Text from the interactive explanation
SPIFFI enables single-shot super-resolution and multidimensional imaging
Wei Guo, Lely Feletti, Aleksandra Radenovic
Original paperGlowing tag molecules shine like tiny antennas: their light is polarized along each molecule's axis. SPIFFI splits a single camera exposure through four polarizers at once, and reads the disagreement between those views the way other methods read flicker collected over hundreds of frames. The result: about 1.7× finer detail from one shot.
Wei Guo, Lely Feletti & Aleksandra Radenovic, “SPIFFI enables single-shot super-resolution and multidimensional imaging”, Nature Methods 23, 2077–2086 (2026).
1 exposure → 4 polarization channels → 20 fluctuation images → 1 sharper image
Simulated sample computed live in your browser. The four channels are recorded at the same instant; the sweeping view is calculated exactly from them, so you can watch which filaments each angle favours.
Molecules glow like antennas
A fluorophore, the dye molecule that makes a structure glow, doesn't shine equally in every direction. It emits like a radio antenna, a dipole : its light vibrates mostly along the molecule's own axis. Put a polarizer (a filter that passes one vibration direction) in front of the camera, and the molecule looks bright when the two line up and faint when they cross.
That reshapes the blur spot too. The textbook spot, the point spread function or smallest blur a lens can make of one point, assumes a source that glows the same way in all directions. Seen through a polarizer, a real dipole's spot is usually an ellipse with a slimmer waist . Turn and tilt the molecule to see the four channels SPIFFI records.
Simulated Vectorial dipole model of an ideal 1.49-NA oil objective at 600 nm emission, after the imaging model in the paper's Supplementary Note 2. The slimmer polarized waist is one of three sources of SPIFFI's gain; with the complete method, real 100-nm beads narrowed from 294 nm to 173 nm (measured, Fig. 1c), even though freely rotating bead dyes give little polarization contrast.
Two molecules, one blur
Place two molecules 160 nanometres apart, closer than light can separate. The camera sees one smudge. Take one shot and follow what SPIFFI does with it:
Split. The single exposure lands on four quadrants of one camera, behind analyzers at 0°, 45°, 90° and 135°.
Resample. Each channel is split checkerboard-fashion into two interleaved pixel grids and filled back in two directions: five images per channel, 20 in all.
Correlate. For every pixel, measure how strongly its brightness swings across the 20 images.
Read. Where one molecule dominates, the swing is strong. Between two molecules pointing different ways, their swings oppose and cancel, so a gap opens.
Then experiment: turn the arrows so both molecules point the same way, shrink the gap, add wobble, dim the light, or switch the polarizers off. Switched off, every step of the maths stays the same but is fed four identical unpolarized copies: the swings vanish, and so does the gap.
Simplified model In this in-browser model the split is strongest when neighbours point different ways, the polarization-modulated fluctuation the paper describes. The paper's full pipeline (third-order correlation, calibrated deconvolution, background subtraction) resolved about 140 nm even for parallel dipoles in its own simulations, and 160–180 nm in noisy data.
Measured DNA nanorulers carry two dye spots 160 nm apart. From single frames SPIFFI measured 163 nm (s.d. 15, 20 rulers); widefield shows single blobs. Fig. 1d, Guo, Feletti & Radenovic, Nat. Methods 2026, CC BY 4.0; cropped.
No waiting, no smear
Fluctuation methods such as SOFI (super-resolution optical fluctuation imaging) wait for dye molecules to blink, then compute statistics over a window of frames, often 20 to 100 or more. If the cell moves during that window, the statistics mix old and new positions. SPIFFI takes its fluctuations from polarization inside each frame, so every frame is reconstructed on its own.
Simulated Blinking dyes on two filaments sliding across each other, 50-ms frames. The middle panel is a generic second-order fluctuation statistic over the chosen window; set the speed to zero and it sharpens too.
Live COS-7 cell, mitochondria (MitoTracker Red), depth colour-coded. Supplementary Video 4, Guo, Feletti & Radenovic, Nat. Methods 2026, CC BY 4.0; trimmed and re-encoded.
Measured
In live COS-7 cells, SPIFFI followed two microtubules frame by frame as the gap between them shrank from 241 to 198 to 177 nm and they overlapped between 4.6 and 4.7 s, while SACD (20-frame window) and eSRRF (100-frame window) turned the event into a time-averaged blur (Fig. 4i).
As the mitochondrial dye bleached (time constant 53 s), SPIFFI held 178 nm resolution while widefield sat at 479 nm (Fig. 4d). Volumes of five planes revealed a mitochondrial "touch" at a different depth seconds before the two fused (Fig. 4f).
What the real microscope saw
The instrument is a custom widefield microscope: a 100× oil objective, then a 50:50 beam splitter, a half-wave plate that turns one arm by 45°, and two polarizing beam splitters that throw four polarization channels onto one camera. Everything in this section is measured data from the paper.
Showing the 0° channel
One patch of immunolabelled microtubules (fixed COS-7 cell) through the four analyzers, each averaged over 100 frames. Filaments brighten and fade as the analyzer turns: the contrast SPIFFI feeds on. Fig. 2b, Guo, Feletti & Radenovic, Nat. Methods 2026, CC BY 4.0; cropped, arrows redrawn.
A widefield time series and its four polarization channels, recorded in the same exposures. Supplementary Video 1, Guo, Feletti & Radenovic, Nat. Methods 2026, CC BY 4.0; trimmed and re-encoded.
Same cell, six reconstructions
Immunolabelled microtubules, fixed COS-7 cell. Resolution by decorrelation analysis, as reported in Fig. 2c, Guo, Feletti & Radenovic, Nat. Methods 2026, CC BY 4.0; panels cropped, labels redrawn.
Every frame reconstructed on its own: widefield wipes to SPIFFI. Supplementary Video 1, CC BY 4.0; cropped, trimmed and re-encoded.
Across fixed cells, SPIFFI from one frame matched the detail of methods that needed 100 frames (Fourier ring correlation: 166 nm for SPIFFI versus 161 nm for SACD and 153 nm for eSRRF, each from 100 frames), and against a commercial SIM microscope it resolved filaments 131 nm apart. Hardware changes sit entirely in the emission path, so two colours work too: microtubules sharpened from 319 to 155 nm and mitochondria from 328 to 183 nm in the same cell.
Down to 80 nm
Each SPIFFI frame is already a super-resolved image, so a time series of them can be handed to ordinary temporal SOFI. The two stages stack: fluctuations across polarization first, fluctuations across time second.
On DNA nanorulers with three marks 80 nm apart, the two-stage result matched single-molecule localization microscopy (SMLM), the slow gold standard, from a third as many frames. It reached the equivalent of ninth-order SOFI from 1,000 frames, the number normally used for second-order SOFI, while avoiding the artifacts that high orders usually bring.
Long recordings need drift correction: about 50 nm of stage drift over 1,000 frames is enough to wash out 80-nm detail.
80-nm DNA nanorulers (three Cy3B marks). Fig. 3b,d, Guo, Feletti & Radenovic, Nat. Methods 2026, CC BY 4.0; cropped.
Fixed microtubules (CF568). Two neighbouring filaments measured 83 nm apart by the two-stage method and 88 nm by SMLM; sixth-order bSOFI and SACD failed even with 10,000 frames, and eSRRF reached 181 nm. Fig. 3f,g, CC BY 4.0; cropped.
Which way they point
The four channels hold more than sharpness. Two simple contrasts between them, the Stokes parameters, give each pixel's average dye angle and its degree of linear polarization (DoLP): near 1 when the dyes are held rigidly in one direction, near 0 when they wobble or point every which way. Add depth scanning and time, and each recording carries six dimensions: x, y, z, t, orientation and anisotropy.
Simulated Dyes along a bending filament, four channels computed with the same dipole model, then the Stokes formulas above, pixel by pixel. Bending here spreads the dye angles, as the paper suggests happens when dyes pack tightly on a curved filament.
Measured DoLP map, immunolabelled microtubules: mostly low, isotropic signal. Fig. 5a, Guo, Feletti & Radenovic, Nat. Methods 2026, CC BY 4.0; cropped.
Measured
With antibodies, high-DoLP regions reached 0.59, 0.51 and 0.38 with dyes about 51–61° from the filament; elsewhere DoLP fell to 0.11–0.13 with widely scattered angles. A live-cell probe on a short, rigid linker held its dyes nearly perpendicular to the filament (about 83° versus 53° for antibodies). As one live microtubule relaxed from curved to straight, its DoLP rose 0.30 → 0.54 → 0.85.
The paper
Wei Guo, Lely Feletti & Aleksandra Radenovic. “SPIFFI enables single-shot super-resolution and multidimensional imaging.” Nature Methods 23, 2077–2086 (2026). doi.org/10.1038/s41592-026-03196-6
Measured material. All figure panels and video clips on this page come from the article and its Supplementary Videos 1 and 4, © The Authors, licensed under CC BY 4.0. They have been cropped, trimmed, re-encoded and relabelled here; the numbers quoted are those reported in the paper.
Simulations. The living network, the antenna explorer, the two-molecule lab, the moving-filament comparison and the orientation map are simplified teaching models implemented for this page from the paper's description and its imaging model: a vectorial dipole point spread function (600 nm, NA 1.49), four analyzers, checkerboard resampling into 20 images, Richardson–Lucy deconvolution and a second-order correlation. Their numbers illustrate the mechanism; they are not the authors' results.
Code and data. The authors' SPIFFI implementation and demonstration data are available at github.com/wg9661/SPIFFI.
Laboratory. Laboratory of Nanoscale Biology, Institute of Bioengineering, École Polytechnique Fédérale de Lausanne (EPFL), Switzerland.
Topics in this iPaper
Physics
a multichannel polarimetric method for single-shot super-resolution imaging and six-dimensional information extraction
Abstract, page 1 (p. 2077)
Biology
We demonstrate the versatility of SPIFFI through experiments on both fixed and live cells, capturing rapid subcellular dynamics
Abstract, page 1 (p. 2077)
Engineering
The custom-built polarization microscope was configured as follows
Methods, Imaging setup, page 10
super resolution microscopy
SPIFFI achieves instant resolution enhancement from a single exposure
Abstract, page 1 (p. 2077)
super resolution optical fluctuation imaging
Each frame was first processed with third-order SPIFFI to obtain a primary super-resolution image, after which third-order SOFI-AC was applied for a secondary reconstruction.
Results, Dual-stage reconstruction, page 5 (p. 2081)
polarization microscopy
Subsequently, both optical paths were further separated by polarizing beam splitters (PBS251, Thorlabs), yielding four distinct polarization channels in total.
Methods, Imaging setup, page 10
live cell imaging
we imaged mitochondria in a live COS-7 cell labeled with MitoTracker Red to highlight the advantages of fast three-dimensional super-resolution imaging
Results, SPIFFI reveals the fast dynamics in live cells, page 6 (p. 2082)
fluorescence anisotropy
the intrinsic polarimetric design of SPIFFI also enables direct readout of fluorescent dipole orientation and anisotropy
Introduction, page 2 (p. 2078)