Subcellular behavior model enables highly precise temporal super-resolved live-cell imaging

Live-cell imaging is essential for studying dynamic cellular processes but is limited by temporal resolution constraints due to photobleaching and phototoxicity. Video frame interpolation techniques have been applied to address this. However, existing pixel-level methods are insufficient in capturing subcellular dynamics, as they focus mainly on appearance features while neglecting other subcellular characteristics. Here, we model the behavior of subcellular particles and combine behavioral features with appearance features. This approach enhances the accuracy of particle matching and improves interpolation outcomes. We tested our approach on simulated and real datasets and evaluated its performance in downstream particle tracking tasks. Compared to other interpolation methods, our approach effectively handles a variety of challenging scenarios, such as the sudden appearance or disappearance of particles, particle overlap, and motion pattern changes. Consequently, our method offers a novel approach to improve the temporal resolution of subcellular imaging and reduce sample exposure time and phototoxicity.