Department of Safety and Security, AIT Austrian Institute of Technology GmbH, eHealth, Reininghausstraße 13, A-8020 Graz, Austria.
Physiol Meas. 2012 Sep;33(9):1449-61. doi: 10.1088/0967-3334/33/9/1449. Epub 2012 Aug 17.
Although immediate feedback concerning ECG signal quality during recording is useful, up to now not much literature describing quality measures is available. We have implemented and evaluated four ECG quality measures. Empty lead criterion (A), spike detection criterion (B) and lead crossing point criterion (C) were calculated from basic signal properties. Measure D quantified the robustness of QRS detection when applied to the signal. An advanced Matlab-based algorithm combining all four measures and a simplified algorithm for Android platforms, excluding measure D, were developed. Both algorithms were evaluated by taking part in the Computing in Cardiology Challenge 2011. Each measure's accuracy and computing time was evaluated separately. During the challenge, the advanced algorithm correctly classified 93.3% of the ECGs in the training-set and 91.6 % in the test-set. Scores for the simplified algorithm were 0.834 in event 2 and 0.873 in event 3. Computing time for measure D was almost five times higher than for other measures. Required accuracy levels depend on the application and are related to computing time. While our simplified algorithm may be accurate for real-time feedback during ECG self-recordings, QRS detection based measures can further increase the performance if sufficient computing power is available.
虽然记录过程中有关 ECG 信号质量的即时反馈很有用，但到目前为止，可用的质量评估方法并不多。我们已经实现并评估了四种 ECG 质量评估方法。空导联标准（A）、尖峰检测标准（B）和导联交叉点标准（C）是根据基本信号特性计算得出的。D 测度量化了应用于信号时 QRS 检测的稳健性。开发了一种基于高级 Matlab 的算法，该算法结合了所有四种方法，以及一种适用于 Android 平台的简化算法，该算法不包括方法 D。这两种算法都参加了 2011 年计算心脏病学挑战赛进行了评估。分别评估了每种方法的准确性和计算时间。在挑战赛期间，高级算法在训练集中正确分类了 93.3%的 ECG，在测试集中正确分类了 91.6%的 ECG。事件 2 中简化算法的得分为 0.834，事件 3 中简化算法的得分为 0.873。方法 D 的计算时间几乎是其他方法的五倍。所需的准确性水平取决于应用程序，并与计算时间有关。虽然我们的简化算法可能适用于 ECG 自我记录的实时反馈，但如果有足够的计算能力，基于 QRS 检测的方法可以进一步提高性能。
