پیش‌بینی سری‌های زمانی سیگنال‌های دینامیکی آشوبناک با استفاده از روش‌های یادگیری ماشین برای تغییرپذیری ضربان قلب

نوع مقاله : علمی-پژوهشی

نویسندگان

1 گروه آموزش کنترل. دانشکده مهندسی برق و کامپیوتر. دانشگاه کاشان. کاشان. ایران

2 استادیار دانشکده مهندسی برق و کامپیوتر دانشگاه کاشان

چکیده

تغییرپذیری ضربان قلب (HRV)، که از فواصل RR در سیگنال‌های ECG به دست می‌آید، فعالیت سیستم عصبی خودکار را منعکس می‌کند که پیش‌بینی آن به دلیل ماهیت غیرخطی و آشوبناک آن چالش برانگیز است. در این مقاله، ابتدا سیگنال‌های ECG برچسب‌گذاری شده از پایگاه داده Physionet به دست آمدند. سپس، فیلتر میان‌گذر مرتبه چهارم Butterworth و بازسازی فضای فاز با رویکرد پنجره کشویی بر روی این سیگنال‌های آشوبناک اعمال شده و با چهار روش یادگیری ماشین آموزش داده شدند. از آنجایی که LSTM و CNN به ترتیب به دلیل قابلیت‌های زمانی و فرکانسی خود عملکرد بهتری نسبت به سایر روش‌ها نشان دادند، یک روش ترکیبی LSTM+CNN پیشنهاد شد که معیارهای ارزیابی RMSE و R² را بهبود بخشید و نشان داده شد که دارای قابلیت‌های زیر است: مقاومت در برابر نویز سفید و صورتی، توانایی تشخیص سیگنال‌های بیمار آریتمی بدون برچسب از سیگنال‌های سالم، و توانایی پیش‌بینی سیگنال‌های HRV سالم و ناسالم.

کلیدواژه‌ها

موضوعات

عنوان مقاله [English]

Time-Series Forecasting of Chaotic Dynamic Signals by Machine Learning Methods for Heart Rate Variability

نویسندگان [English]

  • Masiha Asgharinezhad 1
  • Alireza Faraji 2
1 Control Engineering Group, Department of Electrical and computer Engineering, University of Kashan, Kashan, Iran
2 Assistant professor of Electrical and Computer engineering Faculty. University of Kashanu. Kashan. Iran
چکیده [English]

Heart rate variability (HRV), obtained from RR intervals in ECG signals, reflects the activity of the autonomic nervous system, which is challenging to forecast due to its nonlinear and chaotic nature. In this paper, labeled ECG signals were first obtained from the Physionet database. Then, fourth-order Butterworth bandpass filter and phase space reconstruction with a sliding window approach were applied to these chaotic signals and trained with four machine learning methods. Since LSTM and CNN showed better performance than other methods due to their temporal and frequency capabilities, respectively, a combined LSTM+CNN method was proposed, which improved the RMSE and R² evaluation criteria and was shown to have the following capabilities: robustness against white and pink noise, the ability to distinguish unlabeled arrhythmia patient signals from healthy signals, and the ability to forecast both healthy and unhealthy HRV signals.

کلیدواژه‌ها [English]

  • Chaotic dynamic signal
  • Time-Series Forecasting
  • Phase space reconstruction
  • Machine Learning
  • HRV
  • RR Intervals

مراجع

[1] T. Ince, S. Kiranyaz, and M. Gabbouj, “A generic and robust system for automated patient-specific classification of ECG signals,” IEEE Trans. Biomed. Eng., vol. 56, no. 5, pp. 1415–1426, May 2009, doi: 10.1109/TBME.2009.2012082. (Cited in Section 1 to support HRV analysis from ECG signals.)
[2] U. R. Acharya, K. P. Joseph, N. Kannathal, C. M. Lim, and J. S. Suri, “Heart rate variability: A review,” Med. Biol. Eng. Comput., vol. 44, no. 12, pp. 1031–1051, Dec. 2006, doi: 10.1007/s11517-006-0119-0. (Cited in Section 1 to provide background on HRV’s clinical importance.)
[3] M. R. Mohebbi and H. Ghassemian, “Prediction of paroxysmal atrial fibrillation using feature-based and deep learning approaches,” Biomed. Signal Process. Control, vol. 57, p. 101737, Mar. 2020, doi: 10.1016/j.bspc.2019.101737. (Cited in Section 1 to highlight ML applications in cardiac prediction.)
[4] J. Y. Zheng, C. D. Lin, and W. J. Wang, “A novel hybrid deep learning model for HRV analysis based on CNN and LSTM,” Biomed. Eng. Online, vol. 19, no. 1, p. 32, May 2020, doi: 10.1186/s12938-020-00773-7. (Cited in Section 1 and 2.3 to support the hybrid LSTM+CNN model.)
[5] P. A. Absil, R. Sepulchre, and R. Mahony, “Continuous-time source separation using geodesic flows,” IEEE Trans. Neural Netw., vol. 15, no. 4, pp. 865–873, Jul. 2004, doi: 10.1109/TNN.2004.828758. (Cited in Section 1 to support ML for chaotic signals.)
[6] G. E. Moody and R. G. Mark, “The impact of the MIT-BIH arrhythmia database,” IEEE Eng. Med. Biol. Mag., vol. 20, no. 3, pp. 45–50, May 2001, doi: 10.1109/51.932724. (Cited in Section 2.2 to reference the dataset source.)
[7] H. W. et al., “Deep learning approaches for ECG signal analysis: A review,” IEEE Rev. Biomed. Eng., vol. 14, pp. 68–84, 2021, doi: 10.1109/RBME.2020.2978752. (Cited in Section 1 to support deep learning for ECG analysis.)
[8] S. Kiranyaz, T. Ince, and M. Gabbouj, “Real-time patient-specific ECG classification by 1-D convolutional neural networks,” IEEE Trans. Biomed. Eng., vol. 63, no. 3, pp. 664–675, Mar. 2016, doi: 10.1109/TBME.2015.2468581. (Cited in Section 1 to support CNN applications in ECG signals.)
[9] Y. LeCun, Y. Bengio, and G. Hinton, “Deep learning,” Nature, vol. 521, no. 7553, pp. 436–444, May 2015, doi: 10.1038/nature14539. (Cited in Section 1, 2.3, and 2.4 to support deep learning and CNNs.)
[10] R. A. et al., “Deep convolutional neural network for the automated detection and diagnosis of heart diseases,” Biomed. Signal Process. Control, vol. 44, pp. 148–157, Jul. 2018, doi: 10.1016/j.bspc.2018.04.013. (Cited in Section 1 and 2.3 to support the hybrid LSTM+CNN model.)
[11] M. S. et al., “Deep learning with convolutional neural networks for EEG decoding and visualization,” Hum. Brain Mapp., vol. 38, no. 11, pp. 5391–5420, Nov. 2017, doi: 10.1002/hbm.23730. (Cited in Section 1 to support CNNs for biomedical signals.)
[12] A. J. Smola and B. Schölkopf, “A tutorial on support vector regression,” Stat. Comput., vol. 14, no. 3, pp. 199–222, Aug. 2004, doi: 10.1023/B:STCO.0000035301.49549.88. (Cited in Section 1, 2.3, 2.4, and 4 to describe SVR and discuss its limitations in chaotic systems.)
[13] C. Schafer and A. Voss, “Short-term heart rate variability—a tool for diagnosing cardiac autonomic dysfunction,” Cardiovasc. Res., vol. 50, no. 2, pp. 211–217, May 2001, doi: 10.1016/S0008-6363(01)00215-0. (Cited in Section 1 to support HRV’s clinical relevance.)
[14] L. Glass and M. C. Mackey, From Clocks to Chaos: The Rhythms of Life. Princeton, NJ, USA: Princeton Univ. Press, 1988. (Cited in Section 1 and 2.1 to explain the chaotic nature of HRV and support the heart model.)
[15] S. H. Strogatz, Nonlinear Dynamics and Chaos: With Applications to Physics, Biology, Chemistry, and Engineering. Boulder, CO, USA: Westview Press, 2015. (Cited in Section 1 and 2.1 to explain the chaotic nature of HRV and support the heart model.)
[16] S. H. Strogatz, “Chaos in biological systems,” in Nonlinear Dynamics and Chaos: With Applications to Physics, Biology, Chemistry, and Engineering, 2nd ed. Boulder, CO, USA: Westview Press, 2015, pp. 287–320. (Cited in Section 2.1 to support the chaotic heart model.)
[17] G. B. Moody and R. G. Mark, “The MIT-BIH Normal Sinus Rhythm Database,” PhysioNet, 2001, doi: 10.13026/C2NK5N. (Cited in Section 2.2 to reference the dataset source.)
[18] F. Takens, “Detecting strange attractors in turbulence,” in Dynamical Systems and Turbulence, Warwick 1980, D. A. Rand and L.-S. Young, Eds. Berlin, Germany: Springer, 1981, pp. 366–381, doi: 10.1007/BFb0091924. (Cited in Section 2.2 to support phase space reconstruction.)
[19] S. Hochreiter and J. Schmidhuber, “Long short-term memory,” Neural Comput., vol. 9, no. 8, pp. 1735–1780, Nov. 1997, doi: 10.1162/neco.1997.9.8.1735. (Cited in Section 2.3 and 2.4 to support LSTM.)
[20] M. H. Khan and M. A. Khan, “Attention-based deep learning for ECG signal classification,” IET Signal Process., vol. 15, no. 6, pp. 401–412, Aug. 2021, doi: 10.1049/sil2.12045. (Cited in Section 4 to suggest future work with attention mechanisms.)
[21] A. R. Hassan and M. I. H. Bhuiyan, “Automated detection of cardiac arrhythmia using deep learning,” IET Signal Process., vol. 14, no. 5, pp. 289–298, Jul. 2020, doi: 10.1049/iet-spr.2019.0357. (Cited in Section 4 to support extending the approach to other physiological signals.)
[22] S. K. Bashar and Y. S. Lee, “Real-time ECG signal analysis for wearable devices using machine learning,” IET Signal Process., vol. 13, no. 3, pp. 312–320, May 2019, doi: 10.1049/iet-spr.2018.5123. (Cited in Section 4 to support real-time applications in wearable devices.)

فایل ها

هم رسانی

ارجاع به این مقاله

آمار