Diffusion in biological tissues is known to be hindered by the structural complexity of the underlying medium. In the heart, improved characterisation on how this complexity influences acquired diffusion weighted signals is key to advancing our interpretation of diffusion magnetic resonance imaging, as well as to propose novel biomarkers to further characterise myocardial microstructure. In this work, we propose stretched Mittag-Leffler signal decay models for the quantification of the anomalous decay observed in acquired diffusion weighted signals. Our results, analysed in ex vivo healthy, fixed rat ventricles, indicate that such a representation suffices to capture the anomalous signal decay observed in the myocardial syncytium. The subdiffusive order of signal decay is shown to encode independent information to that encapsulated by standard diffusion tensor metrics, and thus may provide additional information on tissue microstructure. Moreover, subdiffusion gradients are shown to be indicative of the total structural heterogeneity spanning the left ventricular wall, which includes progressive myolaminae branching and spatially varying densities of perimysial collagen, microvasculature and adipose tissue. The proposed approach may therefore have important implications for the characterisation of tissue microstructure, both in cardiac and other tissue types.