Depth-dependent attenuation and backscattering characterization of optical coherence tomography by stationary iterative method

Yaning Wang, Shuwen Wei, Jin U. KangJournal of Biomedical Optics 2023 · DOI

TL;DR

This paper introduces a new iterative method to accurately measure how light scatters and is absorbed at different depths within biological tissues using Optical Coherence Tomography (OCT). By simultaneously characterizing both attenuation and backscattering, this approach overcomes limitations of previous methods and improves the clarity and diagnostic potential of OCT images, especially for complex tissues.

What Problem Does This Paper Address?

Optical Coherence Tomography (OCT) is a powerful imaging technique that provides high-resolution cross-sectional views of biological tissues. However, the contrast and clarity of OCT images are significantly affected by how light is attenuated (absorbed and scattered) as it penetrates deeper into the tissue. Accurately quantifying this attenuation is crucial for detailed analysis and disease diagnosis.

Previous methods for estimating tissue attenuation coefficients (μ) often relied on a simplifying assumption: that the backscattering fraction (R), which describes how much light is scattered back towards the detector, remains constant throughout the tissue or within specific layers. This assumption is often violated in real biological tissues, which can have complex, heterogeneous structures with varying optical properties. This leads to significant errors, either under- or overestimating the true attenuation, especially in regions with significant changes in backscattering. The challenge, therefore, is to develop a method that can accurately measure both attenuation and backscattering simultaneously and depth-dependently, without requiring prior knowledge of the tissue's detailed structure.

Methods

The core of this paper is a novel stationary iterative method designed to decouple and simultaneously estimate the depth-dependent attenuation coefficient ($\mu(z)$) and backscattering fraction ($R(z)$) from OCT signals.

Here's a breakdown of the approach:

  • Theoretical Foundation: The method starts with the fundamental OCT signal equation, which relates the detected intensity $I(z)$ to the initial irradiance $L(0)$, the attenuation coefficient $\mu(z)$, and the backscattering fraction $R(z)$. The key challenge is that these parameters are coupled.
  • Iterative Decoupling: The authors derive a set of nonlinear integral equations that relate $\mu(z)$ and $R(z)$. Since these equations are underdetermined and transcendental, an iterative approach is employed. The method iteratively refines estimates of $\mu(z)$ and $R(z)$ by using a stationary iteration scheme.
  • Handling Backscattering Variations: A crucial innovation is how they handle variations in $R(z)$. Instead of assuming $R(z)$ is constant, their algorithm accounts for its depth-dependent nature. To manage the complexity, they introduce a "fuzzy processing" step for $R(z)$. This means that only significant changes in $R(z)$ are used to update $\mu(z)$, while minor fluctuations (often attributed to noise) are averaged out. This prevents the algorithm from being overly sensitive to noise and helps preserve signal edges.
  • Constraint Conditions: To ensure stable convergence and prevent unrealistic results, several practical constraint conditions are applied during the iteration. These include bounds on the value of $R(z)$ (between 0 and 1) and criteria for stopping the iteration process (e.g., when the change in $\mu(z)$ between iterations falls below a threshold).
  • Attenuation Compensation: Once accurate depth-resolved $\mu(z)$ and $R(z)$ are obtained, they are used to develop an attenuation compensation model. This model reconstructs the OCT signal intensity as if there were no attenuation loss, thereby improving the visibility of deeper structures and enhancing image contrast.

Datasets

The method was validated using a combination of:

  • Numerical Simulations:
    • A-mode numerical imaging: Idealized digital phantoms were created to test the core algorithm under controlled conditions, avoiding complexities like multiple scattering or noise.
  • Monte Carlo simulation: A more realistic simulation of light transport through a complex, 13-layer retinal model was performed. This allowed for evaluation in a heterogeneous, multi-layered structure.
  • Experimental Data:
    • Multi-layer silicone-dye-TiO2 phantoms: These fabricated phantoms allowed for controlled variations in optical properties and served as a testbed for the algorithm's performance on heterogeneous samples.
  • Ex vivo bovine eyes: Real biological tissue (retinas) was used to demonstrate the method's applicability to complex, in-vivo-like scenarios.

The paper does not introduce a new benchmark dataset but rather uses these diverse datasets to rigorously validate their proposed method.

Key Insights and Ideas

  • The critical role of depth-dependent backscattering: The paper strongly emphasizes that assuming a constant backscattering fraction ($R$) is a major source of error in OCT attenuation estimation. Their method's ability to simultaneously characterize $R(z)$ is a key differentiator.
  • "Fuzzy" processing of R(z) for stability: The approach of averaging out minor variations in $R(z)$ while retaining significant changes is a clever way to balance accuracy with robustness against noise and minor structural variations. This prevents the algorithm from becoming unstable or producing artifacts.
  • Iterative refinement is essential: The iterative nature of the algorithm is crucial for decoupling the coupled parameters $\mu(z)$ and $R(z)$. The process allows for successive approximations that converge towards more accurate values.
  • Improved image clarity without prior segmentation: A significant advantage is that the method does not require pre-segmentation of tissue layers. This is particularly valuable for tissues with fuzzy boundaries or thin layers, where segmentation is often challenging and error-prone.
  • Enhanced diagnostic potential: By providing accurate, depth-resolved optical properties and improved image contrast, the method has the potential to significantly aid in the diagnosis of diseases affecting complex biological tissues.

Key Results

  • Accurate estimation of $\mu(z)$ and $R(z)$: The method demonstrated robust and precise estimation of both attenuation coefficients and backscattering fractions across simulated phantoms, experimental phantoms, and ex vivo retinal tissue.
  • Reduced estimation bias: Compared to conventional methods that assume constant $R$, the proposed iterative method significantly reduced under- and overestimation biases in $\mu(z)$, especially in regions with varying backscattering.
  • Improved OCT image quality: The attenuation-compensated OCT images showed enhanced resolution and visibility of deeper structures, providing more accurate morphological information.
  • Convergence within a few iterations: For most cases, the algorithm achieved satisfactory results within a few (2-4) iteration loops, indicating computational efficiency. Mean percentage errors between estimated and true values were reduced to below 10% after these iterations.

Limitations and Future Work

  • Global Convergence: The authors acknowledge that the algorithm does not guarantee global convergence, meaning it might converge to a local optimum rather than the absolute best solution. This is discussed in the Supplemental Material.
  • "Fuzzy" processing refinement: While effective, the "fuzzy" processing of $R(z)$ might need further fine-tuning for tissues with very subtle layer distinctions or where intra-layer variations in $R(z)$ are critical.
  • Multiple Scattering: The paper notes that multiple scattering can influence quantitative attenuation characterization, especially in highly scattering tissues. While their iterative method helps mitigate some additive errors, further work might explore analytical solutions that explicitly account for multiple scattering.
  • Visible Spectrum Application: The authors suggest that their method could be adapted for the visible spectrum, where light absorption is more significant. This would require adjusting the definition of attenuation ($\mu = \mu_a + \mu_s$) and backscattering fraction.
  • Computational Optimization: Converting the Matlab code to a more efficient programming language and employing parallel processing with GPUs could further optimize computation time, potentially reducing it to milliseconds per B-scan.

Summary updated

Abstract

Our proposed method was able to correct the estimation bias induced by the variations of <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"><mml:mrow><mml:mi>R</mml:mi></mml:mrow></mml:math> and provided accurate depth-resolved measurements of both <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"><mml:mrow><mml:mi>μ</mml:mi></mml:mrow></mml:math> and <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"><mml:mrow><mml:mi>R</mml:mi></mml:mrow></mml:math> simultaneously. The method does not require prior knowledge of the morphological information of tissue and represents more real-life tissues. Thus, it has the potential to help OCT imaging based disease diagnosis of complex and multi-layer biological tissue.
BibTeX
@article{wang2023depth,
  author = {Yaning Wang and Shuwen Wei and Jin U. Kang},
  title = {{Depth-dependent attenuation and backscattering characterization of optical coherence tomography by stationary iterative method}},
  journal = {Journal of Biomedical Optics},
  year = {2023},
  doi = {10.1117/1.jbo.28.8.085002},
}
Figure