1
Whole human heart histology to validate electroanatomical voltage mapping in patients with non-ischemic cardiomyopathy and ventricular tachycardia
- Supplementary data-
Claire A. Glashan, MD1; Alexander F.A. Androulakis, MD1; Qian Tao, PhD2; Ross N. Glashan;
Lambertus Wisse3; Micaela Ebert, MD1; Marco C. de Ruiter, PhD3; Berend J. van Meer3; Charlotte Brouwer, MD1; Olaf M. Dekkers, PhD4; Daniel Pijnappels, PhD1; Jacques M.T. de Bakker5, PhD; Marta
de Riva, MD1; Sebastiaan R.D Piers1, MD, PhD; Katja Zeppenfeld, MD, PhD1
1. Department of Cardiology, Leiden University Medical Centre, Leiden, The Netherlands 2. LKEB – Division of Image Processing, Department of Radiology, Leiden University Medical
Centre, Leiden, The Netherlands
3. Department of Anatomy and Embryology, Leiden University Medical Centre, Leiden, The Netherlands
4. Department of Epidemiology, Leiden University Medical Centre, Leiden, The Netherlands 5. Department of Clinical and Experimental Cardiology, Academic Medical Centre,
Amsterdam, The Netherlands
2
Supplemental methods
Patients and controls
The presence of coronary artery stenosis >50% was excluded by angiography. Patients with HCM, ARVC, non-compaction CM and cardiac sarcoidosis were excluded.
Electroanatomical mapping and ablation
Electrograms were filtered at 30-400Hz (bipolar) and 1-240Hz (unipolar). Epicardial access was obtained through subxiphoid puncture if a prior endocardial ablation had failed or endocardial substrate and/or activation mapping was consistent with an epicardial substrate. In these patients, an aortic root map was obtained for in-vivo CT image integration. Based on previous proposed cut- off values to detect fibrosis, low voltage sites were defined as electrograms with BV<1.5mV or UV<8.27mV for the endocardium and with BV<1.5mV and UV<7.95mV for the epicardium 1-4. In all but one patient (who presented with an incessant VT) substrate mapping was the initial approach, however, in the majority of patients abnormal EGMs that could be targeted during SR were sparse.
Accordingly activation and entrainment mapping facilitated by hemodynamic support (in all but one patient) was performed. Only sites likely related to induced VTs were targeted. All but one patient remained inducible at the end of the procedure and had recurrence of VT.
Post-mortem/post-transplantation image integration
All hearts were excised from the thoracic cavity in their entirety and fixed in 10% formaline.
The hearts were embedded in ballistic gelatin (20% gelatin, 0.9% saline). A semi-automatic slicer (Cater Chef Slicer 300ES-12) was used to slice 5mm short-axis slices, which were photographed. In- house developed software (MASS, Research version 2016, Leiden University Medical Centre, Leiden, The Netherlands) was used to trace endocardial, epicardial and aortic root contours onto the images of the short-axis slices. 3D meshes created from these contours were imported into CARTO and manually merged with EAVM data, as previously described5. The registration matrix, mapping point co-ordinates, mapping point details (e.g. ablation sites), BV and UV were exported from CARTO.
3
Using inverse registration, all points were superimposed on the pathology slices and projected onto the closest contour (MATLAB, R2015b, MathWorks inc. Natick, MA). The accuracy of integration was determined by visual assessment of the spatial relationship between projected ablation sites and distinct anatomical ablation lesions on gross pathology. Points which had to be translocated ≥10mm from the CARTO map to be projected onto a gross pathology contour were excluded. If multiple points were projected to within 2mm of each other, the mean voltages of clustered data were taken.
Histological analysis
Biopsies were embedded in paraffin and 7μm sections were obtained and stained with Picrosirius Red (cross-stained with Weigert's Hematoxylin) to visualize histology with collagen staining red, myocardium yellow and nuclei black. High-resolution microscopy images were taken at 20x magnification (3DHistech, Pannoramic 250 Flash III digital scanner).
Quantification of the amount of fibrosis
Custom software (Python, 2.7) was used to process the image of the scanned biopsies. Wall thickness for each biopsy was measured. Each pixel in the image was classified as red, yellow, black or white; corresponding to fibrosis, myocardium, nuclei and non-staining tissue respectively (Figure S1).
The percentage of fibrosis within a biopsy was calculated by dividing the number of red pixels by the sum of red, black and yellow pixels. As voltages are generated by viable myocardium, the quantity of viable myocardium was calculated as the area of yellow and black pixels (mm2) for each TB. As all biopsies had the same width (5mm), the area of viable myocardium gave a
comparable measure considering both the amount of fibrosis and the wall thickness. As an example:
a myocardial biopsy with a wall thickness of 12mm and no fibrosis would produce an area of viable myocardium of 60mm2. If this same biopsy had 50% fibrosis, the area of viable myocardium would be 30mm2.
4
The area of viable myocardium was used as a surrogate for the volume of viable myocardium present at that location. To validate this, three-dimensional reconstructions from representative biopsies of each patient were created. As a 3.5mm tip catheter was used a 3.5mm x 3.5mm x wall thickness biopsy was taken for each reconstruction. The entire biopsy was slices at 7μm, every 14th slice was mounted and stained to result in a z-resolution of 98μm6. Fibrosis and viable myocardium was quantified in each thin slice as described above and the volume calculated. If the area of viable myocardium in one thin slice is compared to the volume of viable myocardium at that same location a linear relationship is seen (R2 = 0.94) (Supplemental figure S2).
From the control biopsies, 7μm sections were taken, stained and scanned. The images were processed as above and the percentage of fibrosis was calculated. As the control biopsies did not have a fixed width of 5mm, the amount of viable myocardium could not be quantified in a way which would be comparable to the NICM biopsies.
LGE-CMR and analysis of scar extension and heterogenity according to previously published methods
One patient underwent LGE-CMR prior to ablation. A 1.5-T Gyroscan ACS-NT/Intera MRI scanner (Philips Medical Systems, Best, The Netherlands) was used. Images were acquired approximately 15 minutes after bolus injection of gadolinium diethylenetriamine penta-acetic acid (Magnevist, Schering/Berlin, Germany; 0.15 mmol/kg) with an inversion-recovery 3D turbo-field echo sequence with parallel imaging (SENSE, acceleration factor 2). Inversion time was determined with real-time plan scan to null normal myocardial signal. The heart was imaged in 1 breath-hold with 22 imaging levels in the short-axis view (slice thickness 5mm). Image analysis was performed with the MASS software package. In the CE short-axis image series, LV endocardial and epicardial contours were manually traced. Four scar identification methods were used as previously defined:
1.) Max-SI method: Myocardial tissue with an SI value of ≥35% of the maximum SI within a user defined high intensity region was considered scar. The scar core was defined as myocardium
5
with an SI of ≥50 and scar borderzone as a myocardium with SI ≥35% but SI <50% of the maximum SI7.
2.) Modified Full-Width Half Maximum method: scar core was defined as an SI value of >50% of the maximum SI within a user defined high intensity region. A user defined region with low SI in the remote myocardium as identified and the maximum SI with this remote area determined. Borderzone was defined as SI>peak remote SI but <50% of maximum SI8.
3.) 6 SD method: scar was defined as >6SD above the mean SI of a user defined remote region9.
4.) 2-3SD method: Scar core was defined as >3SD above the mean SI of a user defined remote region and scar borderzone was defined as >2SD but <3SD above the mean10.
6
Supplemental Results
Inter-observer agreement
Inter-observer agreement for the pattern and architecture of fibrosis was evaluated by assessing the percentage agreement of classification between two observers.
One hundred and fifty TB were randomly selected for inter-observer agreement analysis (100 endocardial TB and 50 epicardial TB). Inter-observer agreement was 89% for the pattern of fibrosis. Inter-observer agreement was 84% for architecture of fibrosis. Of the 16% disagreement, 15% was partial disagreement (agreement on ≥1 architecture, but disagreement on ≥1 architecture) and 1% was complete disagreement.
Cut-off values to detect amount of viable myocardium
Receiver operating characteristics (ROC) curve analysis was performed to determine the optimal BV and UV cut-off values for assessing a pre-specified amount of viable myocardium, defined as the value maximizing the sum of sensitivity and specificity. Two amounts of viable myocardium were preselected: <40mm2 and <60mm2, based on the median amount of viable myocardium in endocardial TB with >21% or < 21% fibrosis, respectively. ROC analysis yielded optimal cut-offs of 2.9mV BV (AUC 0.68, sensitivity 73%, specificity 56%) and 6.85 mV UV (AUC 0.72, sensitivity 73%, specificity 56%) to identify <40mm2 and 3.93mV BV (AUC 0.67, sensitivity 79%, specificity 53%) and 8.29mV UV (AUC 0.66, sensitivity 78%, specificity 53%) to identify <60mm2.
Performance of currently applied BV and UV cut-off values
Of all endocardial TB, 62 (22%) generated a BV of ≤1.5mV. The majority of them had abnormal amounts of fibrosis (84%, median 29.5% [IQR 22.6-40.1]). However, 68% of biopsies generating BV>1.5mV also had abnormal amounts of fibrosis (median 24.8% [IQR 18.9-32.0]), indicating that a BV cut-off of >1.5mV for excluding the presence of abnormal fibrosis is insufficient.
In addition, although two thirds of TB generating BV ≤1.5mV showed a sub-endocardial or
7
transmural pattern of fibrosis, 30% showed a mid-wall or sub-epicardial pattern, which further supports that BV mapping has a wider field of view than previously suggested.
Of the endocardial TB, 194 (70%) generated a UV of ≤8.27mV. These biopsies showed a sub- endocardial or transmural pattern of fibrosis in 58%, a mid-wall or sub-epicardial pattern in 37% and no fibrosis in 5%. Eighty percent of biopsies generating a UV≤8.27mV had abnormal amounts of fibrosis (median fibrosis 26.8% [IQR 22.0-37.7]). On the other hand, of the biopsies which generated
>8.27mV, 53% also had >21% fibrosis (median 22.5% [IQR 15.3-29.2]), indicating that a cut-off value of >8.27mV performs poorly at excluding the presence of abnormal fibrosis.
Linear relation between wall thickness and electrogram amplitude
The linear relation between wall thickness and electrogram amplitude is at first unexpected.
Activation in a myocardial bundle can be represented by a current dipole. It can be shown that the amplitude of a unipolar electrogram generated by activity in a myocardial bundle is inversely proportional to the power of two of the distance between the site of activation (the current dipole) and the recording site (r, Figure S4). The equation is valid for distances that are larger than the separation between the current source and sink of the dipole (r>>d). Distance d is in the range of 1 mm (conduction velocity x duration upstroke action potential). In our study, the relation between electrogram amplitude and wall thickness was determined for ex-vivo wall thicknesses between approximately 10 and 20 mm. Thus, for these recordings the approximation that the amplitude of the extracellular electrogram is related to 1/r2 is valid.
Figure S4A shows the curve of 1/r2 from r=0,1 to r=2. The decline of the signal is fast between 0,1 and 0,5 (value within this interval varies from 100 to 4) and is much slower between 1 and 2 (from 1 to 0,25). In panel B, the decline of the curve for the interval between 1 and 2 is depicted and shows that the curve is virtually linear between these two points (R2=0,92). Thus for
“large” distances the amplitude of the electrogram varies linear with the distance between recording site and the activated bundle. For short distances, this will not be the case. Also, if one takes into
8
account the fact that the contribution of multiple bundles must be added to account for wall thickness, the relation between distance (wall thickness) and amplitude is virtually a linear one as illustrated in Figure 3C. In this Figure, the contribution of bundles is stacked to reflect the wall thickness. The relation is linear for “large” values of r with R2=0,97 for linearity.
A similar explanation is valid for a bipolar electrogram where the amplitude decreases with the power of three of the distance between the site of activation and recording (Figure S4 and Figure S5 D, E and F). An increase in collagen can be viewed as a decrease in the dipole current strength, which means that the relation between electrogram amplitude and wall thickness remains linear for the wall thicknesses used. The line will only shift downward, as is evident from data of our hearts with increases fibrosis.
The field of view of a bipolar recording (amplitude proportional to 1/r3) is smaller than that of a unipolar one (amplitude proportional to 1/r2), which means that for a same distance r between activation and recording site the bipolar recording is lower in amplitude. The linear increase of the signal with the wall thickness is simply caused by the fact that function 1/r2 (as well as 1/ r3) approaches a linear function for intervals with larger values of r.
9
Supplemental tables
Table S1. Mapping and ablation characteristics
LV endocardial mapping 8 (100%)
Endo- and epicardial mapping 5 (63%) LV endocardium:
Mapping points (n) 255 [170-317]
BV < 1.5mV (n) 99 [56-174]
BV scar size (cm2) 22.9 [14.3-46.6]
UV < 8.27mV (n) 205 [129-270]
UV scar size (cm2) 96.5 [74.6-209.4]
Epicardium:
Mapping points (n) 251 [211-359]
BV < 1.5mV (n) 115 [115-130]
BV scar size (cm2) 169.6 [39-193.7]
UV < 7.95mV (n) 182 [177-194]
UV scar size (cm2) 93.3 [83.2-383.5]
Mechanical support during procedure 7 (88%) Bail out procedures
Bipolar or alcohol ablation 2 (25%)
Surgical cryoablation 2 (25%)
10 Table S2. Electroanatomical and histological characteristics
Location of low voltage areas Histology
Endocardium Epicardium
Patient
Bipolar Voltage
< 1.5mV
Unipolar voltage
< 8.27mV
Bipolar voltage
< 1.5mV
Unipolar voltage
< 8.27mV
Endocardial biopsies (n)
Epicardial biopsies (n)
Fibrosis location macroscopic
histology
Dominant fibrosis pattern
1 Basoseptal
Entire LV except apex and lateral
wall
Entire epicardium except anterior and
inferolateral
Entire epicardium except
anterior and inferolateral 54 58 Baso-anteroseptal Sub-endocardium
2 Baso-anteroseptal Baso-anteroseptal Baso-anteroseptal Baso-anteroseptal 32 37 Basoanterior Mid-wall
3 basal, around aortic
root Basoseptal Antero-septal Anteroseptal 21 17 Basoseptal Transmural
4 Basal, anterior Mitral
annulus Basoseptal Anterobasal Anteroseptal 32 58 Basoanterior Sub-endocardium and
Mid-wall
5 Basoseptal Inferio-anterobasal Entire epicardium except inferolateral
Entire epicardium except
inferolateral 57 60 Basoseptal Sub-epicardium
6 Baso-anteroseptal anterior towards
apex, basoseptum 43 Mid anterior Sub-epicardium
7 Normal voltages Anteroseptal 20 Mid lateral Sub-endocardium
8
Anteroseptal some inferior and apical
involvement
Anteroseptal 18 Apical-
anteroseptal Sub-endocardium
11 Table S3. Architecture of fibrosis in NICM
Dominant fibrosis architecture Total
Interstitial Patchy Diffuse Compact
Secondary fibrosis architecture
Interstitial 0 (0%) 217 (42.8%) 40 (7.9%) 1 (0.2%) 258 (50.9%)
Patchy 6 (1.2%) 0 (0%) 122 (24.1%) 7(1.4%) 135 (26.6%)
Diffuse 0 (0%) 57 (11.2%) 0 (0%) 6 (1.2%) 63 (12.4%)
Compact 0 (0%) 1 (0.2%) 0 (0%) 0 (0%) 1 (0.2%)
No secondary architecture
39 (7.7%) 2 (0.4%) 9 (1.8%) 0 (0%) 50 (9.9%)
Total 45 (8.9%) 277 (54.6%) 171 (33.7%) 14 (2.8%) 507
12
Supplemental figures
Figure S1. Processing of scanned stained biopsies using custom software. Each pixel in input image (left) is classified as red (corresponding to fibrosis), yellow (tissue), black (nuclei) or white (non- staining). Output image (right) consists of pixels with these 4 colours.
13
Supplemental figure S2. Linear relationship between volume and area of viable myocardium.
14
Figure S3. The potential generated by a current dipole at a distance r in a homogeneous conductive medium. For distances r>>d the potential is inversely proportional to the power of two of the distance. “d” is the distance between the current source and sink of the dipole.
15
Figure S4. The bipolar voltage generated by a current dipole at a distance r in a homogenous conductive medium. For distances r>>x the potential is inversely proportional to third power of the distance. “x” is the inter-electrode spacing.
16
Figure S5. A and D: Graph of the function 1/r2 and 1/r3for values of r between 0,1 and 2,0. Note that the function descends steeply for r-values <0,5 and gently for higher values. B and E:the same function for values of r between 1,0 and 2,0. Note that the course of the function in this interval is virtually linear (black line; R2 = 0,92 and R2 = 0,86). C and F: Graphs of the function ∑1/rn2 and ∑1/rn3
for values of r between 1,0 and 2,0 (r1 = 0,1, r2 = 0,2 etc.). Note that the course of the function in this interval is virtually linear (black line; R2 = 0,97 and R2 =0,92).
17 References
1. Desjardins B, Morady F, Bogun F. Effect of epicardial fat on electroanatomical mapping and epicardial catheter ablation. J Am Coll Cardiol 2010;56(16):1320-7.
2. Hutchinson MD, Gerstenfeld EP, Desjardins B, Bala R, Riley MP, Garcia FC, Dixit S, Lin D, Tzou WS, Cooper JM, Verdino RJ, Callans DJ, Marchlinski FE. Endocardial unipolar voltage mapping to detect epicardial ventricular tachycardia substrate in patients with nonischemic left ventricular cardiomyopathy. Circ Arrhythm Electrophysiol 2011;4(1):49-55.
3. Marchlinski FE, Callans DJ, Gottlieb CD, Zado E. Linear ablation lesions for control of unmappable ventricular tachycardia in patients with ischemic and nonischemic cardiomyopathy.
Circulation 2000;101(11):1288-96.
4. Piers SR, van Huls van Taxis CF, Tao Q, van der Geest RJ, Askar SF, Siebelink HM, Schalij MJ, Zeppenfeld K. Epicardial substrate mapping for ventricular tachycardia ablation in patients with non- ischaemic cardiomyopathy: a new algorithm to differentiate between scar and viable myocardium developed by simultaneous integration of computed tomography and contrast-enhanced magnetic resonance imaging. Eur Heart J 2013;34(8):586-96.
5. Piers SR, Tao Q, van Huls van Taxis CF, Schalij MJ, van der Geest RJ, Zeppenfeld K. Contrast- enhanced MRI-derived scar patterns and associated ventricular tachycardias in nonischemic cardiomyopathy: implications for the ablation strategy. Circ Arrhythm Electrophysiol 2013;6(5):875- 83. 6. de Bakker JM, Coronel R, Tasseron S, Wilde AA, Opthof T, Janse MJ, van Capelle FJ, Becker AE, Jambroes G. Ventricular tachycardia in the infarcted, Langendorff-perfused human heart: role of the arrangement of surviving cardiac fibers. J Am Coll Cardiol 1990;15(7):1594-607.
7. Wijnmaalen AP, van der Geest RJ, van Huls van Taxis CF, Siebelink HM, Kroft LJ, Bax JJ, Reiber JH, Schalij MJ, Zeppenfeld K. Head-to-head comparison of contrast-enhanced magnetic resonance imaging and electroanatomical voltage mapping to assess post-infarct scar characteristics in patients with ventricular tachycardias: real-time image integration and reversed registration. Eur Heart J 2011;32(1):104-14.
8. Schmidt A, Azevedo CF, Cheng A, Gupta SN, Bluemke DA, Foo TK, Gerstenblith G, Weiss RG, Marban E, Tomaselli GF, Lima JA, Wu KC. Infarct tissue heterogeneity by magnetic resonance imaging identifies enhanced cardiac arrhythmia susceptibility in patients with left ventricular dysfunction. Circulation 2007;115(15):2006-14.
9. Nazarian S, Bluemke DA, Lardo AC, Zviman MM, Watkins SP, Dickfeld TL, Meininger GR, Roguin A, Calkins H, Tomaselli GF, Weiss RG, Berger RD, Lima JA, Halperin HR. Magnetic resonance assessment of the substrate for inducible ventricular tachycardia in nonischemic cardiomyopathy.
Circulation 2005;112(18):2821-5.
10. Yan AT, Shayne AJ, Brown KA, Gupta SN, Chan CW, Luu TM, Di Carli MF, Reynolds HG, Stevenson WG, Kwong RY. Characterization of the peri-infarct zone by contrast-enhanced cardiac magnetic resonance imaging is a powerful predictor of post-myocardial infarction mortality.
Circulation 2006;114(1):32-9.
