[论文解读] Skin color independent robust assessment of capillary refill time

Raquel Pantojo de Souza Bachour, Eduardo Lopes Dias+1 more|arXiv (Cornell University)|Feb 26, 2021

Optical Imaging and Spectroscopy Techniques被引用 5

一句话总结

本研究提出了一种不受肤色影响、鲁棒性强的毛细血管再充盈时间（pCRT）评估方法，采用RGB摄像头、LED光源及交叉圆偏振片以减少表面反光。通过施加7 kPa压力持续5秒，并利用基于数据的截断点和基于不确定性的标记方法，对绿色通道强度衰减进行指数回归分析，该方法在所有Fitzpatrick皮肤光型I–VI中均实现了高重复性（80%的测量值在个体平均值±20%范围内）。

ABSTRACT

Capillary refill time (CRT) is a method for evaluating peripheral perfusion by visual assessment. CRT is especially useful for quick evaluations in the absence of sophisticated equipment. However, there are repeatability and reproducibility limitations with CRT, especially for dark skin. To test the limits of CRT repeatability and skin color independence, we developed a system and method to perform simple and robust CRT measurements. The system consists of an RGB camera and an LED lamp, with crossed circular polarizers imaging to attenuate the light reflected by the superficial layer of the skin. The capillary refill time is determined using an exponential regression on the time-dependent green channel mean pixel intensity of the region of interest after the compression is released. We limited this regression up to a data-dependent cut-off time, after which we assume the exponential model is invalid, and used the confidence interval of the uncertainty to develop a criterion to flag and discard faulty measurements. We tested the system on twenty-two volunteers with skin phototypes ranging from I to VI on the Fitzpatrick scale, applying to their forearms a 7 kPa compression for 5 s. After the release of measurements flagged as inadequate (about 20\% of measurements) by our regressions, our results indicated good precision, with high repeatability for all skin phototypes. Approximately 80\% of measurements fall within $\pm 20\%$ of the individual's expected value for CRT (mean CRT value). Our results suggest CRT can be used as a quantitative measurement and encourages further developments for the implementation of a similar method on smartphone cameras for quick and robust CRT measurements in patients' triage, monitoring, and telehealth.

研究动机与目标

• 解决手动毛细血管再充盈时间（CRT）测量缺乏标准化及可重复性差的问题，特别是对深色皮肤人群。
• 克服因观察者差异、光照条件、温度及皮肤光型影响导致的视觉CRT评估局限性。
• 开发一种低成本、鲁棒且客观的定量CRT测量方法，实现与肤色和观察者偏倚无关。
• 通过回归不确定性与动态截断时间，标记并剔除低质量读数，确保测量可靠性。
• 为未来在智能手机摄像头上的应用铺平道路，实现床旁、远程医疗及持续监测应用。

提出的方法

• 使用RGB摄像头与LED光源，配合交叉圆偏振片，以最小化表面反光并提升皮肤成像的信噪比。
• 使用带有特氟龙接触面的定制圆柱形装置，对前臂施加7 kPa的受控压力，持续5秒。
• 在释放压力后记录感兴趣区域（ROI）的视频，并提取绿色通道强度的时间序列均值。
• 采用非线性回归对强度恢复曲线拟合指数衰减模型，使用基于数据的截断时间以排除无效的再充盈后阶段。
• 利用回归的置信区间识别并标记不可靠测量结果以进行剔除。
• 将pCRT定义为指数拟合的时间常数，代表在受控、客观条件下测得的毛细血管再充盈时间。

Figure 1: Experimental setup. a) Schematic illustration of the weight and arm support, front view; (1) Standard aluminum cylindrical weight; (2) $4cm^{2}$ thermally insulating Teflon tip, that comes into contact with the subject’s skin; (3) Armrest (dense polyurethane foam). b) Setup with a voluntee

实验结果

研究问题

• RQ1 能否使用低成本、基于摄像头的系统，在所有Fitzpatrick皮肤光型中可靠且可重复地测量毛细血管再充盈时间？
• RQ2 使用交叉圆偏振片是否能显著提升测量的鲁棒性，从而减少表面反光伪影？
• RQ3 基于数据的截断时间与回归不确定性标准是否能有效识别并剔除故障测量？
• RQ4 与传统CRT中使用的较高压力相比，低压力压缩（7 kPa）对测量重复性与准确性有何影响？
• RQ5 该方法在多大程度上独立于观察者偏倚与肤色，使其适用于多样化临床人群？

主要发现

• 该方法实现了高重复性，约80%的pCRT测量值落在每位受试者平均CRT值的±20%范围内，涵盖所有皮肤光型。
• 约20%的测量值因拟合效果差或不确定性高而被标记并剔除，表明质量控制有效。
• 该系统在所有Fitzpatrick皮肤类型（I–VI）中均表现出鲁棒性，证实了pCRT测量对肤色的独立性。
• 采用低至7 kPa的压缩压力提升了测量稳定性并降低了噪声，而较高压力则导致拟合质量下降。
• 结合动态截断时间的指数回归模型显著提高了pCRT估计的可靠性与重复性。
• 结果支持在智能手机摄像头上实现类似pCRT系统的可行性，实现非侵入性、客观且可及的外周灌注监测。

Figure 2: Mean ROI pixel intensities during a pCRT experiment. a) Mean intensities of the R, G and B channels from the pixels inside the ROI. The cylindrical weight blocking the camera during compression causes the sharp drop in intensities observed between 14 s and 19 s. After the weight is lifted,