-to-beat variation, poor image quality, and measurement error ²⁴. In the only randomized trial of CRT in patients with wide QRS ²⁵, the failure of 12 different echocardiographic dyssynchrony parameters to improve outcome is most-likely related to the large inter- and intra-observer variability (4–24% and 7–72%, respectively). These limitations of echocardiography have led to a search for non-echocardiographic imaging techniques to optimize decision making prior to CRT (Table 1).

Table 1.

Comparison of different imaging modalities of CRT

	Echo	CCT	CMR	SPECT
Technical characteristics
Reproducibility	+	+++	+++	+++
Temporal resolution	20–30 ms	60–165 ms	20–40 ms	15–45 ms
Spatial resolution	+++	+++	+++	+
Contrast-to-noise ratio	++	+++	++++	++
Automated software	++	++	++	+++
Simple to use	++	++	+	+++
Time for analysis	~ 10–15 min	~10–15 min	~45–60 min	~1–3 min
Number of data points	16 segment model	540 data point	16 to 600	>600 data points
3-dimensional data	3D TTE	Yes	Yes	Yes
Affected by R-R variability	+++	++	++	+
Performed post-hoc on prior studies	proper frame rate	Prospective gating	Tagged sequence	At least 8 frames/cycle
Ionizing radiation	No	Yes	No	Yes
Availability	++++	++	+	+++
Cost	$	$$	$$$	$$
Fusion of modalities	In evolution	Yes with PET/SPECT	Yes with SPECT	Yes with CT and CMR
Clinical utilities
Scar burden	++ (strain)	++	++++	+++
LV EF, volumes	+++	+++	++++	+++
RV volumes	++ (3D)	+	+++	−
Mitral regurgitation	+++	+	++	−
Simplicity of interpretation	>16 indices	1–3 indices	~5–7 indices	2 indices
Latest site of activation/regional	+++	+	+++	+++
Coronary vein	−	+++	++	−
Post-CRT follow-up	+++	++	−	+++
Clinical experience	++++	+	++	+++

CCT (cardiac computed tomography); CMR (cardiac magnetic resonance imaging); CRT (cardiac resynchronization therapy); EF (ejection fraction); LV (left ventricle); PET (positron emission tomography); SPECT (single photon emission computed tomography); TTE (transthoracic echocardiogram); RV (right ventricle)

Computed Tomography

The role of cardiac computed tomography (CCT) in heart failure has been recently reviewed ²⁶. The technique assesses scar location and burden ²⁷, anatomic location of the phrenic nerve, cardiac venous anatomy ²⁸, LV function, and dyssynchrony indices ⁵ (Table 1, Figure 1).

Figure 1. Non-echocardiographic imaging of CRT.

Scar burden, LV function and mechanical dyssynchrony are visualized with different imaging modalities (panel A-J). Panels A and B show large area of delayed enhancement (scar) (white arrow) by CCT and CMR respectively. Panel C is a polar map of a rest MPI showing large fixed perfusion defect size. Panels D-F represent quantification of LV volumes and EF with CCT, CMR and gated SPECT, respectively. Panel G shows the method of assessing mechanical dyssynchrony with CCT using time to maximal LV wall thickening curves (reproduced with permission from Truong QA et al ⁵). Panel H is tagged CMR which is used to derive strain and dyssynchrony indices. Panels I and J are phase histogram and polar map showing segmental onset of mechanical activation with LV site of latest activation (star, inferior wall), respectively. Panel K is a three-dimensional volume rendered image of the coronary venous anatomy by CCT while planning LV lead placement (* side branches of the posterior vein of the left ventricle) (Reprinted with permission from Van de Veire et al. ²⁸). Panel L is a reconstructed 3D-CMR showing the coronary sinus. CS (coronary sinus); CCT (cardiac computed tomography); CMR (cardiac magnetic resonance); CX (left circumflex coronary artery); EF (ejection fraction); GCV (great cardiac vein); LMV (left marginal vein); LV (left ventricle); PIV (posterior interventricular vein); PVLV (posterior vein of the left ventricle); RCA (right coronary artery); SCV (superior cardiac vein).

Scar location and burden

The role of CCT in assessing scar burden is increasing, but still faces many challenges. Dual-phase evaluation has been shown to detect scar after acute and chronic myocardial infarction ^29,30 (Figure 1-A) and correlates well with gadolinium delayed enhancement with cardiac magnetic resonance imaging (CMR), although the contrast-noise ratio is far superior with CMR ³¹. However, there are no published studies evaluating the impact of scar burden by CCT on CRT response.

Coronary venous anatomy

CCT has an advantage over other techniques in assessment of the coronary venous system (Figure 1-K) which correlates well with conventional catheter-based venography (r = 0.82–0.95) ^28,32 and can register venous anatomy to the site of latest activation ³³. This is particularly important as LV lead placement for CRT is technically challenging in dilated ventricles with prominent tortuousity, stenosis, acute angulation of the coronary veins ^28,34, and especially in the 5% of subjects in whom the posterior vein of the LV and the 39% among whom the left marginal vein are absent ³⁵. In addition, CCT identifies the relationship of the left phrenic neurovascular bundle to the target vein which allows the operator to avoid diaphragmatic stimulation ³⁶.

Mechanical dyssynchrony and CRT response

Little work has been done on mechanical dyssynchrony indices with CCT and with no published data on prediction of CRT response. Recently, three global indices have been described in a small study (N=38) using the following 3 parameters (Figure 1-G); 1) Standard deviation of the time from the R wave to maximal wall thickness (computed as the radial distance between the endo- and epi-cardial borders) (~ 540 data points per patient); 2) standard deviation of time to maximal wall motion (using the endocardial borders and centerline algorithm) (~ 480 data points); and 3) standard deviation of time to minimal systolic area (~90 data points). For segmental dyssynchrony, the average of the maximal difference in time-to-maximal wall thickness or wall motion between each of the 3 pairs of opposing walls were derived (average 180 and 160 data points per patient, respectively). Overall, the reproducibility of global dyssynchrony was much better than regional parameters (intraclass correlation coefficient range 0.71–0.95 versus 0.06–0.91, respectively). Among the global indices, the standard deviation based on time to maximal wall thickening was the most reproducible (intra and inter-observer reproducibility of 0.95 and 0;94, respectively (P<0.0001), with no systematic bias by Bland Altman analysis) ⁵.

The indices correlated moderately with 2D and 3D echocardiography in 14 patients (r = 0.65, p = 0.012 and r = 0.68, p =0.008). However, there was no bias evaluation performed ⁵.

Furthermore, the CCT derived dyssynchrony indices were significantly higher in patients with low EF and wide QRS (N=16) than in a control group. The correlation between electrical and mechanical dyssynchrony was fair (r =0.51, p −0.007), which is expected as the two concepts are not equivalent ⁴. However, the dyssynchrony index could not differentiate patients with low EF and wide versus narrow QRS, and did not correlate with LVEF (r=−0.27, p = 0.17). While the small number of patients in each group is probably the limiting factor, these findings need to be validated in larger studies.

Challenges

Temporal resolution remains a major challenge, and is limited to 165 ms with a typical 64-slice CCT scanner. However, with the emergence of new dual source scanners, the effective temporal resolution can be further reduced to 83 and 42 ms with single and muti-segment reconstruction algorithms, respectively, which is comparable to other imaging modalities (Table 1).

The high radiation dose with computed tomography, especially when performing prospective gating to assess wall thickening and dyssynchrony (~15–25 mSv), or with serial scanning to optimize LV lead placement and follow-up post CRT remain major limitations. However, with the fast scanners and radiation lowering techniques, it will be feasible, in the near future, to limit the radiation burden to ~1–5 mSv allowing multiple scans at a fraction of the typical radiation burden used today ³⁷.

Cardiac Magnetic Resonance Imaging

CMR is a well-studied imaging technique with high reproducibility and high spatial resolution, which provides three-dimensional data on LV function and dyssynchrony, allows visualization of the coronary venous anatomy, and provides viability information ^38,39 (Table 1, Figure 1). These features make it useful in CRT planning.

Scar burden and CRT response

Cardiac MRI is now recognized as the gold standard to assess myocardial viability, with excellent spatial resolution and contrast to noise ratio⁴⁰ (Figure 1-B). Not only can it detect small infarcts which may be missed by nuclear SPECT MPI ⁴¹, it also avoids overestimating LV inferior and posterior scar which can occur with SPECT MPI due to attenuation artifacts ⁴². The scar burden can also be quantified ^43,44. A total scar burden ≥33% of LV myocardial volume, scar transmurality ≥51%, or pacing over a posterolateral scar were associated with poor response to CRT ^12,44. Similarly, another small study showed that a cut-off of 15% scar had a 85% sensitivity and 90% specificity to predict CRT response ⁴⁵. Furthermore, a linear relationship between total scar burden by MRI and LV remodeling or response to CRT has been described ^45–47. T1-mapping has emerged as a potential tool to detect and quantify myocardial interstitial fibrosis ⁴⁸, but there have been no studies to see if it predicts response to CRT.

Coronary venous anatomy

Coronary venous anatomy can also be visualized with CMR ^49,50. The contrast between the coronary veins and surrounding myocardial tissue is adequate to assess the location, dimension, tortuosity and branching angles of the coronary veins. The study is performed within 10–15 minutes with free breathing and without any contrast or radiation burden. However, the spatial resolution (typically 0.5 to 1 mm) is not quite as good as with CCT, and the study cannot be routinely performed in patients with implanted cardiac devices⁵¹.

Mechanical dyssynchrony and CRT response

CMR can be used to quantify dyssynchrony using several basic scan acquisition techniques (Figure 1-H; Figure 2). The relevant indices from these methods are summarized in Table 2 ^38,52–67. While some of these indices were shown to predict CRT response, the studied cohorts were small.

Figure 2. CMR for evaluation of dyssynchrony.

A: SPAMM stripe tagging in 4-chamber view in diastole and systole. B, C: SPAMM grid tagging in 4-chamber (B) and short axis planes (C). Scan details: 1.5 Tesla scanner; fast spoiled gradient recalled echo sequence; 7 mm tag spacing; flip angle 10 degrees; slice thickness 8 mm; reconstructed to 20 cardiac phases. D: 3-dimensional map of maximum principal strain in left and right ventricles, reconstructed from short and long axis co-registered tagging CMR. Timing of maximal mechanical activation/contraction can be discerned across the cardiac cycle for each segment of the left and right ventricle, allowing for assessment of both intra and inter-ventricular dyssynchrony. Scale at right represents value of maximal principal strain; shortening strain is a negative number.

CMR (Cardiac magnetic resonance); LV (left ventricle); RV (right ventricle); SPAMM (spatial modulation of magnetization)

Table 2.

Dyssynchrony evaluation by cardiac magnetic resonance imaging

Technique		Characteristics	Advantages	Disadvantages	Dyssynchrony indices	Prediction of CRT response
Cine	SSFP	Wall motion	Myocardial borders well seen	Automated or semi-automated contour detection often requires manual correction Time consuming	Septal-Lateral delay of wall thickening (>65 ms) Radial dyssynchrony (tissue synchronization index)	SN 90%; SP 59% (N=40) (septal-lateral delay) Could not further stratify patients (radial dyssynchrony) (N=225)
Phase contrast	MR-TVM	Tissue velocity mapping	3D velocity info per pixel Velocities used to derive strain Measures Inter-ventricular dyssynchrony	Similar to above Long scan time to acquire images Respiratory artifact	Aorta–Pulmonary onset flow time difference for interventricular dyssynchrony (RV-LV) (ms)	No studies
	DENSE	Encodes position/tissue displacement	Better image contrast than MR-TVM Used to derive strain	Low temporal resolution (one image/cardiac cycle) Not widely available	N/A	No studies
Tagged imaging	SPAMM	Spatial modulation of magnetization	Deformation is quantified into strain	Long processing time (up to several days) Low spatial resolution (5–7 mm) Tagging fades in diastole	Circumferential strain (SD of time to peak systolic strain) Regional variance of strain Regional variance of vector strain Temporal uniformity index (circumferential uniformity ratio estimate)	PPV 87% NPV 100% to improve patient selection (circumferential uniformity ratio estimate) (N=47)
	CSPAM M	Subtraction of two out of phase tagging grids to give improved persistence of tag lines	Tags last longer in diastole	Longer acquisition time Reduced temporal resolution
	HARP	Analysis of tagged images in the frequency domain	Automated “Faster” 2D and 3D-HARP	Up to 45 min processing time
	SENC	Sinusoidal tags applied in the slice plane	fastest post-processing Instantaneous real-time quantitative strain Higher spatial resolution Through plane encoding allows circumferential and longitudinal strain	Not widely available Analysis is complex

CRT (cardiac resynchronization therapy); CSPAMM (complementary spatial modulation of magnetization); DENSE (displacement encoding by stimulated echo); HARP (harmonic phase analysis); MR-TVM (magnetic resonance tissue velocity mapping); NPV (negative predictive value); PPV (positive predictive value); SD (standard deviation); SENC (strain encoded magnetic resonance imaging); SN (sensitivity); SP (specificity); SPAMM (spatial modulation of magnetization); SSFP (steady-state free precession).

Challenges

The broad use of CMR to plan CRT still faces many challenges. There are problems with imaging patients with implanted devices, although this has been recently attempted in well-selected and monitored patients on a case-to case basis ^68,69; however, the resultant artifact remains a major challenge when analyzing dyssynchrony and response post CRT³⁸ (Figure 3). Furthermore, despite recent advances⁷⁰, the analysis process including use of specialized software is still complicated, time consuming, and not fully automated. Like other imaging modalities used in CRT studies, data have been validated in very small cohorts with different indices. Lastly, access to CMR maybe limited since in many areas it is often performed only in large centers. Even when available, not all centers have expertise in the interpretation of synchrony.

Figure 3. Implantable cardiac device and ventricular lead artifact by CMR.

The figure illustrates the artifacts (arrows) that are produced by an implantable cardiac device and ventricular lead in a well-selected and monitored patient who underwent CMR. Can indicates signal void artifact with no image (only black region) near the subcutaneously implanted pulse generator can (Reprinted with permission from Lardo AC et al. ³⁸).

CMR (cardiac magnetic resonance imaging)

Nuclear Imaging

The role of nuclear imaging with gated SPECT MPI in CRT has been recently reviewed and described as a “one-stop shop” to predict CRT response; it provides data on scar burden and location, LV function, LV site of latest contraction, and mechanical dyssynchrony from a single scan^71–74 (Figure 1).

Scar burden, location and response to CRT

The presence, location and burden of myocardial scar have been shown to affect response to CRT^71,75. In a study by Adelstein et al, an inverse relationship was described between the extent of fixed perfusion defect on MPI and absolute or relative increase in LVEF 6 months post-CRT (r = −0.63 and −0.53, P <0.01, respectively) (N=50). Furthermore, patients who responded to CRT had lower global scar burden and scar density near the LV lead versus non-responders^11,15. Also, the extent of scar around the LV lead correlates negatively with improvement in LVEF¹¹, and is associated with no response in 29% of patients with extensive scar at LV lead site despite having concordant lead with latest site of activation¹⁵. Similar findings showed that a transmural scar (<50% tracer activity) at the site of LV lead placement was associated with no response to CRT¹³. These results are concordant with CMR studies as described in the previous section. An advantage of SPECT MPI is the ability to automatically quantify the scar burden with good reproducibility⁷⁶. However, the low spatial resolution and counts of the images remains a limitation, particularly when assessing viability in dilated ventricles with thin walls as it might overestimate extent of scar. Positron emission tomography (PET) images are performed with higher tracer counts, lower radiation exposure, and better spatial resolution, solving the problem to a great extent⁷⁷. However, there are limited data on dyssynchrony or CRT response using PET images.

Coronary venous anatomy

SPECT MPI plays no role in identifying coronary venous anatomy.

Mechanical dyssynchrony

a. Technique characteristics

Nuclear imaging was used to evaluate mechanical dyssynchrony several decades ago, in the era of gated equilibrium radionuclide angiography ^78–81. In recent times, gated single photon computed tomography (SPECT) has quickly emerged as an attractive alternative to quantify dyssynchrony ⁸² (Figure 1). The technique of phase analysis for SPECT MPI has been extensively described by Chen et al using the Emory Tool Box (SyncTool, Emory University, Atlanta, GA)^71,82, with other software in development ⁸³.

Briefly, a three-dimensional count distribution is extracted from each of the LV short axis data sets; a one-dimensional fast Fourier transform is applied to the count variation over time for each voxel, generating a 3D phase distribution that describes the timing of LV onset of mechanical contraction over the entire R-R cycle (Figure 1-I). Two clinically relevant dyssynchrony indices are derived: standard deviation and histogram bandwidth⁸². The normal values have been published and validated ^82,84,85. The technique is fully automated, has effective temporal resolution of ~15 ms for a heart rate of 60/min⁸⁶, inter- and intra-observer reproducibility of 99% ⁸⁷, high repeatability⁸⁸, independent of the type of camera used ⁸⁹, the type of image reconstruction ⁹⁰, or the tracer dose ⁹¹ (Table 1).

b. Dyssynchrony, LV lead placement and CRT response

Although mechanical dyssynchrony is associated with the extent of cardiomyopathy (P =0.01), scar burden (P<0.0001), and QRS duration (P = 0.04)⁹², the correlation with each of these parameters is far from perfect (r =−0.49; 0.50–0.65; 0.25–0.50, respectively) ⁸⁵ (Figure 4), suggesting that mechanical dyssynchrony may provide incremental value. In 42 patients with heart failure, QRS >120 ms, and LVEF <35% undergoing CRT, a cut-off value for SD of 43 ° was shown to have 74% sensitivity and 81% specificity to predict clinical response to CRT and LV reverse remodeling⁹³.

Figure 4. Prevalence of LV dyssynchrony by gated SPECT (phase SD >43°) in patients with severe (EF <35%) and mild-moderate (EF 35–50%) LV dysfunction stratified by QRS duration (ms).

EF (ejection fraction); LV (left ventricle); SD (phase standard deviation); SPECT (single photon emission computed tomography)

Furthermore, in a recently published study of 90 patients undergoing CRT, Boogers et al showed that concordance of latest LV activation by phase analysis with lead placement during CRT implantation was associated with improvement in LV reverse remodeling¹⁵ (Figure 5). The site of latest activation was determined using a 6 segment model ^15,19; the mean phase of every segment was calculated, and the highest value corresponded to the latest activated segment. These segments were located in the posterior (42%), lateral (23%), inferior (13%), anterior (16%) walls (intra and inter-observer agreements of 93% and 87%, K = 0.96 and 0.92, respectively)¹⁵. Conversely, the LV lead position was determined by biplane fluoroscopy (Figure 5). The response rate to CRT was 79% (concordant lead) versus 26% (discordant). In addition, there were 7 patients with extensive scar at latest activation site; excluding those patients, the response was 92%¹⁵. Figure 6 illustrates some examples of CRT response.

Figure 5. Concordance of left ventricular lead placement with latest site of activation.

The figure represents the distribution of the left ventricular latest site of mechanical activation with phase analysis of gated single photon emission computed tomography (6 segment model) and the left ventricular lead placement. There was 58% concordance (N = 90).

Figure 6. Phase analysis of gated SPECT pre and post CRT.

The upper panel shows a responder to CRT. The patient had baseline LV dyssynchrony, and small scar at the apex and inferolateral wall. The LV pacing lead was placed at the posterolateral wall, concordant with the site of latest activation. The patient had favorably responded to CRT. The LV dyssynchrony parameters were reduced immediately post-CRT. This patient had been followed up for >1 year, and showed no endpoint outcomes (cardiac death, heart failure hospitalization, shocks, CRT deactivation).

The lower panel shows a non-responder to CRT. The patient had baseline LV dyssynchrony but no scar. The latest activation site was at the inferolateral wall, however, the LV pacing lead was placed at the anteroseptal wall. The patient had deteriorated LV dyssynchrony immediately post-CRT. This patient had CRT deactivation due to worsened symptoms 15 days post-CRT.

BW (phase histogram bandwidth); EDV (end-diastolic volume); EF (ejection fraction); ESV (end-systolic volume); CRT (cardiac resynchronization therapy); LV (left ventricle); SD (phase standard deviation)

The change of dyssynchrony parameters after CRT occurs immediately after implantation, and may predict long-term LV remodeling⁹⁴. It would therefore be possible, using gated SPECT with single tracer injection, to change CRT parameters to optimize response, especially with the excellent repeatability of the technique to measure dyssynchrony⁸⁸.

Challenges

The major challenges with gated SPECT for CRT are the radiation burden with serial scans when assessing LV remodeling and improvement in dyssynchrony indices post therapy, and the inability to visualize coronary venous anatomy, although the fusion of SPECT-CT imaging could potentially address the latter issue.

Future directions

The search for the best modality for non-invasive cardiac imaging to predict CRT response is still ongoing. The different modalities bring different strengths and weaknesses to this process (Table 1). A universal first step should be to assess LV function; while CMR is the gold standard, other imaging modalities including 3D echocardiography with contrast provide reliable LVEF. It is important to realize that the prognostic benefit of CRT is not based on the prediction of “response” (itself a controversial topic), but in some circumstances, prediction of symptomatic response may be the main consideration driving the decision to implant a device. The identification of significant mechanical dyssynchrony may be performed by detection of the site of latest activation by SPECT or echocardiography with speckle tracking. In patients with ischemic heart disease, CMR is the gold standard for scar quantification, but SPECT is a good alternative. Finally, CCT could be used to assess coronary venous anatomy in selected patients.

Further multicenter studies that evaluate all modalities together or an integrative approach displaying anatomy and function are needed to define which single test or combination of techniques can best guide initial patient selection, procedural approach, and post-implant optimization.

Abbreviations

BW

histogram bandwidth

CCT

Cardiac computed tomography

CMR

Cardiac magnetic resonance

CRT

Cardiac resynchronization therapy

EF

Ejection fraction

LV

Left ventricle

MPI

Myocardial perfusion imaging

PET

Positron emission tomography

SD

phase standard deviation

SPECT

Single photon emission computed tomography