High-resolution scanning fluorescence imaging through scattering via speckle replica alignment and variance computation

Abstract

Fluorescence imaging is an essential diagnostic tool in many fields, but diffraction-limited optical imaging at depth is limited by scattering. Here, we present a method based on multiple random illuminations, combined with a computational framework that retrieves high-resolution images by aligning local speckle replicas and computing their pixel-wise variance. We demonstrate its versatility in two regimes: linear wide-field one-photon (1P) fluorescence imaging and nonlinear two-photon (2P) fluorescence imaging where the object is excited by a scanned speckle field and detected with a single-pixel detector. This approach outperforms standard autocorrelation techniques in terms of resolution and convergence.

Figures (6)

• Figure 1: Schematic of the experimental setup and reconstruction principle. a, When a highly scattering medium obstructs the direct view of an object, light from a point source in the object plane is scattered, creating a diffuse light field and a random speckle on the camera. This speckle pattern can be regarded as the system's PSF. Consequently, the captured speckle image of the object on the camera is the convolution of the object with the PSF. b–f, The principle of the proposed reconstruction strategy: A series of speckle images is first recorded and then preprocessed according to the specific experimental setup. After preprocessing, the speckle images are transformed into the input dataset (b). This input dataset is then divided into subimages (c), with the size of each subimage determined by the object's approximate dimensions. d, A subimage is selected as a reference, and cross-correlation alignment is performed between the reference and all other subimages. e, During the cross-correlation alignment process: first, the cross-correlation between the reference and an unaligned subimage is computed; second, the unaligned subimage is shifted based on the cross-correlation result; third, the aligned subimage is summed with the reference to update the reference image. f, The intermediate image is achieved by calculating the variance over the aligned data. f, Then, the final reconstruction is produced by performing a dual-deconvolution between this intermediate image and the input dataset. h, 1P fluorescence imaging scenario: A fluorescent object is illuminated with randomly modulated, unknown laser patterns, and a series of fluorescence speckle images is recorded by the camera as the illumination changes randomly between frames. The pixel-wise variance of the captured speckle images is computed and then used as the input dataset. Subsequently, the input dataset is divided into subimages. i, 2P fluorescence imaging scenario: A fluorescent object is excited by a pulsed laser with random, unknown modulation. A speckle image is captured using a photomultiplier tube (PMT) by scanning the speckle pattern across the sample while the rotating diffuser remains fixed during each scan. A series of speckle images is acquired, with the modulation varying between different frames. The captured speckle images act as input data in the 2P fluorescent case. They are then divided into subimages. SM: scattering medium.
• Figure 2: 1P fluorescence imaging through scattering media. a-f, Fluorescence speckle images measured at the camera. b-g, Input data, which is preprocessed data from (a-f). c-h, The intermediate image. d-i, The final reconstruction. e-j, Ground truth image taken from the side without the scattering media.
• Figure 3: 2P fluorescence imaging through scattering media. a-e, Fluorescence speckle images captured using PMT. (Inset) Incoherent sum of speckle images. b-f, The intermediate image. c-g, The final reconstruction. d-h, Ground truth images taken from the side without the scattering media.
• Figure 4: Comparison of resolution between incoherent summation and higher-order SOFI over aligned subimages.a, 2P fluorescence speckle images captured using a PMT. b–e, Incoherent sum, and second-, third-, and fourth-order SOFI images, respectively. f, Ground-truth images taken from the side without scattering media. g–h, Intensity profiles extracted along the dashed lines in b–e. g, Vertical profiles; h, Horizontal profiles. i Intensity profiles extracted along the green dashed lines in b–e. j, Legends for g–i.
• Figure 5: Experimental setups. (a) 1P Setup: A sequence of random phase patterns is displayed on a spatial light modulator (SLM), generating uncorrelated speckle illumination patterns at the sample plane (O). This is achieved through a beam splitter (BS) and a high-numerical-aperture objective (OB) with a NA of 0.85, which focuses the light onto fluorescent beads. The emitted fluorescence is transmitted through a scattering medium (SM) and collected by a low-NA OB (NA = 0.25). The signal is then imaged onto a fluorescence camera (Cam. fluo) via a tube lens (TL). The camera ground truth (Cam. GT) is used to capture the ground truth image from the side without SM. (b) 2P Setup: A femtosecond laser beam is modulated by a rotating diffuser (RD), introducing random phase variations. The modulated beam is directed through a galvanometric mirror (GM) unit for raster scanning and focused onto the back focal plane of an excitation OB (NA = 1.0), exciting the object located behind a SM. The emitted fluorescence is collected by a second OB (NA = 1.3), filtered, and detected by a photomultiplier tube (PMT) after passing through optical components including a dichroic mirror (DM), tube lens (TL), and filter (F). The Cam. GT is operated to capture the ground truth image from the side without SM. The beam shown in green in (a) (in yellow in (b)) corresponds to the illumination light, whereas the beam shown in lime yellow in (a) (in vivid orange in (b)) corresponds to the emission light. Abbreviations: SLM – spatial light modulator; BS – beam splitter; OB – objective; SM – scattering medium; CAM – camera; PMT – photomultiplier tube; GM – galvo mirror; SL – scanning lens; DM – dichroic mirror; TL – tube lens; F – filter; O – object; Cam. GT - camera ground truth.