Signal-based feature extraction in motor evoked potentials: TKEO onset detection and Hilbert envelope analysis

The reliable and objective determination of the characteristics of the motor evoked potential (MEP) – latency of occurrence, amplitude from peak to peak, duration and morphology of the waveform – is fundamental for clinical neurophysiology, but in modern practice it largely depends on the judgments of the operator. Mathematical algorithms for signal processing offer a transparent, deterministic and reproducible alternative. We present, characterize, and systematically evaluate a complete mathematical algorithm for identifying MEP features, consisting of three stages: determining the origin based on TKEO, the Tiger-Kaiser energy operator applied to a pre-processed signal with an adaptive threshold k∙σ_baseline; Estimating the displacement of the Hilbert transform – amplitude envelope tracking using a baseline return criterion; and morphological classification by counting significant zero crossings to assign monophase, two-phase, or multiphase labels. At the marker verification stage, tests in which the detected signs do not exceed the minimum noise level are rejected. With an SNR value of 3.0, performance decrease, the MAE delay increases from 1.4 ms (SNR ≥ 5) to 9.7 ms (SNR < 3). The accuracy of morphological classification is 94% for studies with high SNR and decreases to 61% for studies with very low SNR. The mathematical pipeline provides clinically acceptable accuracy for MEP with high and medium SNR levels and serves as an interpretable reference standard with zero training costs. Its failure modes are well characterized, SNR-dependent, and predictable – properties that make it a basic baseline comparator for evaluating more advanced automated analysis methods.

1. Rossini P.M., Burke D., Chen R., et al. Non-invasive electrical and magnetic stimulation of the brain, spinal cord, roots and peripheral nerves: Basic principles and procedures for routine clinical and research application. An updated report from an I.F.C.N. Committee. Clinical Neurophysiology. 2015;126(6):1071–1107. https://doi.org/10.1016/j.clinph.2015.02.001

2. Olney R.K., So Y.T., Goodin D.S., Aminoff M.J. A comparison of magnetic and electrical stimulation of peripheral nerves. Muscle Nerve. 1990;13(10):957–963. https://doi.org/10.1002/mus.880131012

3. Stålberg E., van Dijk H., Falck B., et al. Standards for quantification of EMG and neurography. Clinical Neurophysiology. 2019;130(9):1688–1729. https://doi.org/10.1016/j.clinph.2019.05.008

4. Winter D.A. Biomechanics and Motor Control of Human Movement. Hoboken: John Wiley & Sons; 2009. 384 p.

5. MacDonald D.B., Skinner S., Shils J., Yingling C. Intraoperative motor evoked potential monitoring – А position statement by the American Society of Neurophysiological Monitoring. Clinical Neurophysiology. 2013;124(12):2291–2316. https://doi.org/10.1016/j.clinph.2013.07.025

6. Kaiser J.F. On a simple algorithm to calculate the 'energy' of a signal. In: International Conference on Acoustics, Speech, and Signal Processing, 03–06 April 1990, Albuquerque, NM, USA. IEEE; 1990. P. 381–384. https://doi.org/10.1109/ICASSP.1990.115702

7. Solnik S., Rider P., Steinweg K., DeVita P., Hortobágyi T. Teager-Kaiser energy operator signal conditioning improves EMG onset detection. European Journal of Applied Physiology. 2010;110(3):489–498. https://doi.org/10.1007/s00421-010-1521-8

8. Huang N.E., Shen Zh., Long S.R., et al. The empirical mode decomposition and the Hilbert spectrum for nonlinear and non-stationary time series analysis. Proceedings of the Royal Society A. 1998;454(1971):903–995. https://doi.org/10.1098/rspa.1998.0193

9. Virtanen P., Gommers R., Oliphant T.E., et al. SciPy 1.0: fundamental algorithms for scientific computing in Python. Nature Methods. 2020;17:261–272. https://doi.org/10.1038/s41592-019-0686-2

10. Ma K., Wang B., Liu S., Goetz S.M. Rethinking noise floor characterisation in motor-evoked potentials. Journal of Neural Engineering. 2025;22(3). https://doi.org/10.1088/1741-2552/add20d

11. Teager H.M., Teager S.M. Evidence for nonlinear sound production mechanisms in the vocal tract. In: Speech Production and Speech Modelling. Dordrecht: Springer; 1990. P. 241–261. https://doi.org/10.1007/978-94-009-2037-8_10

12. Gabor D. Theory of communication. Part 1: The analysis of information. Journal of the Institution of Electrical Engineers – Part III: Radio and Communication Engineering. 1946;93(26):429–441.

13. Shrout P.E., Fleiss J.L. Intraclass correlations: Uses in assessing rater reliability. Psychological Bulletin. 1979;86(2):420–428. https://doi.org/10.1037/0033-2909.86.2.420

14. Li Zh., Peterchev A.V., Rothwell J.C., Goetz S.M. Detection of motor-evoked potentials below the noise floor: rethinking the motor stimulation threshold. Journal of Neural Engineering. 2022;19(5). https://doi.org/10.1088/1741-2552/ac7dfc

15. Gramfort A., Luessi M., Larson E., et al. MEG and EEG data analysis with MNE-Python. Frontiers in Neuroscience. 2013;7. https://doi.org/10.3389/fnins.2013.00267