Deep learning enables structured illumination microscopy with low light levels and enhanced speed

Abstract

Structured illumination microscopy (SIM) surpasses the optical diffraction limit and offers a two-fold enhancement in resolution over diffraction limited microscopy. However, it requires both intense illumination and multiple acquisitions to produce a single high-resolution image. Using deep learning to augment SIM, we obtain a five-fold reduction in the number of raw images required for super-resolution SIM, and generate images under extreme low light conditions (at least 100× fewer photons). We validate the performance of deep neural networks on different cellular structures and achieve multi-color, live-cell super-resolution imaging with greatly reduced photobleaching.

Turn-key super-resolution mapping of cell receptor force orientation and magnitude using a commercial structured illumination microscope

“…SIM-MFM also enables rapid super-resolution imaging of structural proteins such as paxillin in parallel with super-resolution imaging of force magnitude and orientation, and facile, photostable timelapse imaging of molecular force orientation. We expect that ongoing SIM and pSIM technique developments will continue to improve SIM-MFM acquisition [68][69][70] . Perhaps eventually, SIM will be performed with dipoles that are not parallel to the coverslip, which would enable nondegenerate force vector mapping 40 .…”

Chemical Analysis of Single Cells and Organelles

“…An overview of the recent advances in SRM techniques is given in Figure 3 and discussed more throughout. The SRM techniques shown in Figure 3 include deep learning assisted SIM ( Figure 3 A), 45 multicolor SOFI ( Figure 3 B), 46 raster ( Figure 3 C, left) compared to smart ( Figure 3 C, right) scanning, 47 and automated maS 3 TORM multiplexing ( Figure 3 D). 48 …”

“…(D) Automated maS 3 TORM multiplex setup and experimental workflow. Panel A is reproduced from Jin, L.; Liu, B.; Zhao, F.; et al Nature Communications 2020 , 11 , 1934 (ref ( 45 )). Panel B is reproduced from Grußmayer, K. S.; Geissbuehler, S.; Descloux, A.; et al Nature Communications 2020 , 11 , 3023 (ref ( 46 )).…”

Progress and prospect of biomedical optical imaging

“…深度学习技术通 过学习建立高低分辨率图像之间的映射, 能够从低分 辨图像推断出高分辨图像, 且推断出的高分辨图像与 低分辩图像保持时间分辨率一致 [98] . 深度学习技术可 从低信噪比图像预测其相应的高信噪比图像, 在缩短 曝光时间、减少光毒性的情况下, 实现长时程高时空 分辨率成像 [99] . .…”