Sorelli Michele, Ince Can, Bocchi Leonardo
Dept. of Information Engineering, University of Florence, Florence, Italy.
Physiol Meas. 2017 Feb;38(2):358-373. doi: 10.1088/1361-6579/aa56ab. Epub 2017 Jan 4.
In recent years the development of portable microscopes, which enable the noninvasive bedside evaluation of the sublingual microcirculation in critically ill patients, has expanded the clinical research on this level of the cardiovascular system. Several semi-quantitative scores have been defined in order to provide researchers with a standardized framework for the offline assessment of the microcirculation status. Among those, space-time diagrams (STDs) constitute an established method for obtaining an estimate of the red blood cells (RBCs) flow velocity in capillaries. However, STDs have the drawback of being time-consuming, inherently subjective, and difficult to manage when the flow is not regular.
In this work we propose an automated method for calculating erythrocyte flow speed, aiming to provide a fast and objective tool for the evaluation of peripheral blood perfusion.
The proposed method exploits an image segmentation module for estimating the positions of candidate flowing cells. A multi-object tracking algorithm based on Kalman filters analyzes and matches the positions corresponding to specific erythrocytes within consecutive frames. Thus, the output of the filter enables to estimate the displacement of each cell, yielding their instantaneous speed.
The method has been validated against the results obtained by the manual analysis of STDs, proving a good agreement for speeds up to 300 μm s. At higher speeds, RBC tracking becomes unstable due to the currently limited video acquisition rate (25 Hz) of state-of-the-art devices, that makes the matching between objects appearing in consecutive frames very challenging.
近年来，便携式显微镜的发展使得对危重症患者舌下微循环进行无创床边评估成为可能，这拓展了对心血管系统这一层次的临床研究。为了给研究人员提供一个用于微循环状态离线评估的标准化框架，已经定义了几种半定量评分。其中，时空图（STDs）是一种用于估计毛细血管中红细胞（RBCs）流速的既定方法。然而，时空图存在耗时、本质上主观以及当血流不规则时难以处理的缺点。
在这项工作中，我们提出一种计算红细胞流速的自动化方法，旨在为外周血灌注评估提供一种快速且客观的工具。
所提出的方法利用一个图像分割模块来估计候选流动细胞的位置。一种基于卡尔曼滤波器的多目标跟踪算法分析并匹配连续帧中特定红细胞对应的位置。因此，滤波器的输出能够估计每个细胞的位移，从而得出它们的瞬时速度。
该方法已根据通过时空图手动分析获得的结果进行了验证，结果表明对于高达300μm/s的速度具有良好的一致性。在更高速度下，由于当前最先进设备有限的视频采集速率（25Hz），红细胞跟踪变得不稳定，这使得连续帧中出现的物体之间的匹配非常具有挑战性。
