Mathematical modeling of fluorescence diffuse optical imaging of cell\n membrane potential changes

Habib Ammari, Josselin Garnier, Laure Giovangigli

TL;DR

This paper establishes a rigorous mathematical framework for imaging changes in cell membrane potential using fluorescence diffuse optical tomography (FDOT). By coupling light diffusion models with electrical cell physics, the authors derive analytical formulas for the "resolving power" of this technique, providing a roadmap for tracking cellular dynamics at microscopic resolutions even in noisy environments.

What Problem Does This Paper Address?

Understanding the electrical potential across a cell membrane is fundamental to biology—it’s how neurons fire and how cells signal one another. While optical methods using fluorescent dyes (fluorophores) have been used to measure these potentials since the 1970s, doing so deep within biological tissue is challenging. Biological tissue is highly scattering, meaning light doesn't travel in straight lines, making it difficult to reconstruct a clear image of the cell membrane.

The authors tackle the "inverse problem": how can we take measurements of light at the boundary of a tissue and work backward to create a high-resolution map of the electrical potential changes on a specific cell membrane? Specifically, they aim to define the mathematical limits of this imaging—how small of a change can we actually see before measurement noise washes it out?

Methods

The paper builds a multi-layered mathematical model:

  1. Coupled Diffusion Equations: The authors use two diffusion equations to model light. One describes the "excitation" light entering the tissue, and the second describes the "emitted" fluorescent light coming from the cell membrane.
  2. Electrical Modeling: They incorporate a model for the voltage potential ($u$) within the medium. Crucially, they assume the fluorophore concentration on the membrane responds linearly to the "potential jump" (the difference in voltage) across that membrane.
  3. Layer Potential Techniques: To solve these complex boundary value problems, the authors employ layer potentials—a sophisticated tool in partial differential equations that allows them to represent solutions as integrals over the cell boundary.
  4. Linearized Inverse Problem: To reconstruct the cell's shape, they assume the cell is a small "perturbation" of a known shape (like a disk in 2D or a sphere in 3D). This allows them to use Fourier analysis to estimate the shape deformation ($h$) from boundary light data.
  5. Stability and Resolution Analysis: They introduce a Signal-to-Noise Ratio (SNR) framework to derive the "resolving power," determining which "modes" (frequencies of shape change) can be reliably recovered.

Datasets

Not discussed in the paper. This is a theoretical and mathematical modeling paper. While it uses "realistic physical values" for parameters like absorption and scattering coefficients (sourced from previous experimental literature), it does not utilize or introduce a specific experimental dataset.

Key Insights and Ideas

  • The Fluorophore as a Voltmeter: The core "aha" moment is the formal coupling of the electrical potential jump to the fluorophore concentration. This turns the cell membrane into a localized, light-emitting sensor of its own electrical state.
  • The Resolution-Noise Trade-off: The paper quantifies exactly how measurement noise limits spatial resolution. It shows that while low-frequency changes in membrane shape are easy to see, high-frequency (fine-grained) details require exponentially higher SNR to resolve.
  • Minimal Resolving Radius ($R^*$): The authors define a threshold for the smallest feature or cell size that can be imaged given a specific noise level. This provides a practical limit for engineers designing these imaging systems.
  • Frequency Sensitivity: The choice of the modulation frequency ($\omega$) for the light source is critical. The model shows how the physics of light diffusion interacts with the timing of fluorescence to dictate the final image quality.

Key Results

  • Analytical Resolving Power: The authors successfully derived explicit formulas (Theorems 4.13 and 4.14) that predict the resolving power of FDOT based on SNR and physical tissue parameters.
  • Stability Conditions: They demonstrated that for a circular cell, the reconstruction of the membrane perturbation is unbiased, and they provided a precise formula for its variance in the presence of Gaussian noise.
  • Generalization: While much of the paper focuses on perturbations of a disk, they provided a generalized framework (Proposition 4.15) for arbitrary cell shapes in two dimensions.

Limitations and Future Work

  • Small Perturbation Assumption: The current direct imaging algorithm relies on the cell being a relatively simple perturbation of a disk or sphere. Reconstructing highly complex or "cluttered" geometries would likely require the iterative algorithms the authors suggest but do not fully develop here.
  • Homogeneity: The model assumes the background tissue is largely homogeneous. Real-world biological environments are much "messier," which could introduce additional artifacts not accounted for in this pure mathematical model.
  • Future Directions: The authors highlight the potential for using this model to track dynamic processes like cell division or the behavior of "cluttered" live cell populations in real-time.

Summary updated

Abstract

The aim of this paper is to provide a mathematical model for spatial\ndistribution of membrane electrical potential changes by fluorescence diffuse\noptical tomography. We derive the resolving power of the imaging method in the\npresence of measurement noise. The proposed mathematical model can be used for\ncell membrane tracking with the resolution of the optical microscope.\n
BibTeX
@inproceedings{ammari2013mathematical,
  author = {Habib Ammari and Josselin Garnier and Laure Giovangigli},
  title = {{Mathematical modeling of fluorescence diffuse optical imaging of cell\textbackslash\{\}n membrane potential changes}},
  booktitle = {Quarterly},
  year = {2013},
  doi = {10.1090/s0033-569x-2013-01334-x},
}
Figure