Super-resolution fluorescence microscopy is one of the most important breakthroughs in optical microscopy in the 21st century, and is awarded with the Nobel Prize for Chemistry in 2014. Super-resolution localization microscopy (SRLM) is a representative super-resolution imaging technique. With the combination of single-molecule fluorescence microscopy and high-precision molecule localization, SRLM is able to use a simple optical set-up to provide 20~30 nm spatial resolution that enables unprecedented opportunities for biomedical researches.
Originally, SRLM relies on sparse excitation and sparse localization of fluorescence emitters. Thousands or even tens of thousands of raw images are required to accumulate a sufficient number of localized emitters for reconstructing a super-resolution image, and tens of seconds or even several minutes are required for imaging a field of view (FOV), thus limiting the application of SRLM in high-throughput imaging (which images usually more than several hundreds of FOV). In recent years, researchers have tried to develop multi-emitter localization algorithms to maximize the imaging speed of SRLM. Multi-emitter localization algorithm allows more emitters to be localized within a raw image, thus can be used to accumulate a sufficient number of localized emitters using less raw images. However, the slow image analysis speed of reported multi-emitter localization algorithms limits their usage in mostly off-line image processing with small image size.
Under the guidance of Prof. Zhen-Li Huang from Britton Chance Center for Biomedical Photonics, Wuhan National Laboratory for Optoelectronics, Luchang Li (a PhD candidate in Huang’s group) and other group members developed fast maximum likelihood estimation (MLE) algorithm for sparse emitters, weighted MLE for sparse emitters with signal contamination, and MLE algorithms for multi-emitters. They further adopted the well-known divide and conquer strategy in computer science and combined these MLE-based algorithms to present a fitting-based method called QC-STORM for fast multi-emitter localization. QC-STORM achieves 3-4 orders of magnitude in the image processing speed than the popular fitting-based ThunderSTORM and the up-to-date non-iterative WindSTORM, with similar localization precision and detection rate. Therefore, QC-STORM is capable of providing real-time full image processing on raw images with 100 µm × 100 µm field of view and 10 ms exposure time. QC-STORM is beneficial for pushing the development and practical use of SRLM in high-throughput or high-content imaging of a large cell population. This study is reported recently in Optics Express, 22nd July, 2019, Vol. 27, Iss. 15, pp. 21029-21049.
Figure 1. Evaluation of the localization performance among QC-STORM, ThunderSTORM and WindSTORM using experimental 2D images.
