Compressive Fourier-Domain Intensity Coupling (C-FOCUS) enables near-millimeter deep imaging in the intact mouse brain in vivo

Renzhi He, Yucheng Li, Brianna Urbina, Jiandi Wan, Yi Xue*,
University of California, Davis,
Paper arXiv Video Code Data

Abstract

Two-photon microscopy is a powerful tool for in vivo imaging, but its imaging depth is typically limited to a few hundred microns due to tissue scattering, even with existing scattering correction techniques. Moreover, most active scattering correction methods are restricted to small regions by the optical memory effect. Here, we introduce compressive Fourier-domain intensity coupling for scattering correction (C-FOCUS), an active scattering correction approach that integrates Fourier-domain intensity modulation with compressive sensing for two-photon microscopy. C-FOCUS enhances fluorescence intensity by over 20-fold in vivo, accelerates scattering correction by 5-fold compared to previous approaches, and enables multipatch correction across fields of view spanning hundreds of microns beyond the memory effect range. Using C-FOCUS, we demonstrate high-resolution imaging of YFP-labeled neurons and FITC-labeled blood vessels at depths exceeding 900 μm in the intact mouse brain in vivo. Furthermore, we achieve transcranial imaging of YFP-labeled dendritic structures through the intact adult mouse skull. C-FOCUS provides a broadly applicable strategy for rapid, deep-tissue optical imaging in the living brain.

Overview

Matting Example

C-FOCUS enables two-photon imaging of YFP-labeled pyramidal neurons up to 940 $\mu m$ below the dura in the somatosensory cortex using intensity-based scattering correction with compressive sensing.

a, Optical schematic of the C-FOCUS system, based on the 2P-FOCUS framework. b, an uncorrected image is first acquired and segmented into subregions, each containing a local intensity peak that serves as a correction target. c, Correction masks for the subregions are measured and calculated sequentially using compressive sensing. d, In vivo imaging of YFP-labeled pyramidal neurons in the somatosensory cortex of a Thy1-YFP-H mouse through a cranial window using 1035 nm excitation, with and without scattering correction. With C-FOCUS, fluorescence intensity is enhanced fivefold, enabling imaging to a depth of 800 μm (8.5 EAL).

Results

Fluorescent Beads through a Mouse Skull

Progressive Encoding for Neural Optimization

Quantitative evaluation of C-FOCUS by imaging red fluorescent beads through a 250 μm mouse skull ex vivo.

a, Correction masks computed using 10–5,000 measurements and the corresponding corrected images. The fully sampled case uses 10,000 measurements. b, Image of the same fluorescent bead without correction (magnified 38-fold for display). c, Image of the bead using a low numerical aperture (NA) mask with size comparable to the DC component of the correction mask. d, Image using a correction mask generated by 2P-FOCUS with 10,000 measurements. e, Overlap of correction masks from 2P-FOCUS (cyan, 10,000 measurements) and C-FOCUS (green, 3,000 measurements; magenta, 5,000 measurements). f, Comparison of peak fluorescence intensity without correction (cyan), with low NA (green), with 2P-FOCUS (purple), and with C-FOCUS (magenta) using correction masks from a. g, Optimal DMD output-to-input power ratio balancing resolution and correction effectiveness. h, Fluorescence intensity, i, image contrast, j, lateral ,and k, axial resolution at various power ratios. l, m, PSF cross-sections of the bead along the x- and z-axes, respectively.

Neurons

D-NeRF

C-FOCUS visualizes axons of pyramidal neurons up to 900 μm deep in layer 6 and white matter of the visual cortex in vivo.

a-d, Comparison of image volumes with and without correction. The left column shows representative images at depths from 30 to 620 μm without correction. e, Comparison of volumetric views of images without correction (left) and with correction (right). f-i, Side-by-side comparison of images without correction (first column), with correction (second column), zoomed-in views (third column; scale bar, 10 μm), and fluorescence intensity profiles along the dashed lines shown in the zoomed-in views (fourth colume), at \textbf{f.} 720 μm, \textbf{g.} 780 μm, \textbf{h.} 850 μm, and \textbf{i.} 900 μm depths. In the zoomed-in views, intensities of the uncorrected images are magnified (as indicated) for visualization purposes. j, Schematic of in vivo imaging of layer 5 pyramidal neurons expressing YFP through a cranial window. k, Subregions and corresponding correction masks at 900 μm depth, labeled by colored frames.

Blood Vessels

Blood Vessels

Adult wild-type mice were imaged through a cranial window under anesthesia following intravenous injection of FITC.

a, Volume rendering of FITC-labeled blood vessels after scattering correction. The left column shows maximum intensity projection (MIP) images from 0 to 600 μm depth without correction. The right column shows the volume view of the $252 \times 252 \times 910 \mu m^3$ image stack. b-d, Comparison between images without correction (left column), with correction (middle column), and zoomed-in views (right column, scale bar: 20 μm) at \textbf{b} 700 μm, \textbf{c} 820 μm, and \textbf{d} 910 μm depths. These images were acquired from different mice under the same labeling protocol. In the zoomed-in views, the intensity of the uncorrected images is digitally magnified (scaling factor indicated in each panel) for display purposes. e-g, Fluorescence intensity profiles without and with correction along the dashed lines in the zoomed-in views. h-j, Subregions and representative correction masks from the boxed regions in \textbf{b-d} (highlighted by colored frames). /p>

Neurons through Intact Skull

NR-NeRF

C-FOCUS visualizes the apical dendrites of YFP-labeled pyramidal neurons through 110 μm-thick intact skull in vivo.

a, The volume view of the skull (second-harmonic signal (SHG)) and YFP-labeled dendrites without and with correction. The left column shows the representative SHG image at 50 μm depth (1.1 EAL) and YFP images at 170 μm (3.2 EAL) and 220 μm (3.4 EAL) depth without correction. b-d, Comparison between the images without correct (left column) and the images with correction (middle column), and zoomed-in views (right column, scale bar, 10 μm) at \textbf{b} 430 μm (3.9 EAL), \textbf{c} 490 μm (4.2 EAL), and \textbf{d} 560 μm (4.5 EAL) depth. In the zoomed-in views, the intensity of images without correction are magnified (as labeled in the images) for display purpose. e-g, Fluorescent intensity along the dashed line in the zoomed-in views. h-j, Subregions and correction masks on the same $z$-plane as \textbf{b-d}. k, Schematic of in vivo transcranial imaging of YFP neurons through an intact skull. The coverglass is to protect the exposed skull after survive surgery. /p>

BibTeX



      @misc{he2025compressivefourierdomainintensitycoupling,
      title={Compressive Fourier-Domain Intensity Coupling (C-FOCUS) enables near-millimeter deep imaging in the intact mouse brain in vivo}, 
      author={Renzhi He and Yucheng Li and Brianna Urbina and Jiandi Wan and Yi Xue},
      year={2025},
      eprint={2505.21822},
      archivePrefix={arXiv},
      primaryClass={physics.optics},
      url={https://arxiv.org/abs/2505.21822},}