Introduction to blind-SIM
Structured Illumination Microscopy (SIM) is a successful super-resolution fluorescence microscopy method. For more information on the experimental method, please read the fast SIM page.
Data acquired on a SIM instrument require sophisticated processing to yield a final super-resolution image. The classical reconstruction algorithm relies on precise knowledge of the illumination pattern (phase, grating period…) and is prone to artefacts in the case of non-ideal experimental conditions. For instance, strongly scattering or thick samples can lead to a distorsion of the fringes.
Blind-SIM is a novel, deconvolution-based, approach which does not require a priori knowledge of the illumination pattern. It should therefore widen the range of applications of SIM imaging. So far, it has been demonstrated with both periodic and speckle illumination schemes on 2D simulated and experimental samples .
Recently, the performance of the blind-SIM algorithm has been improved by using partial a priori knowledge of the illumination pattern. This so-called filtered blind-SIM has enabled to reconstruct samples that were excited by strongly distorted light fringes, a situation where the classical reconstruction produces artefacts .
Further development includes using novel smart illumination schemes, and extending the method to 3D.
Thick slice blind-SIM
The blind-SIM algorithm as it is published in references  and  can only reconstruct 2D images. The novel implementation of blind-SIM, called thick slice blind-SIM, is able to remove the out-of-focus light contribution out of data acquired in a single plane . This approach further improves blind-SIM reconstructions in thick samples but where the information of interest lies in a single plane.
(a) Used simulated sample. A π/2-rotated version of the same object is placed 800 nm out-of-focus.
(b) 2D deconvolved widefield image where some out-of-focus blurred features are visible
(c) 3D deconvolved widefield image where the out-of-focus information is removed but where the resolution is low (diffraction limit).
(d) Simulated SIM raw image (first grating position) for a straight illumination grating.
(e) Processed SIM result with 2D blind-SIM: the resolution has been improved by a factor of 2 but the out-of-focus is still present.
(f) Processed SIM result with thick slice blind-SIM: the out-of-focus blur is rejected and the resolution doubled with respect to the WF deconvolved image.
(g) Distorted illumination grating used for a second simulation of SIM data.
(h) Reconstructed grating using thick slice blind-SIM: blind-SIM does not only retrieve the sample but also the illumination pattern.
(i) Processed SIM result with thick slice blind-SIM: the image quality is similar to the non-distorted case: thick slice blind-SIM can reject out-of-focus and reconstruct super-resolution details also in the case of distorted illumination.
Thick slice blind-SIM applied to experimental data
The sample is an MCF-7 actin-labelled cell which was prepared by Michael Reuter. The raw SIM data was acquired on the commercial Elyra-S1 microscope (Carl Zeiss) by Elena Tolstik.
(a) 2D widefield deconvolved image of a 200×200 pixels region of the sample.
(b) 2D blind-SIM reconstruction. Note the resolution enhancement indicated my the magenta arrow: two filaments appear merged in the WF images and are resolved in the SIM reconstructions.
(c) 3D WF deconvolution
(d) Thick slice blind-SIM reconstruction. Note that the out-of-focus information is rejected, for example the two filaments indicated by the pair of green arrows belonging to another focal plane are not visible.
(e) Comparison to the reconstruction using the commercial software in the focal plane. The filaments are also quite well rejected.
(f) Comparison to the reconstruction using the commercial software in the plane where the filaments appear in focus (200 nm distance along optical axis). This proves that the observation made in the thick slice blind-SIM is not an artefact.
The blind-SIM algorithm is included in a generic deconvolution toolbox. Several types of samples and modes of processing are implemented. This flexibility makes it a very powerful tool, however also increase its complexity. At the moment, the algorithm is not stable enough for the contributors to release it. Work is being done in this direction. However, before the distribution of a fixed version, we welcome colleagues to send us their SIM data for reconstruction. We hope for understanding that, even though such collaborations are warmly appreciated, these reconstructions cannot be given first priority.
 Mudry, E.; Belkebir, K.; Girard, J.; Savatier, J.; Le Moal, E.; Nicoletti, C.; Allain, M. & Sentenac, A.
Structured illumination microscopy using unknown speckle patterns
 Ayuk, R.; Giovannini, H.; Jost, A.; Mudry, E.; Girard, J.; Mangeat, T.; Sandeau, N.; Heintzmann, R.; Wicker, K.; Belkebir, K. & Sentenac, A.
Structured illumination fluorescence microscopy with distorted excitations using a filtered blind-SIM algorithm
Optics Letters, 2013, 38 (22), 4723-4726
 Jost, A.; Tolstik, E.; Feldmann, P.; Wicker, K.; Sentenac, A. & Heintzmann, R.
Optical Sectioning and High Resolution in Single-Slice Structured Illumination Microscopy byThick Slice Blind-SIM Reconstruction
PLoS ONE, 2015, 10(7)
PhD Student: Aurélie Jost