The development of voltage-sensitive dyes has revolutionized cardiac electrophysiology and made optical imaging of cardiac electrical activity possible. Photon diffusion models coupled to electrical excitation models have been successful in qualitatively predicting the shape of the optical action potential and its dependence on subsurface electrical wave orientation. However, the accuracy of the diffusion equation in the visible range, especially for thin tissue preparations, remains unclear. Here, we compare diffusion and Monte Carlo (MC) based models and we investigate the role of tissue thickness. All computational results are compared to experimental data obtained from intact guinea pig hearts. We show that the subsurface volume contributing to the epi-fluorescence signal extends deeper in the tissue when using MC models, resulting in longer optical upstroke durations which are in better agreement with experiments. The optical upstroke morphology, however, strongly correlates to the subsurface propagation direction independent of the model and is consistent with our experimental observations.