Department of Electrical and Computer Engineering, Kansas State University, Manhattan, KS 66506, USA,
IEEE Trans Biomed Circuits Syst. 2012 Feb;6(1):54-63. doi: 10.1109/TBCAS.2011.2157822.
Onboard assessment of photoplethysmogram (PPG) quality could reduce unnecessary data transmission on battery-powered wireless pulse oximeters and improve the viability of the electronic patient records to which these data are stored. These algorithms show promise to increase the intelligence level of former "dumb" medical devices: devices that acquire and forward data but leave data interpretation to the clinician or host system. To this end, the authors have developed a unique onboard feature detection algorithm to assess the quality of PPGs acquired with a custom reflectance mode, wireless pulse oximeter. The algorithm uses a Bayesian hypothesis testing method to analyze four features extracted from raw and decimated PPG data in order to determine whether the original data comprise valid PPG waveforms or whether they are corrupted by motion or other environmental influences. Based on these results, the algorithm further calculates heart rate and blood oxygen saturation from a "compact representation" structure. PPG data were collected from 47 subjects to train the feature detection algorithm and to gauge their performance. A MATLAB interface was also developed to visualize the features extracted, the algorithm flow, and the decision results, where all algorithm-related parameters and decisions were ascertained on the wireless unit prior to transmission. For the data sets acquired here, the algorithm was 99% effective in identifying clean, usable PPGs versus nonsaturated data that did not demonstrate meaningful pulsatile waveshapes, PPGs corrupted by motion artifact, and data affected by signal saturation.
在板评估光体积描记图 (PPG) 的质量可以减少电池供电的无线脉搏血氧计上不必要的数据传输，并提高电子病历的生存能力，这些数据存储在这些电子病历中。这些算法有望提高以前的“哑”医疗设备的智能水平：这些设备获取和转发数据，但将数据解释留给临床医生或主机系统。为此，作者开发了一种独特的板载特征检测算法，以评估使用定制反射模式、无线脉搏血氧计采集的 PPG 的质量。该算法使用贝叶斯假设检验方法来分析从原始和抽取的 PPG 数据中提取的四个特征，以确定原始数据是否包含有效的 PPG 波形，或者它们是否受到运动或其他环境影响的干扰。基于这些结果，该算法进一步从“紧凑表示”结构计算心率和血氧饱和度。从 47 名受试者收集 PPG 数据，以训练特征检测算法并评估其性能。还开发了一个 MATLAB 接口来可视化提取的特征、算法流程和决策结果，其中所有与算法相关的参数和决策都是在传输之前在无线单元上确定的。对于这里采集的数据，该算法能够有效识别 99%的干净、可用的 PPG 与未饱和数据，未饱和数据不显示有意义的脉动波形、运动伪影干扰的 PPG 和受信号饱和影响的数据。
