Quantifying The Imaging Performance Of Low-light Cameras In Super-resolution Localization Microscopy

With the recent rapid development, super-resolution localization microscopy (SRLM) has been an important tool for biology research at nanoscale. Furthermore, it has the potential to provide key technical support for the worldwide Human Brain Project. As is well known, fluorescence probe, low-light camera and localization algorithm are the most essential factors for providing high quanlity super-resolution images considering super-resolution localization microscopy technique is naturally a single molecule imaging technique. Serving as a connection link between the preceding single molecule signal generation and the following single molecule localization, low-light camera records the single molecule signal and thus affects the final image resolution. Therefore, selecting a suitable low light detector is undoubtedly critical for achieving optimal localization performance, especially in the study of ultrafine structure where tiny resolution improvement would have great benefit. However, there is still no systematic study and direct experimental comparison on the performance of low-light cameras. Here, we present a direct and detailed experimental comparison on the performance of low-light cameras taken into consideration the effects of current fluorescent probes and camera-specific algorithms, the main contents are as follows:(1) Theoretical basis for quantifying the imaging performance of low-light cameras in SRLM. We analysed the interrelationship of fluorescence probe, low-light camera and localization algorithm, then pointed that the performance of cameras relied on the features of signal and localization algorithm. Based on this conclusion, we pointed out the deficiencies of previous works and completed camera-specific algorithms. This will lay the foundation for the next experimental and analytical methods.(2) Experimental basis for quantifying the imaging performance of low-light cameras in SRLM. We improved our optical setup for direct comparing two low-light cameras to rigorously control experiment quality. The detection path was designed delicately so that the splitting ratio is close to1:1and the pixel size at sample plane for both cameras is matched (100nm). With these improments, we believed that EMCCD and sCMOS cameras can record images at identical conditions. After that, we characterized the system stability and ability of single molecule imaging, making sure that our system was competent enough for this study.(3) Quantifying the imaging performance of low-light cameras in SRLM based on rigorous data analysis method. Using normal and camera-specific MLE algorithms, we regarded the localization precision as primary index to compare the performance of EMCCD and sCMOS in traditional signal and background conditions. Our results showed:(i) With normal MLE algorithms, the EMCCD camera provides better localization precision than sCMOS at weak signal and background level, benefiting from relative low read noise. While at relatively high signal intensities, both cameras exhibit comparable localization performance. In addition, the EMCCD camera was more susceptible to higher background due to the excess noise, and thus leads to lower localization precision;(ii) on the other hand, for the EMCCD camera, using the camera-specific MLE algorithm is not indispensable since it provides similar or only slightly better localization precision as the normal MLE at typical signal levels; but for the sCMOS camera, it is advantageous to use the camera-specific MLE algorithm when the signal is relative weak (corresponding to most fluorescence proteins used in SRLM);(iii) benefiting from the camera-specific MLE algorithm, the sCMOS camera exhibits excellent localization performance in most traditional cases. Finally, we compared the FRC resolution of reconstructed super-resolution images in cell sample and verified the conclusions mentioned above.