Avanços na monitoração da saturação de oxigênio em tecidos biológicos e vigilância do SARS-CoV-2: integração da imagem no domínio da frequência espacial com rede neural artificial

Abstract

The infectious disease coronavirus 2019 (COVID-19), caused by Severe Acute Respiratory Syndrome (SARS) Coronavirus 2 (CoV-2), also known as SARS-CoV-2, can range from asymptomatic to severe respiratory problems and multiple organ failure. The complexity and rapid progression of the disease, often without noticeable signs, highlight the need for close monitoring and early intervention. Among the possible approaches, visualization of oxygenation in large tissue areas can significantly improve the management of patients in intensive care. The Spatial Frequency Domain Imaging (SFDI) technique has shown promise for measuring the optical properties of biological tissues, offering rapid and accurate measurements over large areas and enabling this type of monitoring. In this study, we developed a technique that combines SFDI with Artificial Neural Network (ANN) to determine oxyhemoglobin (Coxy), deoxyhemoglobin (Cdeoxy), and oxygen saturation (SpO2) concentrations in biological tissues. Using optical phantoms made of polydimethylsiloxane (PDMS) matrix, titanium dioxide (TiO2) as scattering agent, and India ink as absorbing agent, we evaluated their optical properties with the Inverse Adding-Doubling (IAD) and SFDI techniques to validate and calibrate the SFDI optical device. To apply the IAD technique, we used an integrating sphere coupled to a spectrometer to measure the spectrum of transmitted and diffusely reflected light. SFDI images were obtained by projecting two-dimensional light patterns with different spatial frequencies onto the sample and analyzing the effects of scattering and absorption on the diffuse reflectance (Rd) values captured by a camera. Using a MATLAB algorithm, we created spatial maps of light absorption and scattering on the surface of the samples. The use of ANN allowed us to directly determine the concentrations of oxyhemoglobin and deoxyhemoglobin from the SFDI images, without the need to measure absorption coefficients at different wavelengths. The model showed correlations of 0.997 and 0.982 for the concentrations of oxyhemoglobin and deoxyhemoglobin, respectively, with average errors of 0.98% and 0.99%. In an in vivo study, the technique demonstrated feasibility in obtaining functional images of the concentrations of oxyhemoglobin, deoxyhemoglobin, and oxygen saturation. Oxygenation imaging provided quantitative measurements in irregularly shaped tissues, without direct contact and with a fast response. Although promising, the technique faces limitations, including the need for post-processing of the images and the lack of three-dimensional measurements of the tissue profiles, which limit accuracy. Addressing these limitations will be crucial to adapt the technique to more precise applications and to increase its sensitivity to different experimental conditions.