2.1.

Animal Studies

All animal studies were approved by the Institutional Animal Care and Use Committee of Washington University. New Zealand white rabbits ( $n=18$ , 2.5 to 3 mos. old, $2\text{to}3\mathrm{kg}$ ) were anesthetized with $100\mathrm{mg}∕\mathrm{kg}$ of pentobarbital and 1000 intravenous units of heparin intravenously, after which a midsternal thoracotomy was performed and the heart removed. The heart was Langendorff perfused with oxygenated ( $95\%{\mathrm{O}}_2$ , $5\%\mathrm{C}{\mathrm{O}}_2$ ) Tyrode’s solution at $37°\mathrm{C}$ and received $50\mu \mathrm{L}$ of $5\text{-}\mu \mathrm{M}$ di-4-ANEPPS (Molecular Probes, Eugene, Oregon) over $5\min$ . The heart was dissected in cold Tyrode’s solution $(0°\mathrm{C})$ to create a preparation containing the atrioventricular (AV) junction, sinoatrial node, or the right ventricular free wall. Each type of preparation was pinned to silicone with the endocardium as flat as possible. The preparations were superfused at $50\mathrm{mL}∕\min$ with perfusate containing $15\mathrm{mM}$ of the excitation-contraction uncoupler 2,3-butanedione monoxime (Sigma, St. Louis, Missouri) to prevent motion artifacts.

2.2.

Optical Mapping

A $16\times 16$ photodiode array (C4675, Hamamatsu, Hamamatsu City, Japan) was used with an optical mapping system that has been described previously.¹⁷ Briefly, a ring of light emitting diodes (LED ring) with a wavelength centered at $520\mathrm{nm}$ illuminated the tissue to excite di-4-ANEPPS. Fluorescent emission passed through a $50\text{-}\mathrm{mm}$ lens (1:1.4, AF Nikkor, Nikon, Tokyo, Japan) and a dichroic mirror, was long pass filtered at $610\mathrm{nm}$ , and focused on the photodiode array. Optical signals were amplified (Innovative Technologies Network, Gaithersburg, Maryland), sampled at $1.5\mathrm{kHz}$ (PCI 6031E, National Instruments, Austin, Texas), and then lowpass filtered at $120\mathrm{Hz}$ . The focusing lens could be adjusted to change the spatial resolution from $0.56\text{to}2\mathrm{mm}$ per photodiode, depending on the preparation. Preparations were digitally photographed and optical mapping fields of view were marked with stainless steel pins in the tissue. Activation times were defined as the time of the maximum first derivative of the fluorescent $\left(F\right)$ signal, $dF∕d{t}_{\mathit{max}}$ , and activation maps were constructed from the activation times of each photodiode. Conduction velocity vector fields were calculated from activation maps using a polynomial fitting method developed by Bayly ¹⁸

2.3.

Tissue Preparation

Following optical mapping experiments, the tissue and silicone disk were placed in 3.7% formaldehyde diluted in phosphate buffered saline (Sigma-Aldrich, St. Louis, Missouri) overnight at $4°\mathrm{C}$ . The following day, samples were transferred to a 20% sucrose solution and kept at $4°\mathrm{C}$ for two days to dehydrate the tissue. Optical coherence tomography imaging was then performed with samples submerged in sucrose solution at room temperature on the second day.

2.4.

OCT

A microscope-coupled OCT system was used to image the samples.¹⁹ A $10\text{-}\mathrm{mW}$ super luminescent diode $1310\text{-}\mathrm{nm}$ light source (Covega SLD 1021, Jessup, Maryland) with $70\text{-}\mathrm{nm}$ resolution was used. The axial and lateral resolution of the system was approximately $10\mu \mathrm{m}$ (in air). Individual OCT images were averaged 3 to 6 times depending on the tissue sample at the time of acquisition.

2.5.

Image Processing

The OCT volumes were acquired as data sets that were $400\times 400\times 650\text{pixels}$ in the $x$ , $y$ , and $z$ dimensions, respectively, corresponding to a tissue volume of $4.5\times 4.5\times 3.2{\mathrm{mm}}^3$ . For each of the preparations imaged with OCT, the tissue surface was never completely horizontal across the entire imaging field of view. Depending on the tissue, surface variations ranged vertically from $0.5\text{to}1.5\mathrm{mm}$ across the $4.5\times 4.5\text{-}{\mathrm{mm}}^2$ horizontal field of view [Fig. 1a ]. Because of the surface variation, slicing the 3D data set in the $z$ -plane created artifacts in the resulting images such as shadows and holes, as shown in Fig. 1b. To generate OCT images of the tissue parallel to the surface, we defined a new coordinate system in the $z$ dimension $\left(z^{\prime }\right)$ , where $z^{\prime }=0$ corresponded to the tissue surface [Fig. 1c].

Fig. 1

(a) The surface of a representative OCT data set. Shading represent the $z$ coordinate of the surface. As shown in panel (b), sectioning the OCT data in the $z$ dimension creates an image with holes due to plane of section and shadows due to the variable depth of the section from the surface. (c) A new coordinate system $z^{\prime }$ was defined with $z^{\prime }=0$ corresponding to the tissue surface. Sectioning the tissue in $z^{\prime }$ created images that were free from artifacts due to plane of section, as seen in panel (d), an image $0.15\mathrm{mm}$ below the surface. This data set was sectioned in $z^{\prime }$ until fibers were no longer recognizable, which occurred at a depth of $z^{\prime }=0.9\mathrm{mm}$ .

To define the tissue surface, each $xz$ scan was filtered with a two-dimensional Weiner filter (MATLAB, The Math Works, Natick, Massachusetts) to remove noise before the $xz$ scans were combined to generate a 3D data set. Next, an intensity threshold was defined, and each column of the 3D data set was searched in the $z$ dimension for the pixel where the threshold was crossed. This point in each column corresponded to the tissue surface. All surface points were combined in three dimensions to generate the surface of the tissue. Once the surface was defined, images parallel to the surface were created by stepping through the data set below the surface by one pixel at a time in the $z^{\prime }$ dimension [Fig. 1d].

2.6.

Fiber Orientation Determination

To calculate fiber orientation from two-dimensional (2D) OCT images,²⁰ a modification of an algorithm described in Karlon ²¹ was used. First, preprocessing of the data is necessary because of the low contrast nature of OCT images. A Wiener filter was used to remove noise and a 2D Butterworth high pass filter was used to remove the low frequency background intensity (MATLAB).

The $x$ and $y$ intensity gradients ( $G_x$ and $G_y$ , respectively) were computed by convolving the horizontal and vertical Sobel filters with the 2D $z^{\prime }$ OCT images. For each pixel, the magnitude and angle of the gradient, $G(i,j)$ and ${\theta }^{\prime }(i,j)$ , were calculated

2.7.

Histology

Following OCT imaging, the tissue was placed in optimal cutting temperature medium (Thermo Fisher Scientific, Waltham, Massachusetts), frozen, and kept at $-80°\mathrm{C}$ until use. Tissue was sectioned into $16\text{-}\mu \mathrm{m}$ sections and stained with Masson trichrome.

3.1.

OCT of Myocardium

The optical properties of cardiac fibrous tissue and fat create optical contrast within the myocardium in OCT images. Figure 2a illustrates the rabbit atrioventricular junction (AVJ), and corresponding OCT volumes are shown in Fig. 2b. Comparing corresponding histology and OCT images in Figs. 2c and 2d, and Figs. 2e and 2f, it is clear that the fibrous tissue (which appears blue in Masson’s trichrome histology) and adipose tissue surrounding the conduction system in the rabbit AVJ produce less backscatter in OCT images than the myocardium, providing the necessary contrast to visualize the AVJ conduction system with OCT, free of sectioning artifacts usually present in histology.

Fig. 2

(a) Preparation of the rabbit AVJ is shown, with the anatomic landmarks that surround the AV conduction system, and the location and components of the conduction system are shown in green. (b) Surface images of three OCT data sets taken from the AVJ. (c, d) Masson trichrome and the corresponding OCT image of the conduction system. Arrows denote areas of fibrous tissue within the conduction system that can be seen in the OCT image. (e, f) Masson trichrome and the corresponding OCT image of the conduction system at a different location. AVN: atrioventricular node; CS: coronary sinus; IAS: interatrial septum; INE: inferior nodal extension; His: bundle of His; tT: tendon of Todaro; Tricuspid: tricuspid valve; VS: ventricular septum. (Color online only.)

3.2.

Optical Mapping Coupled with OCT

Cardiac conduction is faster in the direction of fiber orientation than transverse to fibers, due to factors such as cell orientation, size, geometry, and gap junctional distribution.²² Therefore, by coupling optical mapping with OCT, we expected to find a strong correlation between conduction velocity and fiber orientation. To validate the coupling of OCT and optical mapping, we initially used a 2D system so that optical signals originating below the surface of the tissue did not confound our interpretation of the data. For this purpose, we used the rabbit sinoatrial node (SAN, $\mathrm{n}=5$ ) because the rabbit SAN is $200\text{to}300\mu \mathrm{m}$ thick and several centimeters wide²³ and, therefore, is essentially two dimensional with one layer of tissue contributing to optical signals [Fig. 3a ]. Figure 3b displays an activation map recorded during spontaneous SAN activity. Activation began in the middle of the SAN and spread anisotropically, which is indicated by conduction velocity vectors. Velocity vectors were binned in $20\text{-}\mathrm{deg}$ intervals in Fig. 3c, which indicated that conduction spread most rapidly at $19\mathrm{cm}∕\mathrm{s}$ at an angle of $130\mathrm{deg}$ . Conduction was slowest orthogonally at $3\mathrm{cm}∕\mathrm{s}$ .

Fig. 3

(a) Rabbit sinoatrial node (SAN) preparation, with the crista terminalis (CT), right atrial free wall (RA), and superior vena cava (SVC) marked. The optical mapping and OCT fields of view (FOV) are indicated. Small blue diamond marks location of leading pacemaker. (b) Activation map of pacemaking activity in the SAN as determined with optical mapping. Conduction spread anistropically in an elliptical pattern. The block zone denotes an inexcitable region of tissue. Vectors indicate direction and magnitude of conduction velocity. (c) Plot of conduction velocity as a function of angle, as well as a histogram of the fiber orientation vectors shown in panel (f). (e) An OCT image taken $100\mu \mathrm{m}$ below the surface. (f) Same image with fiber orientation vectors superimposed. No fiber vectors were analyzed in the trabeculated section of the image because of the artifacts induced by trabeculae with our method of sectioning. (Color online only.)

A $z^{\prime }$ OCT image of the SAN $100\mu \mathrm{m}$ below the surface is shown in Fig. 3d with the leading pacemaker location indicated, and fiber orientation of the tissue is shown in Fig. 3e. A histogram of the fiber vectors is shown in Fig. 3c, which indicates that the dominant fiber angle in the tissue is $130\mathrm{deg}$ . Therefore, the direction of maximal conduction velocity was aligned with the fiber orientation of the tissue. We found the same result in all studied SAN preparations $(n=5)$ .

3.3.

OCT Can Resolve Multilayer Conduction Seen in Optical Mapping

As mentioned previously, optical mapping data consists of fluorescence that originates not only in the tissue surface, but also from tissue layers below the surface. In tissue with multiple layers of conduction such as the AVJ (Fig. 2), optical action potentials (OAPs) contain multiple components originating in different layers. The upstroke of individual OAP components can be used to create activation maps and conduction velocity vectors to determine the direction of wave front propagation. As shown in Fig. 3, conduction velocity vectors and fiber orientation vectors are strongly correlated. Therefore, we hypothesized that fiber orientation vectors obtained with OCT and conduction velocity vectors constructed from OAP components recorded from multilayer tissue could be used to identify which layer of tissue was the origin of individual optical signal components.

Bimodal imaging in the AVJ was performed $(n=7)$ to reconstruct the origin of OAP components. Figure 4a shows an activation map of retrograde conduction through the AVJ that was induced by pacing the His bundle. Figure 4b displays the OCT volume from this preparation. As can be seen in Figs. 4c, 4d, 4e, components of the activation map can be assigned to different layers of tissue in the OCT volume based on conduction velocity and fiber orientation vectors. In Fig. 4c, conduction spread from right to left at $20\mathrm{cm}∕\mathrm{s}$ . An example OAP in Fig. 4c exhibits two distinct upstrokes. The portion of the activation map shown in Fig. 4c was constructed from any OAP that contained the initial upstroke. In the deep tissue layers seen in OCT ( $700\text{to}1100\mu \mathrm{m}$ below the surface), the fibers of the conduction system were oriented approximately parallel to the conduction velocity vectors [Fig. 4c].

Fig. 4

(a) Activation map of retrograde conduction recorded from the AVJ, where activation began at the lower right, spread to the left, and after a delay spread quickly across the rest of the tissue. Black lines represent lines of conduction block. (b) Volume of OCT data recorded from the same AVJ. The colored lines represent the location of the OCT images shown in panels (c) to (e). (c) Initial $15\mathrm{ms}$ of the activation map shown in panel (a), with conduction velocity vectors. An OAP recorded from photodiode marked is also shown, with the dotted lines representing the time window corresponding to the $15\mathrm{ms}$ of the activation map. The OCT image indicates the location of the conduction system and the fiber orientation that was present in the image. (d) Subsequent $18\mathrm{ms}$ of the activation map shown in panel (a) with an OAP recorded from the photodiode marked, and the dotted lines representing the time period shown in the activation map. The OCT image indicates the interface between the conduction system and the atrial myocardium and the fiber orientation in the image. (e) Final $7\mathrm{ms}$ of the activation map shown in panel (a), with conduction velocity vectors and OAP recorded from the photodiode marked and the dotted lines representing the time period shown in the activation map. The OCT image indicates the fiber orientation in the superficial atrial myocardium. (Color online only.)

In Fig. 4d, conduction propagated very slowly with OAPs showing no distinct upstroke during this period. Instead, the small initial depolarization of the OAP in Fig. 4d is consistent with electrotonic activation in this region. The OCT data from this region shows that $400\mu \mathrm{m}$ below the tissue surface is the interface between the conduction system and the atrial myocardium, where conduction spread transverse to fiber orientation. Finally, the activation map in Fig. 4e, constructed from OAPs that only contained one upstroke that corresponded in time to the second upstroke seen in Fig. 4c, demonstrates that activation spread rapidly $(65\mathrm{cm}∕\mathrm{s})$ in all directions. Optical coherence tomography of the superficial layer of atrial tissue, $50\mu \mathrm{m}$ below the surface, revealed that fiber orientation in the atrial tissue matched the initial direction of rapid conduction. Accounting for the 3D conduction of activation from the conduction system below the surface to the atrial layer above, conduction spread approximately $3\text{to}5\mathrm{cm}∕\mathrm{s}$ between the two layers transversely to fiber orientation. Similar 3D reconstructions of conduction were reproduced in all studied AVJ preparations $(n=7)$ .

3.4.

Structural Heterogeneities Contributing to Arrhythmia

Optical coherence tomography and optical mapping can also be used to investigate structural heterogeneities contributing to arrhythmias. Previous studies have shown that heterogeneities such as disrupted cell-cell coupling and scar tissue can provide the substrate to anchor reentrant arrhythmias,^{24, 25} converting paroxysmal arrhythmia to sustained monomorphic tachycardia. Here, we explored the role of microscopic structural heterogeneities in arrhythmia maintenance in the rabbit isolated right ventricle [Fig. 5a ]. Reentrant arrhythmias were induced in four preparations, and reentry tended to anchor at only one or two locations. These reentry core locations were imaged with OCT. In Fig. 5a, a preparation with two reentry cores is shown, with a phase map of reentry shown in Fig. 5b. Surface OCT imaging of each core is shown in Figs. 5c and 5d. The endocardium in Fig. 5c is highly trabeculated, and we postulate that this contributed to reentry anchoring at this location.

Fig. 5

(a) Picture of rabbit right ventricular free wall preparation. Arrhythmias were anchored in two locations in this preparation marked by the yellow and green boxes as reentry core 1 and reentry core 2. Optical coherence tomography was performed on these regions. (b) Phase map of reentrant arrhythmia, which rotated around reentry core 2 located in the yellow box. (c, d) Surface images of OCT data sets recorded from reentry cores 1 and 2 marked with the the green and yellow locations shown in (a). The complex surface in (c) may have been responsible for anchoring arrhythmias in this location, however the surface of (d) is relatively smooth. (e to g) Progressively deeper OCT images of the data shown in (d) sectioned in $z^{\prime }$ from 60, 200, and $350\mu \mathrm{m}$ below the surface, respectively, with fiber orientation vectors color coded according to angle. (h) Distributions of the fiber orientation vectors shown in (e to g), which indicate a shift in dominant angle from the superficial layers to the deeper layers in the tissue. Blue, red, and black shading in (h) represent the color coding of the vectors in (e to g). (Color online only.)

The reentry core of Fig. 5d shows little or no endocardial structural heterogeneity. However, OCT imaging sectioned in the $z^{\prime }$ dimension indicated two distinct myocardial sheets merged in this area. Figures 5e, 5f, 5g show OCT images and fiber orientation vectors from superficial to deeper layers: 60, 200, and $350\mu \mathrm{m}$ below the surface, respectively. The dominant fiber angle changed from approximately $70\mathrm{deg}$ in the superficial layer to $130\mathrm{deg}$ below it [Fig. 5h]. Although this area appears quite smooth on the endocardium, the complicated fiber structure revealed with OCT may have played a role in anchoring reentrant arrhythmias in this region.

3.5.

Imaging Tissue Damage

Functional studies of cardiac tissue often induce damage, either intentionally in the form of ablations, or unintentionally due to pacing. In Fig. 6a , an AVJ preparation was paced sequentially at each circle from an electrode that roamed in a grid pattern throughout the AVJ. A surface image of the OCT data set [Fig. 6b] clearly indicated that pacing at each location produced a small amount of tissue damage. As shown in Fig. 6c, a section perpendicular to the surface through one of the areas of tissue damage showed tissue swelling and decreased backscattering surrounding the area of injury, indicating that edema surrounded the injury. In addition, it is clear that the area of injury was confined to the superficial atrial layer; the His bundle below the surface was unaffected.

Fig. 6

(a) Preparation of the rabbit AVJ is shown, with different pacing locations marked with circles. The field of view of OCT imaging is indicated. (b) Surface image of OCT data recorded from this preparation. Many areas of tissue damage (marked with black arrows) are present, with the white arrowhead marking the damaged area shown in panel (c). (c) An OCT image perpendicular to the surface shown in (b). The area of tissue damage is surrounded by an area of reduced backscatter. The image shows that the tissue damage did not extend to the His bundle. CS: coronary sinus; IAS: interatrial septum; His: bundle of His; VS: ventricular septum.