High-speed image reconstruction in dual-channel super resolution FRET microscopy

Structured illumination-based super-resolution Förster resonance energy transfer microscopy (SIM-FRET) enables the investigation of molecular structure and function in sub-diffraction regions of live cells. However, limitations in acquisition and reconstruction speeds pose challenges for fast SIM-FRET imaging. To address this, we integrated a GPU-accelerated SIM-FRET reconstruction method with a dual-channel SIM-FRET microscope, enabling near real-time image reconstruction. We optimized the cross-correlation parameter estimation algorithm (iCOR-GPU) by implementing kernel-level GPU parallel computing, reducing parameter estimation time from ∼2837 ms to ∼34.7 ms. Additionally, fine-grained timing control via FPGA improved acquisition speed from ∼249 ms to ∼66 ms. Using the dual-channel SIM-FRET microscope with iCOR-GPU, we successfully monitored mitochondrial dynamic tubulation and cristae remodeling in living MCF7 cells. SIM-FRET imaging of U2OS cells expressing ActA-G17 M demonstrated accurate FRET efficiency and acceptor-to-donor concentration ratio. With a 512 × 512 field of view, we achieved SIM-FRET image acquisition at 15 frames per second and reconstruction at 79 frames per second.