Departments of Mathematics and Geosciences, University of Trieste, Via Valerio, 12/1, 34127, Trieste, Italy.
DAI Emergenza Urgenza ed Accettazione, Azienda Sanitaria Univeritaria integrata di Trieste, Trieste, Italy.
J Clin Monit Comput. 2021 Apr;35(2):289-296. doi: 10.1007/s10877-020-00469-z. Epub 2020 Jan 28.
Ineffective effort during expiration (IEE) occurs when there is a mismatch between the demand of a mechanically ventilated patient and the support delivered by a Mechanical ventilator during the expiration. This work presents a pressure-flow characterization for respiratory asynchronies and validates a machine-learning method, based on the presented characterization, to identify IEEs. 1500 breaths produced by 8 mechanically-ventilated patients were considered: 500 of them were included into the training set and the remaining 1000 into the test set. Each of them was evaluated by 3 experts and classified as either normal, artefact, or containing inspiratory, expiratory, or cycling-off asynchronies. A software implementing the proposed method was trained by using the experts' evaluations of the training set and used to identify IEEs in the test set. The outcomes were compared with a consensus of three expert evaluations. The software classified IEEs with sensitivity 0.904, specificity 0.995, accuracy 0.983, positive and negative predictive value 0.963 and 0.986, respectively. The Cohen's kappa is 0.983 (with 95% confidence interval (CI): [0.884, 0.962]). The pressure-flow characterization of respiratory cycles and the monitoring technique proposed in this work automatically identified IEEs in real-time in close agreement with the experts.
无效呼气努力（IEE）发生在机械通气患者的需求与机械通气在呼气期间提供的支持之间不匹配时。这项工作提出了一种呼吸不同步的压力-流量特征描述，并验证了一种基于该特征描述的机器学习方法，以识别 IEE。考虑了 8 位机械通气患者产生的 1500 次呼吸：其中 500 次被纳入训练集，其余 1000 次被纳入测试集。每个呼吸都由 3 位专家进行评估，并分类为正常、伪影或包含吸气、呼气或循环脱落不同步。实现所提出方法的软件使用训练集的专家评估进行训练，并用于在测试集中识别 IEE。结果与三位专家评估的共识进行了比较。该软件对 IEE 的分类具有敏感性 0.904、特异性 0.995、准确性 0.983、阳性预测值 0.963 和阴性预测值 0.986。Cohen's kappa 为 0.983（95%置信区间（CI）：[0.884，0.962]）。这项工作提出的呼吸周期的压力-流量特征描述和监测技术可以自动实时识别 IEE，与专家的评估结果非常吻合。
