🔬 physics

Quantitative dynamic microscopy through scattering media without image reconstruction: a tool for biomedical optics

This paper presents a novel quantitative dynamic microscopy technique that measures the motion of microscopic objects hidden behind scattering media by exploiting temporal correlations of spatial Fourier modes, thereby enabling high-resolution, calibration-free biomedical imaging without the need for image reconstruction or coherent light sources.

Original authors: Fabio Giavazzi, Matteo Brizioli, Roberto Cerbino

Published 2026-09-07

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Fabio Giavazzi, Matteo Brizioli, Roberto Cerbino

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

In the world of optics, seeing clearly through a foggy window or a thick cloud of smoke is a familiar frustration. When light hits a chaotic, cloudy material, it bounces around wildly, scrambling the image of anything hidden behind it. For decades, scientists have tried to fix this by building complex mathematical models to reconstruct the original picture from the scrambled mess, a process that often requires powerful lasers, precise calibration, or extensive computer training. However, many real-world biological scenes, like blood flowing inside a living body, are too fast and too messy for these reconstruction methods to work, especially when using the simple, incoherent light found in standard microscopes. The challenge has been to find a way to measure how tiny things move inside or behind such cloudy layers without ever trying to see a clear picture of them in the first place.

A team of researchers has now demonstrated that you do not need to reconstruct an image to understand the motion of objects hidden behind a scattering layer. Instead of trying to unscramble the visual noise to reveal a face or a cell, they developed a method to listen to the rhythm of the light itself. By analyzing how the patterns of light change over time in a mathematical space called the Fourier domain, they can extract precise information about how particles are moving, even when the image on the camera looks like a completely featureless blur. This approach, which works with standard bright-field microscopes and ordinary light bulbs, allows scientists to measure the speed of blood flow and the jittery motion of microscopic particles hidden behind thick, cloudy barriers, achieving a level of sensitivity that was previously thought impossible without image reconstruction.

The researchers tested their idea using a simple setup: a glass tube filled with tiny plastic beads suspended in water, placed behind a layer of Parafilm, a common material used to mimic the cloudy nature of biological tissue. When they looked at this setup through a microscope, the image was completely scrambled; the individual beads were invisible, and the view appeared as a static, grainy noise. In a traditional imaging scenario, this would be the end of the road. However, the team applied a technique called differential dynamic microscopy, which does not look at the image as a picture but rather as a collection of light waves. They tracked how the intensity of these waves fluctuated over time. Even though the image was destroyed by the cloudy layer, the timing of these fluctuations remained linked to the actual movement of the beads. By measuring these temporal correlations, they could calculate exactly how far the beads moved over time, recovering their diffusion speed with nanometer precision, despite the signal being buried deep below the noise floor.

To ensure this method worked beyond the limits of simple, thin layers, the team developed a new statistical framework that accounts for how thick, cloudy materials mix different parts of the light signal. They discovered that while the cloudy layer scrambles the image, it does not erase the information about motion; it simply redistributes that information across different scales. By mathematically separating the effects of the scattering layer from the actual motion of the particles, they created a self-correcting process that works without needing to know the specific properties of the cloudy material beforehand. They validated this by testing layers of Parafilm ranging from 125 to 625 micrometers thick, as well as other materials like chicken breast and tracing paper. In every case, the method successfully recovered the true movement of the hidden particles, even when the signal was more than a hundred times weaker than the background noise.

The power of this technique was further demonstrated by measuring the flow of whole blood through a tiny channel hidden behind a scattering layer. Blood cells moving in a stream create a different kind of motion than the random jitter of the plastic beads, and the researchers adapted their method to track this directed flow. They were able to measure the speed of the blood cells as they moved at velocities typical of human capillaries, ranging from 2 to 25 micrometers per second. The results matched perfectly with measurements taken when the scattering layer was removed, proving that the method could accurately quantify the speed of a fluid flowing through a vessel that was completely invisible to the naked eye. This success suggests that the technique could be immediately useful in biomedical laboratories for studying microcirculation and other dynamic processes in tissues where direct imaging is impossible.

The significance of this work lies in its simplicity and robustness. Unlike other advanced techniques that require coherent lasers, guide stars, or complex training data, this method works with the incoherent light found in almost any standard microscope. It does not require the researcher to calibrate the cloudy medium or to know its optical properties in advance. The researchers showed that the degradation of image quality caused by multiple scattering does not necessarily mean a loss of dynamical information. By shifting the focus from reconstructing a picture to analyzing the statistical correlations of light over time, they have opened a new path for observing the microscopic world hidden behind the opaque barriers of living tissue. This approach offers a practical tool for biomedical optics, allowing scientists to measure the speed and movement of objects in environments that were previously considered too cloudy to study.

Technical Summary: Quantitative Dynamic Microscopy Through Scattering Media Without Image Reconstruction

Problem Statement
Quantifying the motion of microscopic objects hidden within or behind turbid (scattering) media remains a significant challenge in biomedical optics. Conventional imaging relies on a one-to-one correspondence between object points and image points, which breaks down when multiple scattering mixes light from different sample regions, destroying image contrast. While existing approaches often attempt to reconstruct images using transmission matrices, wavefront shaping, or deep learning, these methods face critical limitations: they typically require coherent illumination, prior calibration of the medium, or extensive training data. Furthermore, they often fail under low-coherence illumination (standard in biomedical microscopes) or when dealing with rapidly moving objects where image averaging or iterative feedback becomes impractical. Specifically, the quantitative characterization of fast microscopic dynamics (e.g., Brownian motion or blood flow) through thick, static turbid layers has remained largely inaccessible because the sample-induced signal is often buried below the noise floor.

Methodology
The authors propose a method that bypasses image reconstruction entirely, instead recovering sample dynamics directly from the temporal correlations of spatial Fourier modes in microscopy movies. The approach is built upon Differential Dynamic Microscopy (DDM), a technique that extracts the intermediate scattering function (ISF), $f(q, \tau)$ , from the temporal correlations of Fourier modes.

The methodology proceeds in two stages:

Memory Effect Regime: The authors first demonstrate that within the angular memory effect regime (where the scattering layer preserves the angular correlation of the incident field), the temporal correlations of Fourier modes are preserved even if the image itself is severely scrambled. In this regime, standard DDM analysis of the image structure function, $D(q, \tau)$ , can recover the correct sample dynamics despite the signal being visually obscured.
Generalized Framework Beyond the Memory Effect: To extend the method beyond the limited angular memory effect range, the authors develop a statistical model of the transmission kernel, $T(x_1, x_2)$ $T (x_{1}, x_{2})$ . They model the kernel as a random speckle pattern modulated by a bell-shaped envelope (the "shower-curtain effect") and a fluctuating component (the "memory effect").
- They derive a generalized relation (Eq. 10) linking the measured image structure function $D(q, \tau)$ to the true intermediate scattering function $f(q, \tau)$ . This relation accounts for two distinct contributions: a "shower-curtain" term (direct proportionality at low wavevectors) and a "memory effect" term involving a convolution that couples different Fourier modes at intermediate wavevectors.
- A self-consistent, calibration-free inversion procedure is developed to extract the sample dynamics (e.g., Mean Squared Displacement, MSD) from the measured data. This procedure iteratively optimizes a parameter ( $q_a$ ) representing the memory effect range to minimize the dispersion among MSD curves derived from different wavevectors, effectively correcting for mode mixing without prior knowledge of the scattering medium's optical properties.

Key Contributions

Reconstruction-Free Dynamics: The paper establishes that quantitative dynamic information can be recovered from turbid media without reconstructing an image, provided one analyzes temporal correlations in Fourier space.
Generalized Statistical Framework: The authors derive a generalized expression for the image structure function that remains valid beyond the conventional memory-effect regime, explicitly describing how multiple scattering redistributes dynamical information among neighboring Fourier modes rather than erasing it.
Incoherent Illumination Compatibility: The method operates effectively under spatially and temporally incoherent light (e.g., standard halogen lamps), making it immediately deployable in standard biomedical laboratories without the need for coherent sources or complex optical setups.
Calibration-Free Operation: The inversion procedure requires no prior calibration of the scattering medium's transmission matrix or optical properties.

Results
The method was validated through two primary experimental scenarios:

Brownian Dynamics of Colloids: The authors measured the diffusion of submicron polystyrene particles hidden behind scattering layers of varying thickness (up to 625 µm of Parafilm) and different materials (ground glass, tracing paper, chicken breast).
- They successfully recovered the MSD of the particles, achieving nanometer displacement sensitivity and millisecond temporal resolution.
- The measured diffusion coefficients ( $D_s \approx 0.42 \pm 0.02 \, \mu m^2/s$ ) agreed excellently with the Stokes-Einstein prediction and single-particle tracking results obtained without the scattering layer.
- The method remained robust even when the signal amplitude was more than two orders of magnitude below the noise level (effective SNR $\approx 10^{-2}$ ).
Whole Blood Velocimetry: The framework was applied to measure the flow velocity of whole blood in a microfluidic channel hidden behind scattering layers.
- The method accurately recovered flow velocities in the range of $2-25 \, \mu m/s$ , typical of microcirculation.
- The measured velocities matched the imposed flow rates with high accuracy, demonstrating the technique's applicability to directed flow (ballistic motion) in addition to diffusive motion.

Significance and Claims
The paper claims that this work provides a practical framework for recovering microscopic dynamics in scenarios where conventional image-based analysis is impractical. By exploiting the robustness of Fourier-space temporal correlations to optical scrambling, the method offers a complementary approach to existing techniques like wavefront shaping (which requires coherence and calibration) or deep learning (which requires training data).

The authors emphasize that their approach is immediately accessible to standard biomedical optics laboratories as it requires no coherent source, no guide stars, and no prior calibration of the medium. They position the work as a tool for studying optically inaccessible microcirculation and other biological dynamics hidden behind static turbid layers. The paper modestly notes limitations, specifically that the current framework assumes the scattering medium is static on the timescale of the sample dynamics; if the medium itself fluctuates rapidly (e.g., highly perfused living tissue), the factorization of the transmission kernel breaks down, requiring future extensions.

More like this