Real-time structured illumination microscopy based on Fourier ptychographic

Structured illumination microscopy is more suitable for super-resolution imaging of live cells and holds promising prospects in bioscience thanks to its fast imaging speed and low photodamage. However, in order to achieve fast, real-time and long-term imaging of live cells, further enhancements in imaging speed and reductions in photodamage are necessary. In this work, we optimize from both algorithm and hardware aspects. First, a Fourier ptychographic SIM algorithm (FP-SIM) using only three patterns is proposed to reduce the number of frames required for reconstruction. Second, we use CUDA to design appropriate parallel methods on hardware to accelerate the algorithm. We demonstrate that GPU improves the speed by nearly 200-fold in the most time-consuming iteration part of the algorithm. At 512×512 pixels, real-time super-resolution microscopy imaging at approximately 16Hz can be achieved. Faster imaging speed and less photodamage enable our method to provide a promising tool for life science research and biomedical measurement.