Fig. 1.

ZOLA-3D optical setup and algorithm. a Sketch of the optical system, featuring the objective lens, the deformable mirror and the camera. The deformable mirror is placed in the Fourier plane of the emission light path; its shape is controlled by 40 actuators (see Supplementary Fig. 1). b–d PSF calibration from bead images. Scale bars, 1 μm. b A z-stack of one or more subdiffraction sized fluorescent beads is acquired. Here, 40 images are taken over an axial range of 4 μm, with Δz = 100 nm steps. Top row shows one image for every 1 μm step, the bottom row shows a (y, z) slice of the entire stack. c ZOLA uses these images to compute a maximum likelihood estimation of the phase (shown on the right, with color indicating phase) as a linear combination of Zernike polynomials (the twelve Zernike functions with the highest calculated coefficients are shown as small images, with their coefficients beneath). d PSF model computed by ZOLA from the retrieved phase, shown as in b. Unlike the 3D bead image, the PSF model is continuous, i.e., can be computed for any subpixelic position. e Fifty images of a fluorescent bead are taken at each of 40 z positions (with Δz = 100 nm, i.e., over 4 μm). Scale bar, 1 μm. f Localization precisions as a function of z. Blue, red and black dots are average experimental localization precisions, defined as standard deviations of computed coordinates x, y, and z, respectively. Solid curves are theoretical precision limits, assuming a mean photon number of 4677 and mean background of 18.2 as in the bead images. g A single molecule image sequence is processed by ZOLA (a single frame is shown; scale bar, 5 μm). Processing includes detection, localization, merging consecutive localizations, filtering, drift correction, and super-resolution image rendering. h 3D super-resolution image of the mitochondrial protein TOM22 in a Cos7 cell, with color indicating depth z. Scale bar, 5 μm. The right panel shows a (y, z) slice at the position indicated by the pink dashed line. Supplementary Movie 2 shows an animated 3D view