FW6D.4.pdf
Frontiers in Optics 2017 © OSA 2017
Increasing Acquisition Speeds in Structured Illumination Microscopy and its Limits Florian Ströhl and Clemens F. Kaminski Department of Chemical Engineering and Biotechnology, University of Cambridge, Philippa Fawcett Drive, CB3 0AS Cambridge, UK
[email protected]
Abstract: We present a framework for structured illumination microscopy (SIM) reconstructions that enables superresolution from 3 raw frames. This pushes SIM’s speed to its boundaries but also defines a limit beyond which deconvolution microscopy is superior. OCIS codes: (100.6640) Superresolution; (100.3010) Image reconstruction techniques
1. Introduction Structured illumination microscopy, SIM, is an optical superresolution technique, which excels in live cell imaging due to its low phototoxicity and fast acquisition rates. Superresolution is achieved by illuminating a fluorescent sample with a patterned excitation light field. The resulting beat pattern of illumination and sample encodes fine, usually lost, sample details. This frequency mixing thus opens the possibility to increase resolution and provide optical sectioning. In conventional SIM imaging, this procedure requires between 9 and 15 raw frames for a single superresolution image [1,2]. A reconstruction from less than 9 raw frames is hence an elegant way to increase SIM framerate without the need for hardware adaptions. Here, we present both the theory and implementation of a novel reconstruction approach, which uses a joint Richardson-Lucy algorithm [3] to lower acquisition times far below the conventional performance by greatly reducing the amount of required raw data for image reconstruction, while still being able to resolve features below the diffraction limit. 2. Methods In particular, we treat the Fourier spectrum of the raw acquisitions as composites of many individual segments. Incorporating information from all raw frames simultaneously thus enables reconstructions without the need for redundant frequency information as it was required in previous algorithms. This gives rise to a mixing matrix that links 66 unknown ‘superresolution’ spectral segments to 56 differently weighted and overlaid segment triplets from the raw data spectra [4]. A constrained iterative inversion of this underdetermined equation system then yields a SIM reconstruction (3SIM) of quality similar to conventional SIM reconstructions (9SIM). In our experimental validations we reconstructed 9SIM data with the same algorithm, which solves the minimization problem:
(1) Here, io are the acquired raw images and eo are excitation patterns for orientations o. The PSF is h and the SIM image estimates are is. 3. Results and Discussion We validated the procedure experimentally (see Figure 2) and benchmarked the achievable performance for fast imaging in silico to compare our approach to the conventional SIM technique using 9 raw frames.
Figure 1. Fluorescent beads were image on a home-built SIM microscope [5]. Nine raw frames were reconstructed using the jRL-SIM algorithm into a conventional SIM image (9SIM). A reconstruction from just three raw images (3SIM) with the same algorithm shows a similar resolution increase beyond widefield or deconvolution microscopy (using Richardson-Lucy deconvolution).
FW6D.4.pdf
Frontiers in Optics 2017 © OSA 2017
Furthermore, we link the number of raw frames to the maximally recordable sample velocity and the achievable reconstruction quality. Our results show that there exists a limit above which deconvolution microscopy becomes superior to SIM in both acquisition speed and resolution or image quality respectively (see Figure 2).
Figure 2. A double slit target was simulated to move at increasing speed v. The 180nm wide slit is initially resolved by both SIM modalities and just about by deconvolution microscopy (according to the Sparrow criterion). Artefacts in the reconstruction cap 9SIM quality at a target velocity of about 1.3 nm/ms and 3SIM at 3.0 nm/ms. At those velocities a fit to the slit contrast was found to be zero. Deconvolution microscopy is hence superior to SIM when studying highly dynamic samples of greater movement speeds. Increases in photon fluxes and faster raw frame rates can shift the found numbers absolutely but are still inherently limited by fluorophore photon budgets, phototoxicity, and hardware restrictions.
This is due to motion artefacts, which compromise the image quality of SIM reconstructions severely, but are far less pronounced in deconvolved widefield imaging. 4. Conclusion Reducing raw frames is possible in SIM via joint iterative deconvolution. This procedure opens the possibility for faster acquisition speeds or potential decreases of phototoxic effects at lower acquisition speeds. In future work an extension to 3D SIM might be possible. For studies of ultrafast dynamics the resolution gain of SIM is completely overshadowed by motion artefacts and reconstruction are not a trustworthy representation of the true sample structure anymore. In this case the method of choice for image acquisition is widefield microscopy with successive deconvolution or an analog SIM implementation [6] despite its greater phototoxicity. 5. References [1] Gustafsson, M. “Surpassing the lateral resolution limit by a factor of two using structured illumination microscopy”. J. Microsc. 198, 82–87 (2000)). [2] Gustafsson, M, et al. "Three-dimensional resolution doubling in wide-field fluorescence microscopy by structured illumination." Biophys. J. 94.12 (2008): 4957-4970. [3] Ströhl, F. & Kaminski, C. F. “A joint Richardson-Lucy deconvolution algorithm for the reconstruction of multifocal structured illumination microscopy data”. Methods Appl. Fluoresc. 3 (2015) [4] Ströhl, F. & Kaminski, C. F. “Speed limits of structured illumination microscopy”. Opt. Lett. (2017) [5] Young, L. J., Ströhl, F. and Kaminski, C. F. “A guide to structured illumination TIRF microscopy at high speed with multiple colors.” J. Vis. Exp. 111 (2016) [6] York, A., et al. "Instant super-resolution imaging in live cells and embryos via analog image processing." Nat. methods 10.11 (2013): 11221126.
