Flexible Event-Driven Biological Imaging via Bayesian Inference

Abstract

Event-driven imaging is gaining more and more attention in the field of biological imaging, where conventional frame-based imaging techniques confront with the bottleneck of capturing weak and fast-changing biological signals. However, existing methods for processing event camera signals, such as detection, de-noising, etc, are mainly designed for computer vision problems like tracking and surveillance and do not well adapts to biological applications. Particularly, a typical biological imaging system focuses on static scenes with fast-changing and weak illumination. There lacks effective detection algorithms under this extreme lighting condition. To this end, we proposed a novel framework for recovering the weak biological signals from the noisy event camera measurements. Based on Bayesian inference, the proposed approach offers a flexible solution for diverse imaging modalities. We validated our method on real-world biological imaging experiments, demonstrating its robustness and effectiveness of detecting weak and fast varying targets. This unlocks new possibilities for weak biological signal detection under extreme lighting conditions in a variety of imaging applications.