Time-lapse RESOLFT nanoscopy empowered by deep learning.

Deep learning strategies enable 10-fold reduction in light dose and 4-fold increase in frame rate for RESOLFT nanoscopy, enhancing live cell imaging.

• Why it matters: Current RESOLFT microscopy is limited by photobleaching and spatial sampling, restricting the duration and temporal resolution of live cell imaging, which hampers understanding dynamic cellular processes.
• What they did: Researchers developed two deep learning-based approaches: Hierarchical DivNoising (HDN) to restore low-SNR images and extend imaging duration, and Fast-MoNaLISA to accelerate acquisition time and increase frame rate, tested on RESOLFT data.
• The result: These methods enable longer, faster imaging at sub-organelle resolution, facilitating observation of dynamic cellular events such as vesicle budding and actin meshwork formation, advancing live cell nanoscopy.