周波数分離相関による高忠実度超解像光学ゆらぎイメージング

High-fidelity super-resolution (SR) imaging is critical for probing subcellular structures and dynamics. While current SR techniques often require complex hardware, SR optical fluctuation imaging (SOFI) enables rapid SR reconstruction on conventional wide-field microscopes by leveraging fluorescence blinking. However, higher-order SOFI suffers from structural artifacts and discontinuities due to statistical errors in cumulant estimation. We propose frequency separation correlation (FSC), a computational method that decomposes pixel-wise blinking signals into single-frequency components via Fourier analysis. By accumulating auto-correlation cumulant of these components while eliminating cross-frequency correlations, FSC enhances statistical precision. FSC-enhanced SOFI outperformed conventional SOFI, balanced SOFI (bSOFI), and super-resolution auto-correlation with two-step deconvolution (SACD) across simulated and experimental datasets. It maintained structural continuity under challenging conditions (e.g., dense labeling, weak blinking, short sequences). Using only 20 frames, FSC-enhanced bSOFI (fs-bSOFI) and fs-SACD achieved lateral resolutions of 96 nm and 92 nm, respectively, while preserving image fidelity. Large-field fs-SACD imaging of microtubule networks (166 × 166 µm²) was accomplished within ∼2 seconds. FSC-enhanced SOFI provides a robust tool for high-speed, high-fidelity SR imaging, particularly for live-cell applications.