|Home | About | Journals | Submit | Contact Us | Français|
To find evidence of diffuse fibrosis in dilated cardiomyopathy (DCM) patients by comparing measurements on clinical late gadolinium enhancement (LGE) cardiovascular magnetic resonance (CMR) studies between DCM and healthy subjects.
LGE-CMR and the Look-Locker images from 20 DCM patients and 17 healthy controls were analyzed. Blood SNR, myocardium SNR, and blood-to-myocardium CNR were measured on the LGE-CMR images. The optimal TI to null blood and myocardium was determined on the Look-Locker images. The post-contrast T1 was estimated using a phantom study which correlated optimal TI and heart rate to T1.
The blood SNR was lower, myocardium SNR was higher, and the blood-to-myocardium CNR was lower (6.6 ± 0.7 vs. 10.3 ± 0.9, p = 0.004) on DCM LGE-CMR images, as compared to controls. The blood-myocardium optimal TI difference (ΔTI) was lower (38 ± 2 ms vs. 55 ± 3ms, p < 0.001) in DCM, and the estimated blood-myocardium T1 difference (ΔT1) (116 ± 6 ms vs. 152 ± 8 ms, p = 0.001) was also lower.
DCM patients have reduced blood-myocardium ΔTI and ΔT1, and lower CNR as compared to controls, suggesting the presence of diffuse fibrosis. This may impact the interpretation of LGE data.
Patients with non-ischemic dilated cardiomyopathy (DCM) constitute 26-35% of the heart failure population (1-3). On pathological examination, the heart is usually dilated with increased mass and pale appearing myocardium (4). On microscopic examination, diffuse myocardial interstitial and perivascular fibrosis are evident (4-6). Biopsy studies have shown that DCM is associated with diffuse myocardial fibrosis in 60% of patients (7). Late gadolinium enhancement (LGE) cardiovascular magnetic resonance (CMR) has become an important tool in assessing myocardial fibrosis/scar. Recent studies have shown that the presence of focal LGE in DCM patients pertains a worse prognosis, including increased all-cause cardiac mortality (8,9). In these studies, LGE was considered present when there were focal areas of myocardium demonstrating peak signal intensity of > 2 standard deviations (SD) above mean signal intensity of remote myocardium. However, in the presence of diffuse myocardial fibrosis there may not be remote “normal” myocardium, since the post-contrast T1 time of the entire myocardium may be decreased. Figure 1 shows how diffuse fibrosis, with a shorter T1 than normal myocardium, will result in a shorter inversion time (TI) for optimal nulling and a smaller difference in inversion time between blood and myocardium (ΔTID is smaller than ΔTIN). Because of the shorter TI required to null diffusely fibrosed myocardium, blood signal in the LGE scan acquired with the shorter optimal TI will be reduced (MzD is smaller than MzN). We sought to measure these effects in the DCM population in clinical CMR.
Data from 20 consecutive adult patients (age 57 ± 11 years, 13 (65%) males, and left ventricular (LV) ejection fraction (EF) 29±9%) with clinical diagnosis of DCM referred for assessment of LV function and fibrosis between June 2006 and July 2008 were included in the study. Fourteen patients (70%) had cardiac catheterization demonstrating the absence of epicardial coronary artery disease (CAD). The remaining six patients had no ischemia on perfusion imaging (N=3), negative CT coronary angiogram (N=1), or were young women (N=2, ages 34 and 42 years). Seventeen adults (age 31 ± 13 years, 4 (24%) males, LVEF 63±4%) including 13 who were referred for CMR for other indications (atypical chest pain (1), family history of cardiomyopathy (2), suspected hypertrophic cardiomyopathy (4), or rule out adult right ventricular dysplasia (6)), and four healthy adult volunteers imaged between January 2006 and August 2008 with identical LGE protocols as DCM patients served as controls. All control subjects had no clinical history of myocardial infarction, hypertension, CAD, sudden cardiac death, or valvular heart disease, in addition to having normal LV and right ventricular (RV) volumes and systolic functions on CMR. The institutional Committee on Clinical Investigations approved the study protocol. Written informed consents were obtained from volunteers and waived for existing clinical data sets.
CMR imaging was performed on a 1.5 T Philips Achieva MR scanner (Philips HealthCare, Best, NL), equipped with a commercial 5-element cardiac coil. Breath-hold short-axis steady state free processing images covering the entire LV were acquired as previously described (10). 2D breath-hold LGE images were acquired in the same orientation about 10-20 minutes post injection of 0.2mmol/kg gadolinium-diethylenetriamine pentaacetic acid (Gd-DTPA) (Magnevist, Schering, Germany). Imaging parameters were: 2D spoiled gradient echo inversion recovery, with 160 × 160 matrix, field of view (FOV) 320 × 320 mm2, 8 mm slices, with 2 mm gaps, TR/TE/Flip angle(α) = 4.3ms/1.5ms/20°, partial echo, fat saturation, 1 RR between inversions, 2 signal averages. ECG-gating in late-diastole and breath-holding were used to reduce motion artifacts. TI was determined visually using a 2D Look-Locker sequence (11) immediately prior to LGE imaging, 10 to 15 minutes after injection of contrast agent. Scan parameters were: Breath-hold 2D non-selective inversion recovery multi-gradient echo sequence with echo train length of 9 views, TR/α = 40 ms/15°, 1 RR between inversions. Matrix size was 144 × 144, FOV = 320 × 320 mm2, 10 mm thickness, mid-wall short-axis orientation. After the QRS was detected, the non-selective 180° pulse was applied, followed by the acquisition of multiple cardiac phases in 90% of the RR window (i.e. a 10% trigger window).
Regions of interest (ROIs) were used to measure signal-to-noise ratio (SNR) and blood to myocardium contrast-to-noise ratio (CNR) on the short-axis diastolic mid-wall 2D LGE images (Figure 2 (A and B)). ROIs of 50 mm2 were placed in the septum, anterior, lateral, and inferior walls within the myocardium (9). ROIs of 1000 mm2 were placed in the blood pool, avoiding papillary muscles and trabeculations. Noise was measured as the SD of the signal in the air-space anterior to the chest wall by taking as large an ROI as possible. SNR was calculated as the mean signal multiplied by the correction factor of 0.655 divided by the SD of the noise (12).
The Look-Locker data was used to estimate the true zero-crossing for the blood and myocardial signal (optimal TI). ROIs were placed in the LV blood cavity and in the septal myocardial wall in each phase of the cardiac cycle. The zero-crossing for the signal was estimated using the linear interpolation between the two time points straddling zero signal intensity. For both DCM and control subjects, the Look-Locker sequence was acquired at a time that varied between 10 and15 minutes after injection. The difference in T1 between normal myocardium and blood after injection of 0.2mmol/kg of Gd-DTPA has been shown to be stable between 10 and 15 minutes after injection, and to decrease by 10 ms between 15 and 20 minutes (13). The difference between optimal TI to null myocardium and optimal TI to null blood (ΔTI), and the difference in blood-myocardium T1 (ΔT1) were calculated as they are less dependent on the exact imaging time post Gd-DPTA injection than myocardial TI or T1 itself.
Phantoms, constructed of distilled water and Gd-DTPA at multiple dilutions, were used to calibrate between optimal TI and post-contrast T1 as a function of heart rate. First, the T1 time of the phantoms (300-500 ms) were measured using a conventional inversion recovery spin echo acquisitions with TR = 5 sec. After establishing T1 values, the phantoms where imaged with the identical clinical Look-Locker sequence to measure correlation between optimal TI and T1 as a function of heart rates (from 50 to 90 beats per minute (bpm)). Linear regression was used to estimate post-contrast T1 values for each subject as a linear function of TI and heart rate.
Data are presented as mean ± SD for age, body surface area (BSA), LV end-diastolic volume index (LVEDVI), LV mass index (LVMI), LVEF, and RVEF measurements. All ventricular volumetric measurements were performed in the standard fashion (10). SNR, CNR, TI time, and T1 time are presented as mean ± standard error of the mean (SEM) unless specified (i.e. the data is presented as mean ± SD in table 3 to show the distribution of the values). Data were analyzed using the two-tailed Student's t-test to compare continuous variables; the Wilcoxon rank sum test was used to compare categorical variables. A p-value of <0.05 was considered significant. All statistical analyses were performed with STATA Version 10 (STATcorp, TX, USA).
Table 1 shows the summary characteristics of the DCM group and the control group. The control group was younger, had fewer males, smaller BSA (accounted for by fewer males), lower heart rates, normal biventricular EF, normal LV volume, and normal LV mass. Our DCM group represents a population with an average age of 57 years, two thirds male, moderate to severely depressed LV function, moderate to severely increased LV volume, mildly depressed to normal RV function, and mildly increased LV mass.
Figure 2 (C and D) shows the measured SNR and CNR on LGE images as described above (all values presented in Table 2). Blood SNR and blood-myocardium CNR were significantly lower while the average myocardial SNR was significantly higher in DCM patients. The septum had the highest SNR compared to other walls in DCM patients, although this difference between the walls did not achieve statistical significance. When septum was excluded, the average SNR of the remaining walls (SNR Avg-sep) remained significantly higher in DCM patients as compared to controls. In control patients, there was no significant difference in SNR in all the four walls with the highest SNR in the lateral wall. No focal LGE was identified on any of the LGE images.
Figure 3 shows the relationship between T1 and TI, and the dependence on heart rate. For higher heart rates and longer T1s, the relationship between T1 and TI becomes less linear. For lower heart rates (<70 bpm) and shorter TI s, the relationship between T1 and TI is linear.
Table 3 shows the blood and myocardial TI values for optimal nulling, measured from Look-Locker images as described, along with the extrapolated post-contrast T1 times. The average optimal blood TI (DCM 187 ± 5 ms vs. control 201 ± 6 ms, p= 0.098) was not significantly different. The optimal TI of the myocardium was significantly shorter in DCM patients (225 ± 6 ms vs. 256 ± 8 ms, p = 0.003) and resulted in a significantly lower ΔTI in DCM patients compared to control subjects (38 ± 2 ms vs. 55 ± 3 ms, p<0.001). The distribution of ΔTI in DCM and control subjects is presented in Figure 4. In DCM patients, the post-contrast myocardial T1 trended lower (p = 0.067) and ΔT1 was significantly lower (116 ± 6 ms vs. 152 ± 8 ms, p =0.001) as compared to controls.
In the subgroup of 14 patients with catheterization proven absence of coronary artery disease compared with the same 17 controls, when adjusted for heart rate, gender, LVEDVI, and LVMI, DCM status remained significantly correlated with ΔTI (p = 0.024).
LGE-CMR is widely used to identify and characterize scar patterns in patients with ischemic and non-ischemic cardiomyopathy, and to identify patients at higher risk for adverse outcome(8,9,14-16). The accepted methodology of identifying areas of LGE in DCM patients by signal intensity over 2SD above the mean “remote normal myocardium” in the same slice (8,9) was an extension from identification of focally scarred myocardium in patients with ischemic heart disease (17-20). This method may have limitations in DCM patients since diffuse fibrosis rather than focal fibrosis may be present. Diffuse fibrosis may shorten the post-contrast T1 time of the entire myocardium. Consistent with this hypothesis, our study of 37 subjects demonstrated the cohort of 20 DCM patients had a shorter post-contrast T1 time leading to a shorter optimal TI to null the myocardium, and a lower blood-myocardium CNR.
Two recent studies also examined issues related to diffuse fibrosis in patients with non-ischemic cardiomyopathy and heart failure (21,22). Jerosch-Herold et al. reported that the Gd-DTPA partition coefficient, which reflected the distribution volume of contrast agent in the myocardium, was significantly higher in 9 patients with clinically diagnosed DCM as compared to 9 healthy controls (21). The authors suggested that increased extracellular matrix volume in DCM patients may account for this change (21). Iles et al. used an ECG triggered, inversion-recovery prepared 2D fast gradient echo T1-mapping sequence with variable temporal sampling of k-space and found decreased post-contrast T1 time in 25 heart failure patients when compared to 20 controls (22). Those heart failure patients were heterogeneous with 36% ischemic cardiomyopathy, 36% restrictive/infiltrative cardiomyopathy, and only 7 (28%) DCM patients (22). We studied 20 clinically diagnosed non-ischemic DCM patients (none had CAD), including 70% with angiographically proven normal coronaries. We found a smaller difference in the post-contrast T1 time between our DCM patients and controls (463 ± 21ms vs. 514 ± 17 ms) as compared to findings by Iles et al. (429 ± 22 ms (in segments absent of focal fibrosis in heart failure patients) vs. 564 ± 23 ms (controls)). This difference in patient post-contrast T1 time may reflect different diseases studied, as the patient age, LVEF, and heart rate were similar in the two studies. Their control group was of similar age, LVEF, and heart rate as ours and the post-contrast T1 time difference cannot be explained by these variables. Although there are numerical differences, our finding of shorter post-contrast T1 time in DCM patients is consistent with their findings. Furthermore, we were able to demonstrate the changes in post-contrast T1 time using routine LGE imaging methods. The findings of ΔTI and ΔT1 are particularly robust, as they are less dependent on choosing the optimal inversion time or a consistent imaging time for LGE images (13). Iles et al also showed that decreased post-contrast T1 time was correlated with increased myocardial collagen content in a group of patients who had undergone transplantation (22).
In our simulation shown in Figure 1, we predicted that if the optimal TI was chosen to null the diffusely fibrosed myocardium, the blood SNR (MzD) would be lower and the myocardial SNR would not significantly differ. In our study, we found a lower blood SNR, as expected, but we also found a significantly higher myocardial SNR in DCM patients than control subjects. This surprising result may be due to our less than optimal clinical real time visual estimation of TI from Look-Locker images. There might be a preference to choose a TI that produces images with higher blood signal intensity since these images provide better SNR, and therefore better apparent quality. When the myocardium is not correctly nulled due to using a longer than optimal TI to produce higher blood SNR images, a higher myocardial SNR can be obtained.
In this study, we did not distinguish “positive LGE” of the septum by > 2SD of remote normal myocardium (as there was no “normal” myocardium), but we did find that in this DCM cohort, the septum had the highest SNR (although not statistically significant when compared to other walls). However, the average myocardial SNR without the septum in the DCM cohort was still higher than controls, supporting the notion that although septum may have the highest collagen content, the remaining myocardium may also be fibrotic.
Iles et al (22) have alluded to the role of heart rate on inversion time, CNR, and SNR in LGE images. In our phantom study, we identified a non-linear dependence of the optimal TI time at long T1 times and higher heart rates (above 70 bpm). We were able to use the phantom calibration result to estimate post-contrast T1 from TI values and heart rate data.
Our study has a few limitations. First, the LGE scans were performed in a slightly varied time frame (10-20 minutes post Gd-DTPA injection) by experienced dedicated CMR technologists. While T1 of blood and myocardium change from 10 to 20 minutes, the T1 difference between myocardium and blood (ΔT1) is very stable (13). Therefore, any differences in LGE scan timing should not strongly influence our data. Second, we were not able to gender or age match the DCM and control subjects. However, previous studies also had younger control groups (21,22). We are not aware of gender or age difference in LGE pattern in normal individuals. Limited information in HCM patients demonstrated no gender difference in LGE in HCM (23). Third, we do not have biopsy data to correlate with our findings. Whether or not patients with lower ΔTI and/or ΔT1 have a higher burden of fibrosis and worse outcome would be important to study in the future.
Finally, the measured T1 values may be influenced by LV blood flow and basal shortening, which are reduced in DCM patients. The effect of slow flowing blood was not addressed in other studies measuring LV blood T1 (13,21,22). The signal intensity at each TI value is influenced by the number of RF pulses experienced by the relaxing tissues, and more excitations (due to lower velocity) will result in a shorter measured T1. This in-flow effect may partly explain the lower blood T1 estimated in DCM subjects (Table 3) and to a lesser extent might reduce the myocardial T1 in DCM subjects compared to controls. The overall effect would be reduction of the true difference in the ΔTI measured in DCM vs. controls.
In conclusion, using LGE methods, DCM patients have significantly shorter blood-myocardium ΔTI and post-contrast blood-myocardium ΔT1, providing evidence for the presence of diffuse fibrosis. This leads to decreased blood-myocardium CNR as a result of a shorter optimal inversion time. The detection of non-discrete fibrosis is a new paradigm for LGE-CMR, which may impact the interpretation of LGE data in DCM patients.
Grant Support: American College of Cardiology Foundation (to YH); Beth Israel Deaconess Medical Center and the Clinical Investigator Training Program: Beth Israel Deaconess Medical Center – Harvard/MIT Health Sciences and Technology, in collaboration with Pfizer Inc. and Merck & Co. (to YH); American Heart Association Scientist Development Award (to DCP); NIH NIBIB Career Research Award K01 (to DCP); Philips Medical Systems Research Grant (to WJM).