Panoramic Voltage-Sensitive Optical Mapping of Contracting Hearts using Cooperative Multi-View Motion Tracking with 12 to 24 Cameras

TL;DR

The paper addresses the challenge of imaging high-resolution action potential waves on contracting hearts, where motion artifacts and reliance on contraction-inhibited preparations have limited electromechanical studies. It introduces a panoramic voltage-sensitive optical mapping system with cooperative multi-view 3D motion tracking across 12 cameras in a soccer-ball imaging chamber to reconstruct the beating ventricular surface and map electrical activity concurrently with mechanics. The authors demonstrate full-surface measurements during sinus rhythm, paced rhythms, and ventricular fibrillation in intact isolated rabbit hearts, achieving submillimeter electrical resolution (~120 μm per pixel) and millimeter-scale mechanical resolution across thousands of surface elements, including observations of electromechanical waves and VF vortex patterns, as well as APD changes under potassium-channel blockade. This approach enables unprecedented insight into heart electromechanics, with potential to inform diagnostics, disease mechanisms, and electromechanical heart models at high spatiotemporal resolution.

Abstract

Voltage-sensitive fluorescence imaging is widely used to image action potential waves in the heart. However, while the electrical waves trigger mechanical contraction, imaging needs to be performed with pharmacologically contraction-inhibited hearts, limiting studies of the coupling between cardiac electrophysiology and tissue mechanics. Here, we introduce a high-resolution multi-camera optical mapping system with which we image action potential waves at high resolutions across the entire ventricular surface of the beating and strongly deforming heart. We imaged intact isolated rabbit hearts inside a soccer-ball shaped imaging chamber facilitating even illumination and panoramic imaging. Using 12 high-speed cameras, ratiometric voltage-sensitive imaging, and three-dimensional (3D) multi-view motion tracking, we reconstructed the entire 3D deforming ventricular surface and performed corresponding voltage-sensitive measurements during sinus rhythm, paced rhythm, and ventricular fibrillation. Our imaging setup defines a new state-of-the-art in the field and can be used to study the heart's electromechanical physiology during health and disease at unprecedented resolutions. For instance, we measured electrical activation times and observed mechanical strain waves following electrical activation fronts during pacing, observed electromechanical vortices during ventricular fibrillation, and measured action potential duration and contractile changes in response to pharmacological blockage of potassium ion channels.

Figures (46)

• Figure 1: Panoramic voltage-sensitive optical imaging of action potential waves across the entire surface of contracting and strongly deforming isolated hearts. A Spherical 3D-printed soccer ball-shaped imaging chamber with 24 windows, 12 synchronized high-speed video cameras (acA720-520um, Basler) and constant-pressure Langendorff-perfusion system. B View of isolated rabbit heart inside imaging chamber, see also Supplementary Video \ref{['video:SV1']}. C Close-up of isolated rabbit heart immersed in Tyrode solution inside imaging chamber. D Camera poses (position and orientation) and 3D reconstruction of rabbit heart (not to scale) obtained with 12 cameras, see also Supplementary Video \ref{['video:SV2']}. Imaging was performed at 500 fps with 8 mm and 12 mm lenses with shorter and longer working distances, respectively. E Technical drawing (side & top view) of imaging chamber together with camera poses (right). Horizontal imaging was performed through windows 1-9 and apical imaging through windows 10-12. F Continuous green or pulsed green-blue excitation in odd (1,3,5, ...) and even (2,4,6,...) frames, respectively, with up to 48 light-emitting diodes (LEDs), see also Fig. \ref{['fig:SupplementRatiometry']} for more details. In pulsed mode, we performed 2 separate 3D reconstructions using the green and blue video data and subsequently tracked the heart's motion (red vectors, frame-to-frame displacements) over time using the blue video data, see Methods for details and Supplementary Videos \ref{['video:green-blue-video-3D']} and \ref{['video:green-blue-video-3D-texture']}. Hearts were stained with voltage-sensitive fluorescent dye (Di-4-ANEPPS). G Optical traces (normalized dimensionless units) obtained from a $3 \times 3$ pixel region at the center of the same triangle (black dot in panel F) before (top, green video) and after 3D motion tracking (center, green and blue videos), and after both 3D tracking and ratiometric compensation (bottom), see also Supplementary Video \ref{['video:SV4']}. While the green trace (obtained with green excitation) carries the signal, the blue trace is neutral and serves as a measurement for local illumination changes which arise due to relative motion between the tissue and the LEDs.
• Figure 2: Panoramic 3D reconstruction, tracking, and voltage-sensitive optical mapping of moving heart using multi-view cooperative numerical motion tracking with 12 calibrated cameras. A High-resolution raw mesh resulting from static 3D reconstruction (photogrammetry-like technique or patch-matching with COLMAP) showing 3D rabbit heart surface during systole (160m s). The motion of the heart is initially represented by a sequence of independent raw meshes which each consist of an uncorrelated number of vertices and faces, see Supplementary Video \ref{['video:SV3']}. B The motion is tracked over time using a template mesh (black wireframe mesh), which typically consists of approx. 2,000-10,000 (triangular) polygon faces. 3D Tracking is performed using a multi-view vertex-based mesh tracking approach developed by Klaudiny et al. Klaudiny2011, see also panel E. C After tracking, the moving mesh is texturized with the grayscale video data averaging data from multiple cameras, see also Supplementary Video \ref{['video:SV2']}. D Grayscale texture map containing thousands of triangles used to texturize the mesh in C. The ventricular surface is resolved by about 0.5 - 1.0 million pixels, as each triangle contains approx. 50-150 pixels, at a spatial resolution of about 120µm per pixel. The pixels are superpositions of pixels from multiple cameras, see also Fig. \ref{['fig:SupplementCoverage']}. E Displacement vectors (red) indicating shifts of epicardial surface from diastole (0 ms, shortly before depolarization) to systole (160 ms). Close-up: Tracking is performed per vertex (here shown for vertex P1): a mesh-based kernel consisting of a set of sample points (gray, illustration) surrounding each vertex is projected into and used to identify tissue movements in each camera, see Klaudiny2011 for details and Figs. \ref{['fig:SupplementTrackingElectrode']} and \ref{['fig:SupplementTrackingKernel']}. F Mesh vertices (blue points) superimposed on texturized mesh (3D model) and projected into one of the camera images (2D). The projected vertices follow the motion of the heart in each individual video (e.g. points $P_A, P_B, P_C$), see also Supplementary Video \ref{['video:SV5']}.
• Figure 3: Ventricular wall motion imaged with voltage-sensitive 3D optical mapping during sinus rhythm, see also Figs. \ref{['fig:Sinus2']} and \ref{['fig:Sinus3']}. Wall motion is tracked in 3D space using video data obtained with 12 calibrated high-speed cameras, see Figs. \ref{['fig:Setup']} and \ref{['fig:Fig2']}. A Wireframe meshes of epicardial surface in diastolic, stress-free (black) and deformed, contracted (red) mechanical configuration. The mesh consists of $3,325$ vertices or $6,349$ polygon triangles, respectively. Average nearest-neighbour distance: 0.92 ± 0.21m m (black mesh). B Raw video images of left ventricle (LV) projected onto mesh (same heart as in Fig. \ref{['fig:Sinus1']}A,B, anterior wall). C 3D displacement vector field (red) indicating motion at 32 ms and 108 ms, respectively, with respect to diastolic, stress-free mechanical configuration at 0 ms. D Apical view with displacement vectors (red) indicating motion with respect to the previous frame. The fastest contractile motion occurs between 30 - 50 ms, and the ventricles exhibit rotational, torsional motion around the apex, see also Supplementary Videos \ref{['video:SV2']} and \ref{['video:sinus']}.
• Figure 4: Sinus rhythm imaged in a contracting rabbit heart with voltage-sensitive multi-camera optical mapping with 12 cameras, see also Figs. \ref{['fig:Setup']}, \ref{['fig:Sinus1']}, \ref{['fig:Sinus3']}, \ref{['fig:SupplementSinusAPTraces']}-\ref{['fig:SupplementBaselineCorrection2']} and Supplementary Video \ref{['video:sinus']}. Using 3D motion tracking, we imaged action potential waves and wavefronts across the entire contracting and strongly deforming ventricular surface (LV/RV: left/right ventricle). A Raw grayscale video data mapped onto 3D reconstructed ventricular surface together with motion vectors (red) indicating motion and deformation of the ventricular surface with respect to the mechanical configuration at 0 ms (beginning of depolarization). B Action potential wave front (dark: upstroke, high negative rate of change in fluorescence intensity) imaged with continuous green excitation propagating across ventricular surface triggering the contraction and deformation of the heart, see also Fig. \ref{['fig:Sinus3']}A,B). Noticeable contraction sets in about 30 ms after the beginning of the depolarization of the ventricles, c.f. Fig. \ref{['fig:Sinus2']}C,D). C Action potential wave (bright: resting tissue, black: depolarized tissue / plateau phase, gray: repolarizing tissue) measured with pulsed green-blue ratiometric excitation across ventricular surface in strongly contracting heart. Corresponding optical traces shown in Figs. \ref{['fig:Setup']}G), \ref{['fig:Sinus3']}D) and \ref{['fig:SupplementSinusAPTraces']}. Motion vectors (red) as in A) (but with different scaling).
• Figure 5: Electrical activation, strain, and action potential duration (APD) mapped across entire ventricular surface of isolated rabbit heart during sinus rhythm. A Electrical activation map with local activation times (black: early/0 ms, yellow/white: late/20ms; anterior, posterior, apical view, etc.) computed form the action potential upstroke shown in panel B) and Fig. \ref{['fig:Sinus1']}B). The activation map indicates multiple early-activation or epicardial break-through sites (presumably correlated with the Purkinje system). The entire ventricles get electrically activated within 20 ms. B Action potential wavefront (black: upstroke, high negative rate of change of signal; 0 ms: beginning depolarization, continuous green excitation) as seen from anterior and apical walls, respectively, c.f. Fig. \ref{['fig:Sinus2']}B). C Strain measured as local triangular area change across epicardial surface (blue: dilated, red: contracted, 1.2/0.8 = 20% area increase/decrease, anterior wall). Bottom left: Undeformed, stress-free epicardial surface (0 ms: beginning depolarization, anterior wall). Bottom center and right: Stress-free and deformed epicardial surface (0 vs. 200 ms, posterior perspective), c.f. Fig. \ref{['fig:Sinus2']}. D Top: Comparison of contraction (red vectors, frame-to-frame displacements) with normal and low potassium barium-chloride Tyrode. Upper plot: Action potential measured on posterior wall (black dot in panel E) during normal (black) and abnormal sinus rhythm with low-potassium barium-chloride Tyrode (gray) causing delayed repolarization and prolonged action potential duration (APD). The optical traces were measured at the center of the same triangle (averaged from 7 $\times$ 7 pixels, ratiometric imaging with pulsed green-blue excitation). Center plot: Strain measured as local trianglular area change (normalized units with respect to intial triangle area) is comparable but rate of contraction is faster/stronger with delayed repolarization (arrows). Bottom plot: Contractile speed (magnitude of frame-to-frame displacements) shows that the heart contracts more strongly and relaxes earlier with low potassium barium-chloride Tyrode (arrows). E APD maps computed for normal APD across entire ventricular surface and for prolonged APD on posterior wall during sinus rhythm (APD-90 at 10% height of the action potential). Histogram showing peaks at 170m s $\pm$ 5m s with normal Tyrode and 210m s $\pm$ 5m s with low-potassium barium-chloride Tyrode (only posterior wall). Traces in D) sampled from black and gray dots (same location, same triangle). Heart in panels D,E) different than in panels A-C).