International Journal of Cardiology 181 (2015) 430–436

Contents lists available at ScienceDirect

International Journal of Cardiology journal homepage: www.elsevier.com/locate/ijcard

Improving the diagnosis of LV non-compaction with cardiac magnetic resonance imaging☆ P. Choudhary a,b, C.J. Hsu a,b, S. Grieve a,b,c, C. Smillie f, S. Singarayar a,b, C. Semsarian a,b,d, D. Richmond a,b, V. Muthurangu e, D.S. Celermajer a,b, R. Puranik a,b,⁎ a

The University of Sydney, Faculty of Medicine, Sydney, Australia Royal Prince Alfred Hospital, Department of Cardiology, Sydney, Australia Charles Perkins Centre, The University of Sydney, Sydney, Australia d Agnes Gignes Centre for Molecular Cardiology, Centenary Institute, Sydney, Australia e University College London, United Kingdom f Bankstown Heart Clinic, Bankstown, Sydney, Australia b c

a r t i c l e

i n f o

Article history: Received 30 November 2014 Accepted 21 December 2014 Available online 23 December 2014 Keywords: Left ventricular non-compaction Non-compaction cardiomyopathy Cardiac magnetic resonance imaging Diagnostic criteria Cardiomyopathy

a b s t r a c t Background: Current diagnostic criteria for left ventricular non-compaction (LVNC) poorly correlate with clinical outcomes. We aimed to develop a cardiac magnetic resonance (CMR) based semi-automated technique for quantification of non-compacted (NC) and compacted (C) masses and to ascertain their relationships to global and regional LV function. Methods: We analysed CMR data from 30 adults with isolated LVNC and 20 controls. NC and C masses were measured using relative signal intensities of myocardium and blood pool. Global and regional LVNC masses was calculated and correlated with both global and regional LV systolic function as well as occurrence of arrhythmia. Results: LVNC patients had significantly higher end-systolic (ES) and end-diastolic (ED) NC:C ratios compared to controls (ES 0.21 [SD 0.09] vs. 0.12 [SD 0.02], p b 0.001; ED 0.39 [SD 0.08] vs. 0.26 [SD 0.05], p b 0.001). NC:C ratios correlated inversely with global ejection fraction, with a stronger correlation in ES vs. ED (r = −0.58, p b 0.001 vs. r = −0.30, p = 0.03). ES basal, mid and apical NC:C ratios also showed a significant inverse correlation with global LV ejection fraction (ES basal r = − 0.29, p = 0.04; mid-ventricular r = − 0.50, p b 0.001 and apical r = −0.71, p b 0.001). Upon ROC testing, an ES NC:C ratio of 0.16 had a sensitivity of 70% and a specificity of 95% for detection of significant LVNC. Patients with sustained ventricular tachycardia had a significantly higher ES NC:C ratio (0.31 [SD 0.18] vs. 0.20 [SD 0.06], p = 0.02). Conclusions: The NC:C ratio derived from relative signal intensities of myocardium and blood pool improves the ability to detect clinically relevant NC compared to previous CMR techniques. © 2014 Elsevier Ireland Ltd. All rights reserved.

1. Introduction Left ventricular non-compaction (LVNC) is associated with serious complications such as arrhythmia, systemic thromboembolism and sudden cardiac death [1]. Beyond increased long-term mortality, a diagnosis of LVNC may also influence athletic ability [2] and has

Abbreviations: LV, left ventricle; LVNC, left ventricular non-compaction; CMR, cardiac magnetic resonance; NC, non-compacted; C, compacted; SI, signal intensity; ES, end-systole; ED, end-diastole; ROI, region of interest; SIMYO, signal intensity of compacted myocardium; SIBLOOD, signal intensity of blood; LGE, late gadolinium enhancement; SCMR, Society of Cardiovascular Magnetic Resonance; ANOVA, analysis of variance; LVEF, left ventricular ejection fraction; ROC, receiver operated curve. ☆ Statement of authorship: All authors take responsibility for all aspects of the reliability and freedom from bias of the data presented and their discussed interpretation. ⁎ Corresponding author at: Department of Cardiology, Royal Prince Alfred Hospital, Camperdown, NSW 2050, Australia. E-mail address: [email protected] (R. Puranik).

http://dx.doi.org/10.1016/j.ijcard.2014.12.053 0167-5273/© 2014 Elsevier Ireland Ltd. All rights reserved.

implications for family screening [3]. Diagnostic criteria, however, are controversial [4]. Despite being classified as a “distinct cardiomyopathy” by the American Heart Association [5], there is only limited data that relates the extent of structural left ventricular (LV) derangement to ventricular function or clinical outcomes [6]. In fact recent reports suggest that even LVNC that is defined as “severe” by current methods is clinically benign at long-term follow-up [7]. LVNC is still an “unclassified cardiomyopathy” by the World Health Organisation and the European Society of Cardiology [7,8]. Genetic heterogeneity [9], lack of genotype to phenotype correlation, and the multiple associations of the disease [10–13], complicate attempts at classification. Furthermore, the underlying pathophysiology specifically relating the degree and/or location of LVNC to cardiac dysfunction remains unclear and beyond the scope of the currently most utilised diagnostic criteria. Multiple echocardiographic and magnetic resonance imaging criteria exist, but these are all derived from small data sets of patients with often a severe phenotype. These criteria have poor correlation

P. Choudhary et al. / International Journal of Cardiology 181 (2015) 430–436

with each other [14], vary in their cardiac phase of measurement — i.e. end-systole vs. end-diastole and only provide an opportunity to assess the presence or absence of LVNC in a binary manner. The present clinical conundrum occurs as increasing use of CMR imaging has provided much higher resolution, enabling detailed assessment of the endocardium, particularly at the LV apex. With comprehensive sampling of the entire ventricle, lesser degrees of LV trabeculations are detected. However, there remains no clear method to relate the degree of structural myocardial change directly to cardiac dysfunction or arrhythmia risk. There is also subjectivity in assessing the “worst affected” clinical segment in one of three standard long-axis views that may influence the accuracy of serial measurements using current CMR imaging criteria [15]. We therefore aimed to develop a semi-automated quantification technique using CMR derived signal intensities of myocardium and blood pool, to calculate the ratio of NC to C myocardial mass involving the whole LV, from base to apex. Further, we investigated the functional consequences of NC mass at both global and regional levels to determine its relationship to cardiac function and association with arrhythmia. Finally, we also examined the discriminative value of our technique in both end-systole (ES) versus end-diastole (ED), to optimise our ability to define non-compaction in a clinically meaningful way. 2. Methods 2.1. Patient selection 30 consecutive patients with a diagnosis of isolated LVNC were selected over a fiveyear period from 2008 to 2012. Patients with associated congenital heart disease, hypertrophic cardiomyopathy or abnormal loading conditions were excluded. 20 healthy gender-matched controls were also recruited. Demographic data, clinical status including current symptoms and electrocardiographic information were recorded for each patient. The diagnosis of LVNC was made using a combination of clinical and imaging criteria [15,16]. Of patients that met imaging criteria for LVNC, only patients with an elevated pretest probability such as those with arrhythmia, family history of LVNC or sudden cardiac death, unexplained cardiomyopathy or abnormal echocardiographic findings suggestive of LVNC were included. The study was conducted in accordance with institutional ethical research guidelines. 2.2. Cardiovascular MRI protocol MRI was performed using a 1.5 T MR scanner (GE medical system). 2.2.1. Assessment of ventricular volumes and function using cine MRI 4 chamber and short axis views covering the LV (9–12 contiguous slices) were acquired in the vertical long-axis using retrospectively gated steady-state free precession (FIESTA) cine MR images. Image parameters: TR = 3.2 ms; TE = 1.6 ms; flip angle = 78°; slice thickness = 8 mm; matrix = 192 × 256; field of view-300–380 mm; and temporal resolution = 40 ms, acquired during a breath-hold. Assessments of LV volumes were performed by manual segmentation of short-axis cine images with endocardial outline at end diastole and end systole (performed in Osirix software, version 3.6.1 32 bit). Simpson's rule was used to calculate end-diastolic and end-systolic volumes for the LV;

431

ejection fraction (EF) was calculated from these volumes. Ejection fraction was measured independently by an experienced Level III SCMR accredited consultant in CMR (RP). 2.2.2. Image analysis of compacted and non-compacted myocardium 2.2.2.1. Compacted myocardium. Continuous slices in end-systole and end-diastole through the ventricles as described above were used for subsequent analysis. Regions of interest (ROIs) were traced around the epicardial surface and another one encompassed the endocardial side of the compacted muscle segment. ED and ES compacted mass was derived by summating the respective volumes calculated in each slice and then multiplying by the slice thickness and myocardial tissue density constant — 1.05 g. 2.2.2.2. Papillary muscles. Papillary muscles that came to confluence and were morphologically normal were separately identified and included as compacted mass. The signal intensity of papillary muscle was within the C myocardial range and lower than the preset threshold for NC mass in each case. 2.2.2.3. Non-compacted myocardium (Fig. 1). Signal intensities (SI) of compacted myocardium and blood pool were then measured using Regions of interest (ROIs) of predetermined sample volume size per region, which minimized inter- and intra-operator variability. ROIs analysing the blood pool enclosed areas measured 0.3 cm2, 1 cm2 and 1.5 cm2 for apical, mid and basal slices respectively. Those placed to analyse septal myocardium measured 0.15 cm2, 0.3 cm2 and 0.5 cm2 respectively for the apical, mid and basal slices. The compacted myocardial ROI was placed in the septum due to low frequency of septal compaction abnormalities. ROI placement was performed at a point where the signal intensity was most homogeneous and had a standard deviation ≤15% of the signal intensity. The blood pool ROI was placed in the centre of the LV cavity, where the region had the highest signal intensity but with minimal signal variability. Using a modified Otsu thresholding technique [17], the non-compacted myocardial mass, which typically extended beyond the compacted myocardium and into the LV cavity in each slice, was determined as follows. The lower limits of accepted signal intensities for non-compacted myocardium ranged from that of the compacted myocardium (SIMYO) at the septum in any given slice to an upper signal intensity derived by the following formula; SIMYO + (SIBLOOD − SIMYO ∗ 0.5). Once the lower and upper signal intensity thresholds were defined per slice, semi-automated software was used to define the area of noncompacted myocardium within this range of signal intensities in each slice, which was then multiplied by slice thickness and myocardial tissue density constant — 1.05 g to derive total trabeculated mass. The myocardial mass was then described in terms of absolute NC and C masses as well as a NC:C ratio, describing the relative contribution of NC mass and C mass. 2.3. MR flow calculation using phase contrast imaging A flow sensitive gradient-echo sequence (TR, b5 ms; TE, b3 ms; flip angle, 15°; slice thickness, 7 mm; field of view = 300–380 mm matrix, 256 × 240, temporal resolution = 30 ms) during breath hold was used to acquire aortic flow data. The ascending aorta (sinotubular junction) was used as the imaging plane. Through-plane flow data was acquired by the use of retrospective cardiac gating. Phase contrast images were used to calculate arterial blood flow by use of a semiautomatic vessel edge-detection algorithm (Reportcard, GE) with manual operator correction. 2.4. Regional assessment of NC:C ratios Regional LVNC was evaluated using the 17-segment American Heart Association (AHA) model [18] and counting the number of segments with a NC:C length ratio of N2.3 in diastole in short axis. Global LVNC was defined as involvement of greater than 7 NC segments of the 17 segment AHA model. Involvement of fewer segments was defined as “regional” LVNC. The ventricle was also divided into basal, mid and apical levels using papillary muscles as the dividing anatomical landmark. NC:C mass ratios in ES and ED

Fig. 1. Demonstration of thresholding technique using relative signal intensities of blood pool and compacted myocardium. (a) shows end-diastolic (ED) phase and (b) shows end-systolic (ES) phase. Papillary muscles have a signal intensity of compacted myocardium, as seen in the mid-LV septal wall, and are excluded from the non-compacted mass calculations (manual regions of interest in yellow).

432

P. Choudhary et al. / International Journal of Cardiology 181 (2015) 430–436

3.2. Standard volumetric CMR data (Table 1)

Table 1 Patient characteristics and volumetric measurements. Variable

LVNC (n = 30)

Control (n = 20)

P Value

Age (years) Gender (%) – Male – Female Height (cm) Weight (kg) BSA (m2) BMI (kg/cm2) Heart rate (beats/min) LV end-diastolic volume indexed (mL/m2) LV end-systolic volume indexed (mL/m2) LV mass (g) LV ejection fraction (%) LV stroke volume (mL) Aortic flow (mL) Cardiac index (mL/min/m2)

44 +/− 16

34 +/− 8

0.02

50% 50% 171 +/− 10 73. +/− 17 1.85 +/− 0.3 24.9 +/− 4 68 +/− 8 89. +/− 17

55% 45% 177 +/− 13 72 +/− 14 1.89 +/− 0.2 23 +/− 4 66 +/− 10 86 +/− 15

0.81 0.13 0.74 0.63 0.27 0.43 0.56

39 +/− 16

34 +/− 8

0.19

161 +/− 49 57 +/− 10 92 +/− 19 88 +/− 19 3.4 +/− 0.8

144 +/− 34 61 +/− 4 95 +/− 20 93 +/− 22 3.4 +/− 0.6

0.17 0.07 0.56 0.42 0.96

were measured in the basal, mid and apical slices using our technique. Ventricular function for each slice of the short axis stack was determined by calculating the percentage change in circumferential area in systole compared to diastole and a local “slice ejection fraction” was obtained. 2.5. Late gadolinium enhancement (LGE): scar/fibrosis imaging Segmented phase-sensitive inversion recovery sequences (Image parameters: TR = 2 × RR interval; TE = 3.4 ms; flip angle = 25°; slice thickness = 10 mm; matrix = 144 × 256; field of view = 300–380 mm, acquired during a single breath-hold) were used in the entire short and multiple long axis views of the LV, to identify myocardial scar 10 min post-administration of intravenous contrast (0.2 mmol/kg of gadolinium pentatate, Magnevist). 2.6. Reproducibility Inter-observer variability was assessed by comparing measurements made by the primary operator (PC) and a second blinded CMR cardiologist (CH, Level III SCMR accredited). The coefficient of variation was determined by dividing the difference between the measurements by the mean measurement for the two operators (PC and CH) for each variable measured. Agreement between the measurements and the methods was further analysed using the intra-class coefficient as described by Shrout and Fleiss [19]. The coefficient of variation was 3.2% and the intra-class co-efficient using two-way mixed models for consistency was 0.990. 2.7. Statistical Analysis Statistical analysis was performed using SPSS software (Version 21, Armonk, NY: IBM Corp). Gaussian distribution was assessed using the Kolmogorov-Smirnov test with a p value of b0.05 considered significant. Ventricular mass, volumes and quantified ratios were treated as continuous variables. Differences between LVNC patients and controls of continuous variables were analysed using independent t-tests and differences between three groups (i.e. global NC, regional NC and controls) were assessed using analysis of variance methods (ANOVA). Non-parametric tests were used to compare means for variables that were not normally distributed. Differences in categorical variables were assessed using Fisher's exact tests. Correlations between continuous variables were assessed using Pearson's correlation method. Inter-observer variability was assessed using coefficient of variation as well as by intra-class correlation coefficient. Results were considered significant at a 2-tailed p value of b0.05.

3. Results 3.1. Patient characteristics (Table 1) 30 consecutive patients with isolated LVNC and 20 gender matched healthy controls were studied. Baseline characteristics are shown in Table 1. Of patients with LVNC, 3 had a family history of sudden cardiac death and 10 had a family history of LVNC, 14 had reduced ejection fraction (defined as EF b 55%) and 15 patients had abnormalities on ECG or Holter monitoring. The most common symptoms included syncope in 12 (39%), dyspnea in 7 (23%) and palpitations in 5 (16%) patients.

Indexed left ventricular end-diastolic volumes, ventricular mass, LV EF and cardiac indices were similar between the non-compacted and control subjects. No significant differences in stroke volume or aortic flow were demonstrated between the LVNC and control groups. 3.3. Quantification of non-compacted myocardium (Table 2) NC mass was significantly higher in both ES and ED phases in the LVNC cohort. No significant differences in C mass were observed. NC:C ratio and NC:total ratio were significantly elevated in LVNC patients in both ES and in ED. Total ventricular mass was not significantly different between the LVNC and control groups. When indexed to body surface area, the indexed NC:C ratio and NC:total mass ratios were significantly higher in LVNC patients in both cardiac phases. A strong correlation was observed between the cardiac phases of ED and ES NC mass (r = 0.773, p b 0.01), C mass (r = 0.959, p b 0.01) and total mass (r = 0.968, p b 0.01). The NC:C (r = 0.613, p b 0.01) and NC: total mass ratios (r = 0.617, p b 0.01) also showed a significant correlation between ES and ED phases. 3.4. Relationship between NC:C ratios and Ejection Fraction (Figs. 2 and 3) The NC:C ratio showed significant inverse correlation with ejection fraction in both cardiac phases. This correlation was stronger at endsystole compared to ED (ES r = − 0.58, p b 0.001 vs. ED r = − 0.30, p = 0.03) (Fig. 2b). The ejection fraction also correlated inversely with ES NC mass (r = −0.50, p b 0.001) and with ED NC mass (r = −0.38, p = 0.007). No significant correlations were seen between ejection fraction and C or total mass. Indexed LV end-diastolic volumes also showed a strong positive correlation with both ES NC mass (r = 0.54, p b 0.001) as well as ES NC:C ratio (r = 0.371, p = 0.01). ED NC:C ratios did not significantly correlate with indexed LV end-diastolic volumes (r = 0.14, p = 0.33) or with stroke volume or aortic flow measured by phase contrast imaging. Correlation with ejection fraction was compared to both our criteria (Fig. 3a) and the current Petersen's criteria (Fig. 3b). The left ventricular ejection fraction showed no significant correlation with measured NC:C ratio by Petersen's criteria (r = 0.06, p = 0.75), compared to a significant inverse correlation with ES NC:C ratio (r = − 0.54, p b 0.001) derived by our proposed method. Within the LVNC cohort, the ES NC mass (34.6 [SD 15] g vs. 23.4 [SD 8] g, p = 0.017), NC:C ratio (0.27 [SD 0.1] vs. 0.17 [SD 0.06], p = 0.003) and NC:total mass ratios (0.21 [SD 0.06] vs. 0.15 [SD 0.04], p =

Table 2 Comparison of NC and C masses in LVNC patients vs. controls in both cardiac phases of ES and ED. LVNC (n = 30)

Controls (n = 20)

P Value

End-systole (ES) NC mass (g) C mass (g) NC:C ratio Indexed NC:C NC:total mass ratio Indexed NC:total

28.2 +/− 13 136.4 +/− 38 0.21 +/− 0.09 0.12 +/− 0.06 0.17 +/− 0.06 0.09 +/− 0.04

16.6 +/− 5 134.1 +/− 33 0.12 +/− 0.02 0.07 +/− 0.02 0.11 +/− 0.02 0.06 +/− 0.01

b0.001 0.83 b0.001 0.001 b0.001 b0.001

End-diastole (ED) NC mass (g) C mass (g) NC:C ratio Indexed NC:C ratio NC:total mass ratio Indexed NC:total Total mass (g)

47.6 +/− 16 121.8 +/− 37 0.39 +/− 0.08 0.22 +/− 0.07 0.28 +/− 0.04 0.16 +/− 0.04 169.4 +/− 51

30.9 +/− 8 120.2 +/− 29 0.26 +/− 0.05 0.14 +/− 0.04 0.20 +/− 0.03 0.11 +/− 0.03 151.1 +/− 35

b0.001 0.87 b0.001 b0.001 b0.001 b0.001 0.17

P. Choudhary et al. / International Journal of Cardiology 181 (2015) 430–436

433

Fig. 2. Relationship between NC:C ratio and ventricular function. (a) compares NC:C ratios among controls (n = 20) and LVNC patients (n = 30) in ES and ED; (b) shows the correlation between NC:C ratio and global LV ejection fraction; (c) describes the differences in ES and ED NC:C ratios when subjects are stratified by LVEF b55% and N55% and (d) describes the differences in in ES and ED NC:C ratios when stratified by LVEF b50% and N50%.

0.003) derived using this technique, were significantly higher in patients with reduced ejection fraction (EF b 55%, n = 14) compared to LVNC patients with preserved ejection fraction (n = 16). The C mass did not significantly differ between LVNC patients with reduced EF and those with EF N 55% (132 [SD 39] vs. 129 [SD 34] g, p = 0.82). For patients with EF b 50% (n = 6) compared to those with EF N 50% (n = 45) of the LVNC cohort, similar findings were noted, with significantly higher ES NC mass (41.84 [SD 18.30] g vs. 18.02 [SD 7.49] g, p b 0.001), NC:C ratios (0.31 [SD 0.13] vs. 0.18 [SD 0.06], p = 0.001) and NC:total mass ratios (0.23 [SD 0.07] vs. 0.15 [SD 0.04], p = 0.001, Fig. 2d). Similarly, C mass did not significantly differ between the two groups (141 [SD 53] vs.127 [SD 31] g, p = 0.42).

3.5. Relationship between NC mass and electrical abnormalities (Fig. 3) In the LVNC cohort, 4 patients had sustained ventricular tachycardia, 1 had non-sustained ventricular tachycardia, 1 had Mobitz type 2 block and 3 patients had resting ECG abnormalities including long QT interval, T wave inversion and left bundle branch block. Ventricular ectopic beats were noted in 4 patients, 2 patients had atrial arrhythmias and 1 had palpitations with no documented arrhythmia. Four patients had syncope. Patients with sustained ventricular tachycardia had a significantly higher ES NC:C ratio (0.31 [SD 0.18] vs. 0.20 [SD 0.06], p = 0.02). Using a composite end-point of any electrical abnormality including

Fig. 3. Clinical utility of the proposed signal intensity based NC:C quantification technique. (a) demonstrates the inverse correlation of our proposed signal intensity (SI) based ES NC:C ratio and global LV ejection fraction in LVNC patients. This relationship is not observed when using the NC:C ratios derived from the traditionally used Petersen's criteria (b). (c) demonstrates significantly elevated ED NC mass using SI technique in LVNC patients with electrical abnormalities (defined as documented arrhythmia or abnormalities on surface ECG or Holter monitoring) compared to those without electrical abnormalities. No significant differences are seen in the ED NC:C ratios using Petersen's technique as shown in (d).

434

P. Choudhary et al. / International Journal of Cardiology 181 (2015) 430–436

documented arrhythmias on the electrocardiogram or on Holter monitoring; (excluding syncope with undocumented rhythm), both ED NC mass (52.82 [SD 16] vs. 37.76 [SD 10], p = 0.005) and ED NC:C ratios (0.42 [SD 0.09] vs. 0.36 [SD 0.06], p = 0.03) were significantly higher in patients with electrical disturbances compared to those with no rhythm abnormalities. 3.6. Sensitivity and specificity of the NC:C ratio in systole and diastole Although no gold standard exists for diagnosing LVNC, the presence of LVNC by current CMR criteria with an additional clinical risk indicator, suggesting that heightened pre-test probability was used to diagnose LVNC. NC:C ratios derived using our proposed technique were assessed against the current best available technique to compare diagnostic performance and to derive a clinically useful cut-off for NC:C ratio. As the relationship to ejection fraction was the strongest in ES phase, the diagnostic performance of ES NC:C ratio was assessed. The sensitivity was plotted against specificity for NC:C ratio in ES and receiver-operated curves were generated. At ES, the area under the curve was 0.86 (CI 0.76–0.96). A NC:C ratio of 0.12 had sensitivity of 87% with 45% specificity; a ratio of 0.15 had a sensitivity of 73% and specificity of 85% while a ratio of 0.16 had a sensitivity of 70% and a specificity of 95%. Higher NC:C ratios of 0.20 had a sensitivity of 50% with specificity approaching 100%. Among patients with ES NC:C ≥ 0.16, ejection fraction was significantly lower (55 [SD 10]% vs. 61 [SD 5]%, p = 0.01) compared to those with ES NC:C b 0.16. 3.7. Assessment of regional non-compaction (Fig. 4) On average, 7.4 [SD 2] segments per ventricle were non-compacted in the LVNC patients with a predominantly apical distribution. Noncompaction involved at least one apical segment in 100%, midsegment in 88% and basal segment in 42%. A mean number of 4

[SD 0.7] of 5 apical segments, 2 [SD 1.3] of 6 mid-ventricular segments and 0.7 [SD 0.95] of 6 basal segments were involved. LVNC patients had significantly higher ES NC:C ratios in the mid and apical levels (mid: 0.26 [SD 0.17] vs. 0.13 [SD 0.04], p = 0.002; apical: 0.29 [SD 0.21] vs. 0.14 [SD 0.05], p = 0.002) compared to controls (Fig. 4a). ED NC:C ratios at basal, mid and apical levels (0.28 [SD 0.15] vs. 0.20 [SD 0.06], p = 0.03 basal; 0.47 [SD 0.14] vs. 0.28 [SD 0.07], p b 0.001 mid and 0.65 [SD 0.26] vs. 0.42 [SD 0.11], p b 0.001 apical) were also higher. A significant incremental change in NC:C ratios was observed between controls and patients with regionally distributed LVNC (≤7 segments involved) as well as between patients with regional LVNC and LVNC patients with global (N7 segments) involvement (controls 0.12 [SD 0.02] vs. regional 0.18 [SD 0.06], p = 0.001, regional 0.18 [SD 0.06] vs. global 0.25 [SD 0.11], p = 0.04; ANOVA p b 0.001) (Fig. 4b). 3.8. Relationship of regional LVNC and ejection fraction ES basal, mid and apical NC:C ratios also showed a significant inverse correlation with global LV ejection fraction (ES basal r = −0.29, p = 0.04; mid-ventricular r = −0.50, p b 0.001 and apical r = − 0.71, p b 0.001). ED apical NC:C ratio also correlated inversely with ejection fraction (r = −0.45, p b 0.001) but this correlation was not significant at the basal and mid-ventricular levels in ED (Fig. 4c). The local effect of increased NC:C mass ratios was assessed by measuring the regional slice ejection fraction for each slice of the short axis stack, typically 8–12 contiguous slices from apex to base (slice EF). LVNC patients had a significantly lower mean slice EF compared to controls (49 [SD 32] vs. 58 [SD 18]%, p b 0.001). ES NC:C ratios showed significant inverse correlation with regional slice ejection fraction (r = −0.38, p b 0.001) as shown in Fig. 4d. Using the previous clinical cut-off for ES NC:C ratio of N 0.16 for significant NC, slice ejection fractions were significantly lower in those with ES NC:C N 0.16 compared to those with ES NC:C b 0.16 (Slice EF = 44 [SD 1]% vs. 63 [SD 1]%, p b 0.001).

Fig. 4. Assessment of functional consequences of regional LVNC. (a) demonstrates higher ES NC:C ratios, based on relative signal intensities, at all ventricular levels (basal, mid and apical) in LVNC patients compared to controls. (b) shows an incremental increase in the measured NC:C ratio with increasing extent of involvement comparing controls, patients with regional LVNC and with a more “global” distribution of LVNC. (c) illustrates the relationship between regional ES NC:C ratio at each ventricular level and global ejection fraction. (d) shows the significant inverse relationship between the ES NC:C ratio at each slice and its individual LV short axis slice ejection fraction.

P. Choudhary et al. / International Journal of Cardiology 181 (2015) 430–436

3.9. Effect of late gadolinium enhancement There was a low prevalence of ventricular scar identified. LGE was observed in only 2 of 31 patients, one with a known prior inferior myocardial infarction in an area corresponding to the infarcted territory. The second patient had a small region of delayed gadolinium enhancement in the basal septum (mid-myocardial pattern) unrelated to the apically located compaction abnormalities. 4. Discussion Our study demonstrates three important concepts regarding LVNC quantification. Firstly, we show that our CMR technique, based on relative signal intensities of the compacted myocardium and blood pool, enables objective and reproducible quantification of NC mass throughout the whole ventricle. Secondly, we describe a robust correlation between NC:C ratio with both regional and global ventricular functions and demonstrate that even regional non-compaction influences global ventricular function. Thirdly, elevated NC:C ratio is associated with both ventricular dysfunction and arrhythmic consequences. We have identified that trabeculations occur in a quantifiable spectrum with associated clinical consequences at an ES NC:C ratio cut-off of N 0.16. To date, no diagnostic criteria for LVNC have been able to demonstrate a continuous link between degrees of structural change with progressive reduction of ventricular function. LVNC is increasingly recognized as a spectrum of disease ranging from abnormal localised trabeculations to severe and extensive LVNC cardiomyopathy with haemodynamic and arrhythmic consequences [3]. Genetic susceptibility and haemodynamic loading conditions may also affect expression of this cardiac phenotype. Variations in the regional patterns of NC have also been recognized, relating to the sequence of the normal compaction process, namely from epicardial to endocardial surface, from base to apex and from septal to the lateral wall [20]. Three main echocardiographic criteria have been proposed which describe varying degrees of single plane NC:C ratio cut-offs in either ES or ED as pathological [21–23]. Kohli et al. [14] demonstrated poor correlation between the three echocardiographic criteria that may be related to limitations of echocardiography including operator dependence and inconsistent visualization of the endocardium, especially at the LV apex. CMR has superior spatial resolution and is able to reproducibly visualize the entire LV, including the apex, in multiple orthogonal views to better assess regional variations in the extent of non-compaction. However, current CMR criteria also rely on a 2-dimensional, single plane binary NC:C ratio measured at the subjectively assessed “worst affected” myocardial segment [15]. Kawel et al. showed a high prevalence of healthy patients meeting the Petersen criteria for LVNC [24]. Subsequently, Zemrak et al. [7] concluded that even severe LVNC assessed by this technique was clinically benign at late follow-up. It remains unclear whether this conclusion has been reached due to limitations of the diagnostic technique utilised or indeed accurately reflects the proposed benign consequences of even marked LV trabeculations. Our attempts to relate the more widely used Petersen technique to ventricular function did not demonstrate a significant relationship, nor did it correlate with electrical sequelae. Our proposed alternative technique, utilising the relative signal intensities of compacted myocardium and blood pool allows for improved inclusion of confluent papillary muscles within the compacted mass and reduces over-inclusion of the blood pool within the non-compacted mass, leading to superior discriminatory capacity than other previously described MR based techniques [25]. Our technique also reduces operator dependency and subjective quantification. Given its semiautomated nature and correlation co-efficient of 3.2%, it has the potential to be a useful tool for serial follow-up of patients over time. Compared to a single plane binary measurement, we derive a NC:C mass ratio for the entire LV and demonstrate a continuous linear relationship to ejection

435

fraction with an incremental increase in this ratio comparing controls to patients with localised, regional involvement to patients with more widespread “global” myocardial involvement. The lack of a gold standard diagnostic test for LVNC complicates prediction of an absolute cut-off for pathological LVNC. Our findings suggest that an ES NC:C ratio N0.16 has a sensitivity of 70% and specificity of 95% for detecting functionally significant LVNC. A higher cut-off of ES NC:C 0.20 markedly reduces the sensitivity of the technique but correlates with worse ventricular function. Stacey et al. [6] suggest a threshold of trabeculated:total mass ratio of N0.40 measured in the ED phase using the technique proposed by Jacquier et al. [25]. This however, only identifies patients with severe trabeculations and who have morphologically obvious LVNC with established cardiomyopathy. Nevertheless in their cohort, the diagnostic performance of the trabeculated:total mass ratio was inferior to ES NC:C dimension ratio measured in the “worst affected” short axis slice in prediction of heart failure or LVNC associated clinical events. Our data supports assessment of LVNC in ES phase as well as in the short axis views, however our technique involves analysis of the entire ventricle from base to apex to include regional variations rather than a subjectively defined “worst affected” segment. Our technique also provides clinically useful information even in patients with LVNC that is in the moderate range. Specifically, we note that even mild-moderate trabeculations may contribute to both contractile dysfunction and arrhythmia. A significant proportion of our cohort had clinically significant arrhythmia, including ventricular tachycardia. Elevated NC mass and NC: C ratio were associated with both ventricular tachycardia as well as a composite of electrical abnormalities either on surface ECG or Holter monitoring. Although larger prospectively designed studies are required, possible mechanisms may include relating the degree of LV trabeculations with alterations in QT interval and/or LV scar. We demonstrate a low prevalence of LGE in our cohort, which is consistent with most large cohort studies [26]. This finding has been controversial with some groups suggesting a higher prevalence [27]. The timing of image acquisition following contrast injection and direct simultaneous comparison of LGE imaging with cine imaging in the same plane is critical as false positive enhancement can occur due to delayed emptying of blood from within endocardial recesses and crypts. Nevertheless, our findings suggest that significant arrhythmias associated with LVNC are not entirely explained by the presence of LV scar and thus further insights into the mechanisms of arrhythmogenesis are important. 5. Limitations One of the limitations of our study includes small patient numbers and retrospective design. However, LVNC is a rare disease with an estimated prevalence of 0.014% [28] and our sample size is comparable to other studies assessing LVNC diagnostic criteria. Longitudinal studies are required to improve our understanding of the sequence of events that lead to serious complications. Multiple groups have described ethnic variations in myocardial compaction [29,30] and our control cohort was not powered to describe the ethnic variations in NC mass. This quantification technique could be utilised to assess this issue further. The LVNC cohort was slightly but significantly older than the control cohort, however, they still demonstrated a higher non-compacted mass. Dawson et al. [31] assessed the effect of age on the extent of trabeculation in a normal cohort and showed an increase in compacted mass, reduction in total non-compacted mass and consequent reduction in NC:C ratios with increasing age. Demonstration of an elevated NC:C ratio in the slightly older LVNC cohort is thus more striking as the NC: C ratio of normal older controls would be expected to be lower. 6. Conclusion We demonstrate that our semi-automated, highly reproducible CMR derived measure of non-compacted and compacted myocardial masses

436

P. Choudhary et al. / International Journal of Cardiology 181 (2015) 430–436

based on relative signal intensities of the compacted myocardium and blood pool, allows for precise and accurate quantification of LVNC. Further, this method also allows precise quantification of clinically important regional trabeculations, with implications for both regional and global LV functions and relates to occurrence of clinically significant arrhythmia. Finally, we conclude that when the global NC:C ratio exceeds 0.16 in end systole, there are significant functional consequences relating to LV systolic function. Conflicts of interest The authors declare that they have no competing financial or nonfinancial interests. Acknowledgements Dr. Choudhary is funded by a NHMRC and National Heart Foundation co-funded post-graduate research scholarship (# 1055773). References [1] E.N. Oechslin, C.H.A. Jost, J.R. Rojas, P.A. Kaufmann, R. Jenni, Long-term follow-up of 34 adults with isolated left ventricular noncompaction: a distinct cardiomyopathy with poor prognosis, J. Am. Coll. Cardiol. 36 (2) (2000) 493–500. [2] S. Gati, N. Chandra, R.L. Bennett, M. Reed, G. Kervio, V.F. Panoulas, et al., Increased left ventricular trabeculation in highly trained athletes: do we need more stringent criteria for the diagnosis of left ventricular non-compaction in athletes? Heart 99 (6) (2013) 401–408. [3] R.T. Murphy, R. Thaman, J.G. Blanes, D. Ward, E. Sevdalis, E. Papra, et al., Natural history and familial characteristics of isolated left ventricular non-compaction, Eur. Heart J. 26 (2) (2005) 187–192. [4] G. Captur, A.S. Flett, D.L. Jacoby, J.C. Moon, Left ventricular “non-noncompaction”: the mitral valve prolapse of the 21st century? Int. J. Cardiol. 164 (1) (2013) 3–6. [5] B.J. Maron, J.A. Towbin, G. Thiene, C. Antzelevitch, D. Corrado, D. Arnett, et al., Contemporary definitions and classification of the cardiomyopathies: an American Heart Association Scientific Statement from the Council on Clinical Cardiology, Heart Failure and Transplantation Committee; Quality of Care and Outcomes Research and Functional Genomics and Translational Biology Interdisciplinary Working Groups; and Council on Epidemiology and Prevention, Circulation 113 (14) (2006) 1807–1816. [6] R.B. Stacey, M.M. Andersen, M.S. Clair, W.G. Hundley, V. Thohan, Comparison of systolic and diastolic criteria for isolated LV noncompaction in CMR, J. Am. Coll. Cardiol. Img. 6 (9) (2013) 931–940. [7] F. Zemrak, M.A. Ahlman, G. Captur, S.A. Mohiddin, N. Kawel-Boehm, M.R. Prince, et al., The relationship of left ventricular trabeculation to ventricular function and structure over a 9.5-year follow-up: the MESA study, J. Am. Coll. Cardiol. 64 (19) (2014) 1971–1980. [8] P. Elliott, B. Andersson, E. Arbustini, Z. Bilinska, F. Cecchi, P. Charron, et al., Classification of the cardiomyopathies: a position statement from the European Society of Cardiology Working Group on Myocardial and Pericardial Diseases, Eur. Heart J. 29 (2) (2008) 270–276 (1). [9] P. Richardson, W. McKenna, M. Bristow, B. Maisch, B. Mautner, J. O'Connell, E. Olsen, G. Thiene, J. Goodwin, I. Gyarfas, I. Martin, P. Nordet, Report of the 1995 World Health Organization/International Society and Federation of Cardiology Task Force on the Definition and Classification of Cardiomyopathies, Circulation 93 (1996) 841–842. [10] G. Captur, P. Nihoyannopoulos, Left ventricular non-compaction: genetic heterogeneity, diagnosis and clinical course, Int. J. Cardiol. 140 (2) (2010) 145–153.

[11] C. Stöllberger, J. Finsterer, Cardiologic and neurologic findings in left ventricular hypertrabeculation/non-compaction related to wall thickness, size and systolic function, Eur. J. Heart Fail. 7 (1) (2005) 95–97. [12] R. Lubrano, P. Versacci, G. Guido, E. Bellelli, G. Andreoli, M. Elli, Might there be an association between polycystic kidney disease and noncompaction of the ventricular myocardium? Nephrol. Dial. Transplant. 24 (12) (2009) 3884–3886. [13] B.E. Stähli, C. Gebhard, P. Biaggi, S. Klaassen, E. Valsangiacomo Buechel, C.H. Attenhofer Jost, et al., Left ventricular non-compaction: prevalence in congenital heart disease, Int. J. Cardiol. 167 (6) (2013) 2477–2481. [14] S.K. Kohli, A.A. Pantazis, J.S. Shah, B. Adeyemi, G. Jackson, W.J. McKenna, et al., Diagnosis of left-ventricular non-compaction in patients with left-ventricular systolic dysfunction: time for a reappraisal of diagnostic criteria? Eur. Heart J. 29 (1) (2008) 89–95. [15] S.E. Petersen, J.B. Selvanayagam, F. Wiesmann, M.D. Robson, J.M. Francis, R.H. Anderson, et al., Left ventricular non-compaction insights from cardiovascular magnetic resonance imaging, J. Am. Coll. Cardiol. 46 (1) (2005) 101–105. [16] S.E. Petersen, CMR and LV noncompaction: does it matter how we measure trabeculations? J. Am. Coll. Cardiol. Img. 6 (9) (2013) 941–943. [17] N. Otsu, A threshold selection method from gray-level histograms, Automatica 11 (285–296) (1975) 23–27. [18] Segmentation AHAWGoM, Imaging: RfC, M.D. Cerqueira, et al., Standardized myocardial segmentation and nomenclature for tomographic imaging of the heart: a statement for healthcare professionals from the Cardiac Imaging Committee of the Council on Clinical Cardiology of the American Heart Association, Circulation 105 (4) (2002) 539–542. [19] P.E. Shrout, J.L. Fleiss, Intraclass correlations: uses in assessing rater reliability, Psychol. Bull. 86 (2) (1979) 420. [20] D. Sedmera, T. Pexieder, M. Vuillemin, R.P. Thompson, R.H. Anderson, Developmental patterning of the myocardium, Anat. Rec. 258 (4) (2000) 319–337. [21] R. Jenni, E. Oechslin, J. Schneider, C.A. Jost, P. Kaufmann, Echocardiographic and pathoanatomical characteristics of isolated left ventricular non-compaction: a step towards classification as a distinct cardiomyopathy, Heart 86 (6) (2001) 666–671. [22] T.K. Chin, J.K. Perloff, R.G. Williams, K. Jue, R. Mohrmann, Isolated noncompaction of left ventricular myocardium. A study of eight cases, Circulation 82 (2) (1990) 507–513. [23] C. Stollberger, J. Finsterer, G. Blazek, Left ventricular hypertrabeculation/ noncompaction and association with additional cardiac abnormalities and neuromuscular disorders, Am. J. Cardiol. 90 (8) (2002) 899–902. [24] N. Kawel, M. Nacif, A.E. Arai, A.S. Gomes, W.G. Hundley, W.C. Johnson, et al., Trabeculated (noncompacted) and compact myocardium in adults the multiethnic study of atherosclerosis, Circ. Cardiovasc. Imaging 5 (3) (2012) 357–366. [25] A. Jacquier, F. Thuny, B. Jop, R. Giorgi, F. Cohen, J.-Y. Gaubert, et al., Measurement of trabeculated left ventricular mass using cardiac magnetic resonance imaging in the diagnosis of left ventricular non-compaction, Eur. Heart J. 31 (9) (2010) 1098–1104. [26] J. Wan, S. Zhao, H. Cheng, M. Lu, S. Jiang, G. Yin, et al., Varied distributions of late gadolinium enhancement found among patients meeting cardiovascular magnetic resonance criteria for isolated left ventricular non-compaction, J. Cardiovasc. Magn. Reson. 15 (2013) 20. [27] G. Nucifora, G.D. Aquaro, A. Pingitore, P.G. Masci, M. Lombardi, Myocardial fibrosis in isolated left ventricular non-compaction and its relation to disease severity, Eur. J. Heart Fail. 13 (2) (2011) 170–176. [28] E. Oechslin, R. Jenni, Left ventricular non-compaction revisited: a distinct phenotype with genetic heterogeneity? Eur. Heart J. 32 (12) (2011) 1446–1456. [29] T. Luijkx, M. Cramer, A. Zaidi, R. Rienks, P. Senden, S. Sharma, et al., Ethnic differences in ventricular hypertrabeculation on cardiac MRI in elite football players, Neth. Hear. J. 20 (10) (2012) 389–395. [30] G. Captur, V. Muthurangu, C. Cook, A.S. Flett, R. Wilson, A. Barison, et al., Quantification of left ventricular trabeculae using fractal analysis, J. Cardiovasc. Magn. Reson. 15 (1) (2013) 36. [31] D.K. Dawson, A.M. Maceira, V.J. Raj, C. Graham, D.J. Pennell, P.J. Kilner, Regional thicknesses and thickening of compacted and trabeculated myocardial layers of the normal left ventricle studied by cardiovascular magnetic resonance, Circ. Cardiovasc. Imaging 4 (2) (2011) 139–146.