Loading...

Blood Pressure Estimation from PPG Signal Using Dynamic Time Warping Based Methods

Hajikazem, Helia | 2021

160 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 54953 (05)
  4. University: Sharif University of Technology
  5. Department: Electrical Engineering
  6. Advisor(s): Mohammadzade, Narjes alhoda; Behrozi, Hamid
  7. Abstract:
  8. By continuously measuring blood pressure, we can prevent the irreversible effects of high blood pressure. With the traditional method of using a cuff, it is not possible to measure blood pressure continuously during the day, so for continuous monitoring of blood pressure, it is necessary to use a method without the need for a cuff. Based on previous studies, to estimate blood pressure, Photoplethysmogram and ECG signal features, or temporal and morphological features of Photoplethysmogram signal have been used. In methods that use ECG signals, signal recording is difficult, and methods that use both PPG and ECG signals are even more complex. Using only PPG signals also has its problems. Since the morphology of PPG signals varies from person to person, accurate identification of the key points in these signals is challenging. There are many signals in which the key points are not recognizable at all by the current methods. This thesis tries to solve this problem. In this research, we have presented an automatic algorithm for finding the key points of PPG pulses using a dynamic time warping method. After detecting the key points of PPG pulses, effective features are extracted, and the blood pressure is estimated using different machine learning algorithms. Finally, using the features extracted by the proposed method and using a random forest estimator on MIMIC II dataset which contains PPG signals from 9000 people, the mean absolute error and standard deviation error are obtained 5.8 mmHg and 6.4 mmHg for systolic pressure and 3.22 mmHg and 3.92 mmHg for diastolic pressure respectively, which has 50% improvement compared to the competitor state of the art methods
  9. Keywords:
  10. Dynamic Time Warping ; Photoplethysmography ; Machine Learning ; Blood Pressure Estimation ; Electrocardiogram

 Digital Object List

 Bookmark

No TOC