Panoramic optical mapping provides rich spatiotemporal information on cardiac electrical activity and is becoming a key tool for studying arrhythmogenic mechanisms in experimental models. This work proposes a novel approach to exploring spatially coherent structures in optical recordings by identifying spatial correlation communities from estimated transmembrane potential dynamics. Starting from panoramic voltage-sensitive dye recordings in two Langendorff-perfused rabbit hearts, we preprocessed the optical signals to obtain segmentations of two cardiac cycles and estimated local transmembrane potentials at each pixel. A low-dimensional manifold was constructed on these multivariate signals, and a neighborhood graph was built to capture local temporal similarities. Louvain community detection on this graph yielded spatially coherent clusters that mapped back to connected anatomical regions, naturally reflecting the spatial correlation patterns induced by cardiac bioelectrical propagation. Compared to classical clustering methods applied directly in the signal space, communities better preserve anatomical continuity and avoid fragmentation upon inverse projection onto the cardiac geometry. Similar patterns emerge when analyzing torso recordings from torso potentials. Rather than treating the data as isolated points, spatial correlation communities leverage the underlying dynamic structure of bioelectric signals, providing a more physiologically adapted framework for analyzing activation patterns. This method can help identify relevant regions in complex arrhythmic dynamics and complements existing analysis tools in optical mapping studies.