Flowing cells get organelle-level eyes: three structured-light frames plus self-supervised reconstruction reach about 118 nm at up to roughly 720 cells per second

Cells in continuous flow can now be imaged one by one at organelle scale, not merely measured for overall size and shape. Conventional imaging flow cytometry is diffraction-limited, leaving sub-200 nm organelle structures unresolved; among the few attempts to bring super-resolution into flow, image-scanning and structured-light-sheet approaches doubled resolution but ran at about one to ten cells per second, two orders of magnitude below flow-cytometric throughput. On 100 nm fluorescent beads, SR-IFC full widths at half maximum were about 155 nm along the flow axis and 149 nm across it, corresponding to about 118 nm and 110 nm after correcting for bead diameter, while wide-field imaging of the same beads gave 222 to 227 nm, an approximately twofold gain.

The resolution gain comes from only three single-orientation structured-illumination frames, with transverse detail inferred by a network from the measured cells themselves. Conventional structured-illumination microscopy needs multiple grating orientations, usually nine frames in total, to reach isotropic in-plane resolution, which cannot fit the millisecond window in which a cell crosses the field of view. Across three structurally distinct simulated phantoms, SR-IFC's pixel error, structural similarity and Fourier amplitude error closely followed nine-frame structured illumination; on task-based resolution it matched structured illumination for the bead-pair and ellipsoid phantoms, slightly exceeded it for the frequency phantom (0.75 versus 0.72), and stayed close to the structured-illumination reference as peak relative shot noise rose from 0% to 50%.

Frame-to-frame displacement under flow is estimated from the data and corrected, so organelle imaging in continuous flow does not depend on precise flow-speed control. On a motion stage, bead displacement between frames was nearly identical for every bead; in continuous flow each cell appears at a different, unknown position in each of the three frames, and the displacement varies from cell to cell, whereas structured-illumination reconstruction requires aligning the three frames as phase-shifted views of the same object. For Jurkat cells stained with MitoTracker Green and flowed at 25 mm/s, cross-correlation registration produced artifact-free reconstructions, and SR-IFC resolved mitochondrial fragments and tubular structures that were blurred or merged in wide-field and flow-axis structured-illumination images.

The demonstrated capability is limited to a single-color fluorescence channel and two-dimensional in-plane imaging, with physical structured illumination applied only along the flow axis; throughput is coupled to the camera readout region, creating a trade-off between imaging area and acquisition speed, and the method requires accurate registration of the three phase-shifted frames, which constrains the maximum usable flow velocity. The intended readers are researchers and engineers working on flow cytometry, super-resolution microscopy and single-cell phenotyping.

The resolution gain has so far been demonstrated only in a single color channel and in two dimensions; throughput is limited by camera readout height, so larger cells requiring taller readout regions are imaged at lower rates; the CD8+ T-cell morphology trajectory comes from a single healthy donor, the fused morphotype matches the central-memory fraction only weakly, and its relationship to central memory requires further validation; because the cells were stimulated repeatedly, part of the fragmented morphotype may reflect exhaustion-related remodeling, and exhaustion markers were not assessed.