From the Department of Emergency Medicine (J.N.C.), Saint Vincent Hospital; Morosky College of Health Professionals (J.N.C.), Gannon University, Erie, PA; Department of Mechanical and Industrial Engineering (S.C., H.S.K., C.L.), Southern Illinois University, Edwardsville, IL; Department of Biomedical, Industrial and Systems Engineering (I.O.), Gannon University, Erie, PA; RQI Partners (R.E.G.), LLC, Dallas, TX; and Department of Emergency Medicine (H.E.W.), University of Texas Health Science Center at Houston, Houston, TX.
Simul Healthc. 2020 Jun;15(3):160-166. doi: 10.1097/SIH.0000000000000426.
Endotracheal intubation (ETI) is an important emergency intervention. Only limited data describe ETI skill acquisition and often use bulky technology, not easily transitioned to the clinical setting. In this study, we used small, portable inertial detection technology to characterize intubation kinematic differences between experienced and novice intubators.
We performed a prospective study including novice (<10 prior clinical ETI) and experienced (>100 clinical ETI) emergency providers. We tracked upper extremity motion with roll, pitch, and yaw using inertial measurement units (IMU) placed on the bilateral hands and wrists of the intubator. Subject performed 6 simulated emergency intubations on a mannequin. Using machine learning algorithms, we determined the motions that best discriminated experienced and novice providers.
We included data on 12 novice and 5 experienced providers. Four machine learning algorithms (artificial neural network, support vector machine, decision tree, and K-nearest neighbor search) were applied. Artificial neural network had the greatest accuracy (95% confidence interval) for discriminating between novice and experienced providers (91.17%, 90.8%-91.5%) and was the most parsimonious of the tested algorithms. Using artificial neural network, information from 5 movement features (right hand, roll amplitude; right hand, pitch amplitude; right hand, yaw standard deviation; left hand, yaw standard deviation; left hand, pitch frequency of peak amplitude) was able discriminated experienced from novice providers.
Novice and experienced providers have different ETI movement patterns and can be distinguished by 5 specific movements. Inertial detection technology can be used to characterize the kinematics of emergency airway management.
气管插管（ETI）是一项重要的急救干预措施。只有有限的数据描述了 ETI 技能的获取，并且这些数据通常使用庞大的技术，不容易转化到临床环境中。在这项研究中，我们使用小型便携式惯性检测技术来描述经验丰富和新手插管者之间插管运动学的差异。
我们进行了一项前瞻性研究，包括新手（<10 次以前的临床 ETI）和有经验的（>100 次临床 ETI）急救提供者。我们使用惯性测量单元（IMU）跟踪插管者双侧手和手腕的滚动、俯仰和偏航运动。受测者在模拟人身上进行了 6 次紧急插管。使用机器学习算法，我们确定了最能区分有经验和新手提供者的运动。
我们纳入了 12 名新手和 5 名有经验的提供者的数据。应用了 4 种机器学习算法（人工神经网络、支持向量机、决策树和 K-最近邻搜索）。人工神经网络在区分新手和有经验的提供者方面具有最高的准确性（95%置信区间）（91.17%，90.8%-91.5%），并且是测试算法中最简洁的算法。使用人工神经网络，来自 5 个运动特征（右手，滚动幅度；右手，俯仰幅度；右手，偏航标准差；左手，偏航标准差；左手，峰值幅度的俯仰频率）的信息可以区分有经验和新手提供者。
新手和有经验的提供者在 ETI 运动模式上存在差异，可以通过 5 个特定的运动来区分。惯性检测技术可用于描述紧急气道管理的运动学。
