Dynamic-OCT simulation framework based on mathematical models of intratissue dynamics, image formation, and measurement noise

TL;DR

The paper tackles the challenge of interpreting dynamic OCT signals by constructing a comprehensive DOCT simulation framework that couples intracellular/intratissue motion models with DSM-based signal formation and realistic noise. It implements three DOCT metrics (LIV, OCDS, and amplitude-spectrum-based DOCT) and demonstrates how motion parameters, dynamic-scatterer occupancy, and noise influence these contrasts through fully dynamic and dynamic-scatterer-ratio studies. The framework enables mechanistic probing of motion-to-signal relationships and supports future algorithm design by providing an open-source Python implementation that can guide DOCT interpretation and robustness to noise. Overall, the work provides both methodological tools and practical insights to advance DOCT understanding and development for tissue dynamics imaging.

Abstract

Dynamic optical coherence tomography (DOCT) enables label-free functional imaging by capturing temporal OCT signal variations caused by intracellular and intratissue motions. However, the relationship between DOCT signals and the sample motion behind them remains unclear. This paper presents a comprehensive DOCT simulation framework that incorporates mathematical models of intracellular/intratissue motions, two OCT signal generator types that generate OCT signal time sequences from the moving scatterer models, and representative DOCT algorithms. The theory and algorithms of the framework are described in detail, and the utility of this framework is demonstrated through numerical studies. This framework is available as open source and will enhance the understanding and utility of DOCT.

Figures (10)

• Figure 1: Summary of intracellular/intratissue activities and mathematical models of the intracellular/intratissue motions. Seven activity types are included, and each activity type is expressed using one of three types of motion models.
• Figure 2: Schematic illustrations of the three scatterer motion models. (a) In the random ballistic model, all scatterers are move linearly with the same speed, but the directions of motion are randomly defined for each scatterer. (b) In the diffusion model, the scatterers follow the random walk motion (i.e., Brownian motion). (c) In the mono-directional translation model, all scatterers travel linearly in the same direction at the same speed. $i$ is the time point index, and $t_{i-1}$, $t_{i}$, and $t_{i+1}$ represent three time points.
• Figure 3: The flow to generate 3D OCT volume, i.e., 3D speckle pattern, using the 3D-OCT-volume formation model. The details of the flow is described in Section \ref{['sec:3DsimualtionFlow']}.
• Figure 4: Example time sequences of simulated speckles. The scatterers here follow one of the random ballistic, diffusion, or mono-directional translation models, as indicated on the left of each of the images. The first three rows show 2D cross-sections extracted from the 3D volumes and the last three rows show volume rendering of the 3D speckle. The speckle motion can also be seen in the corresponding videos: Visualization 1 (random ballistic), Visualization 2 (diffusion), and Visualization 3 (mono-directional translation).
• Figure 5: Mean LIV values shown as motion parameters (the scatterer speed and the diffusion coefficient). The error bars indicate the standard deviations among the eight trials. The first label on each horizontal axis represents the motion parameters, and the second label represents the corresponding travel distance during the full simulation time. The vertical dashed lines represent half the wavelength (green) and the lateral (red) and axial (blue) resolutions. Plots (a) to (c) represent (a) the random ballistic motion, (b) diffusion, and (c) mono-directional translation. The mean LIV values were computed for five different SNR configurations. The LIV was found to increase as the relevant motion parameter value increased.