Fig. 4.

An example of model-based image analysis. Detailed methods are available in Supplementary Material. A confocal 3D image of human LV tissue (A) with WGA (green) and Cx43 (magenta) channels was acquired as described in Fig. 2 and 3. WGA ridges were segmented using a watershed filter and morphological operations to give a binary mask (B) of viable myocardium (white). Networks of nodes and edges were constructed within the viable myocardium (grey) (C), and edges weighted with a conductivity distribution field derived from the Cx43 3D image channel (D). Viable myocardium cross-sections are typically complex (red). A reaction-diffusion model of electrical activity was solved on the network for a stimulus on the upper surface (E). Although specified electrical conductivities were isotropic, the activation sequence was anisotropic due to the heterogeneous myocardial coupling abstracted from the WGA and Cx43 images to the network reaction-diffusion model. Potential traces (F) confirm the anisotropic spread of electrical activation and heterogeneous conduction velocity. These results show that relatively simple myocardial coupling derived from detailed imaging is sufficient for emergent functional electrical behaviour that communicates LV tissue structures’ intrinsic topology in the human heart.