Structured illumination-based super-resolution Förster resonance energy transfer microscopy (SIM-FRET) is a powerful tool for resolving molecular interactions within sub-diffractionlimited structures. However, the heavy computational burden of estimating illumination parameters hinders real-time SIM-FRET observation. To address this, we propose an optimized partial cross-correlation (pCOR) algorithm that selectively extracts high–signal-to-noise-ratio (SNR) spectral regions from raw SIM data, significantly reducing computational time while maintaining accuracy. By integrating pCOR with multi-threaded parallelization and GPU acceleration, we reduced the per-frame processing time from ~3 s to ~20 ms for SIM and from ~10 s to ~55 ms for SIM-FRET, using a 512 × 512 field of view. By monitoring dynamic events of mitochondrial tubules and cristae in live cells, we demonstrated the system’s real-time imaging capability. In addition, we achieved accurate SIM-FRET efficiency measurements in U2OS cells. In summary, by combining algorithmic optimization with GPU acceleration, this study demonstrates the feasibility and practicality of real-time SIM-FRET imaging in live cells.
Shang et al. (Sat,) studied this question.