Combined Monte Carlo and k-Wave Simulations for Reconstruction of Blood Oxygen Saturation in Optoacoustics: A Pilot Study

Abstract

Optoacoustic (OA) imaging of biological tissues is a modern technique allowing for three-dimensional blood oxygen saturation mapping based on OA spectroscopy data. Since biological tissues are optically inhomogeneous and the spatial distribution of optical parameters within a biological tissue is a priori unknown, Monte Carlo simulation technique is traditionally used to estimate the distribution of probing illumination within tissues in quantitative OA reconstruction. Currently, machine learning techniques are actively employed for reconstructing 3D distribution of blood oxygen saturation or estimating optical properties of biological tissues based on training datasets. In this paper, systemic calculations of synthetic OA images of a medium with embedded vessel-like structures were performed to create a training dataset for machine learning employing combined application of the Monte Carlo technique for direct solution of optical problem and difference-space pseudo-spectral approach implemented through k-Wave Toolbox calculations for the acoustical part. The calculations were performed for probing wavelengths of 532 nm, 658 nm and 1064 nm, which are commonly employed in spectral OA imaging. Simulated OA data for different orientation, diameter and embedding depth of blood vessels allows analyzing the effect of these parameters on the formation of OA image and the reconstruction of blood oxygen saturation. The ratio of OA signals corresponding to probing wavelengths of 658 nm and 1064 nm was employed for simple reconstruction of blood oxygen saturation in silico for different vessel geometries with the precision of < 3-15% for the most of blood vessels diameters and embedding depths and the range of blood oxygen saturation values ≥ 0.8. The obtained set of synthetic OA data has high potential as a training set for employment in machine learning techniques aiming at mapping blood oxygenation based on spectral OA data.

4D spectral-spatial computational photoacoustic dermoscopy

“…Skin is a complex multilayered heterogeneous tissue, and the depth and direction of light propagation within the skin are determined by the optical properties of the various layers of tissue and blood vessels in the skin, which are wavelength dependent and vary according to the random inhomogeneous distribution of various chromophores and pigments [38] . For simplicity, each layer is typically treated as a homogeneous structure in the computational model, and the optical properties vary between layers but remain constant within each layer [28] , [29] , [30] , [31] , [32] , [33] . Typically, the optical properties of each skin layer include the absorption coefficient ( μ a ), scattering coefficient ( μ s ), anisotropy factor ( g ), and refractive index ( n ).…”

“…For the simulation of acoustics, tools such as the k-Wave toolbox [26] and the finite element method, with platforms like COMSOL, are predominant. The k-Wave, in particular, has gained traction due to its efficiency and simplicity [27] , [28] , [29] , [30] , [31] , [32] . Despite these advances, the current literature has limitations in certain aspects of photoacoustic dermoscopy (PAD) simulations, especially models that incorporate the intricate interactions between excitation optical fields and detectable ultrasonic fields.…”

Machine-learning-based mapping of blood oxygen saturation from dual-wavelength optoacoustic measurements

“…MC modeling of light propagation in tissue-like media includes simulations of a large number of random photon trajectories followed by statistical analysis of calculated data. For this study we applied a previously developed platform for 3D MC simulation in vascularized tissue [17,18]. The 3D maps of absorbed light dose distribution H (r, λ ex ) at the probing wavelength λ ex in a flat tissue-like homogenous medium with a full size of 20 × 20 × 10 mm 3 containing a vascular structure consisting of blood vessels were generated for two probing wavelengths: λ ex = 532 and 1064 nm, which are traditionally employed for OA imaging [19].…”