Transcranial magnetic stimulation (TMS) allows the assessment of stroke patients’ cortical excitability and corticospinal tract integrity, which provide information regarding motor function recovery. However, the extraction of features from motor-evoked potentials (MEP) elicited by TMS, such as amplitude and latency, is performed manually, increasing variability due to observer-dependent subjectivity. Therefore, an automatic methodology could improve MEP analysis, especially in stroke, which increases the difficulty of manual MEP measurements due to brain lesions. A methodology based on time-frequency features of stroke patients’ MEPs that allows to automatically select and extract MEP amplitude and latency is proposed. The method was validated using manual measurements, performed by three experts, computed from patients’ affected and unaffected hemispheres. Results showed a coincidence of 58.3 to 80% between automatic and manual MEP selection. There were no significant differences between the amplitudes and latencies computed by two of the experts with those obtained with the automatic method, for most comparisons. The median relative error of amplitudes and latencies computed by the automatic method was 5% and 23%, respectively. Therefore, the proposed method has the potential to reduce processing time and improve the computation of MEP features, by eliminating observer-dependent variability due to the subjectivity of manual measurements.

Graphical abstract

中文翻译：

---

脑卒中患者经颅磁刺激运动诱发电位的自动选择与特征提取

经颅磁刺激 (TMS) 可以评估中风患者的皮质兴奋性和皮质脊髓束完整性，从而提供有关运动功能恢复的信息。然而，从 TMS 引起的运动诱发电位 (MEP) 中提取特征，例如振幅和延迟，是手动执行的，由于观察者相关的主观性而增加了可变性。因此，自动方法可以改进 MEP 分析，尤其是在中风中，这增加了由于脑部病变而导致的手动 MEP 测量的难度。提出了一种基于中风患者 MEP 时频特征的方法，该方法允许自动选择和提取 MEP 幅度和潜伏期。该方法使用手动测量进行了验证，由三位专家执行，根据患者受影响和未受影响的半球计算。结果显示自动和手动 MEP 选择之间的重合率为 58.3% 至 80%。对于大多数比较，由两位专家计算的幅度和延迟与使用自动方法获得的幅度和延迟之间没有显着差异。自动方法计算的幅度和延迟的中值相对误差分别为 5% 和 23%。因此，所提出的方法有可能通过消除由于手动测量的主观性而导致的依赖于观察者的可变性，从而减少处理时间并改进 MEP 特征的计算。对于大多数比较，由两位专家计算的幅度和延迟与使用自动方法获得的幅度和延迟之间没有显着差异。自动方法计算的幅度和延迟的中值相对误差分别为 5% 和 23%。因此，所提出的方法有可能通过消除由于手动测量的主观性而导致的依赖于观察者的可变性，从而减少处理时间并改进 MEP 特征的计算。对于大多数比较，由两位专家计算的幅度和延迟与使用自动方法获得的幅度和延迟之间没有显着差异。自动方法计算的幅度和延迟的中值相对误差分别为 5% 和 23%。因此，所提出的方法有可能通过消除由于手动测量的主观性而导致的依赖于观察者的可变性，从而减少处理时间并改进 MEP 特征的计算。

图形概要