A preprint posted to arXiv reports that DeepFOCUS produced two-photon images of mouse brain structures as deep as 1,200 micrometers. In separate demonstrations, the system visualized hippocampal neurons to that depth and blood vessels to 1,100 micrometers.
How the correction works
DeepFOCUS is an intensity-based approach to scattering correction. It uses a digital micromirror device, or DMD, at a relayed Fourier plane in a point-scanning two-photon system, with 1,035-nanometer excitation at a repetition rate of 1 MHz.
Rather than relying on a pretrained model or a ground-truth correction mask, the method optimizes its correction-mask representation directly from target fluorescence measurements. The measured fluorescence therefore supplies the information used to shape the mask.
From a skull test to living mice
Before the in vivo tests, parameter sweeps used red fluorescent beads in PDMS behind a mouse skull about 250 micrometers thick. Each parameter was evaluated in an independent experiment.
In the measurement-count sweep, reported enhancement peaked at 42-fold with 10,000 measurements and exceeded 40-fold beyond 5,000 measurements. The study selected 5,000 measurements for later experiments.
The later settings were empirical defaults from the ex vivo optimization, rather than prespecified universal parameters. The study did not establish that those settings will transfer to other animals, labels, tissues or microscopes.
What the images reached
In one wild-type mouse, corrected imaging of FITC-labeled blood vessels reached 1,100 micrometers across a 252-by-252-micrometer viewing area. It resolved capillaries as small as 2.1 micrometers; representative enhancement ranged from 3.1-fold to 3.7-fold at the reported depths.
In the other animal, a male 11-week-old Thy1-YFP-H mouse, corrected imaging visualized neurons in the hippocampus’s CA1 region to 1,200 micrometers. It identified pyramidal neurons at 1,020 micrometers, resolved dendrites to 1,180 micrometers, and showed axons and dendrites about 1.7 to 2.2 micrometers across, with white-matter enhancement of up to 5.9-fold.
The deepest hippocampal example came with a time cost. At 1,180 micrometers, 13 correction masks covered a 126-by-126-micrometer field; measuring each mask took about 9 seconds and computing it about 6 seconds, for a reported total correction time of 191.1 seconds.
What the preprint does not settle
The in vivo evidence rests on one wild-type mouse for vascular imaging and one male 11-week-old Thy1-YFP-H mouse for neuronal imaging; the report leaves the wild-type mouse’s sex and age incomplete. No animal-level replication, confidence intervals or formal hypothesis tests were reported.
The study is an optical image-formation demonstration in mice, not evidence of human or clinical imaging, functional neural activity, behavior or therapeutic benefit. The experiments used cranial windows and anesthesia.
It also lacks a direct, same-condition comparison with three-photon microscopy or another scattering-correction method, so the preprint does not establish superiority over those approaches. The authors instead describe DeepFOCUS as complementary to, rather than a replacement for, three-photon microscopy.
For now, the report’s strongest result is a depth demonstration: structural fluorescence images reached 1.2 millimeters in a hippocampal mouse preparation. Whether the approach can be reproduced across more animals and adapted to awake or functional imaging remains open.
Paper data and sources
Original title: Intensity-based scattering correction enables in vivo two-photon imaging beyond 1 mm
Authors: Yucheng Li, Renzhi He, Yi Xue
Journal/Repository: arXiv
Status: Preprint, not yet peer-reviewed
First online: 2026-08-20
DOI: Not available
Original paper · Full text