Novel Photoplethysmographic and Electrocardiographic Features for Enhanced Detection of Hypertensive Individuals

Hypertension is a major risk factor for many cardiovascular diseases, which are the leading cause of death worldwide. Regular monitoring is essential to provide early diagnosis since most patients with elevated blood pressure (BP) have asymptomatic hypertension. This work presents a method for BP classification using simultaneous electrocardiographic (ECG), photoplethysmographic (PPG) and BP signals. 86 recordings were used, being 35 normotensive, 26 prehypertensive and 25 hypertensive. It has been proposed 23 novel features to improve the discrimination between healthy and hypertensive individuals based on pulse arrival times (PAT) and morphological features from PPG, VPG and APG signal. Moreover, alternative classification models as Support Vector Machines (SVM), Naive Bayes or Coarse Trees were trained with the defined features to compare the classification performance. The classifier that provided the highest results comparing normotensive with prehypertensive and hypertensive subjects were Coarse Tree, providing an F1 score of 85.44% (Se of 86.27% and Sp of 77.14%). The use of new PPG and ECG features has successfully improved the discrimination between healthy and hypertensive individuals, around 7% of F1 score, so this machine learning methodology would be of high interest to detect HT introducing these techniques in wearable devices.

[1]  Spoorthy Kulkarni Hypertension management in 2030: a kaleidoscopic view , 2020, Journal of Human Hypertension.

[2]  K. Kario Management of Hypertension in the Digital Era , 2020, Hypertension.

[3]  Dorairaj Prabhakaran,et al.  2020 International Society of Hypertension Global Hypertension Practice Guidelines. , 2020, Hypertension.

[4]  Derek Abbott,et al.  How Effective Is Pulse Arrival Time for Evaluating Blood Pressure? Challenges and Recommendations from a Study Using the MIMIC Database , 2019, Journal of clinical medicine.

[5]  Mohamed Elgendi,et al.  Hypertension Assessment via ECG and PPG Signals: An Evaluation Using MIMIC Database , 2018, Diagnostics.

[6]  Rabab Ward,et al.  An optimal filter for short photoplethysmogram signals , 2018, Scientific Data.

[7]  Peter Szolovits,et al.  MIMIC-III, a freely accessible critical care database , 2016, Scientific Data.

[8]  Survi Kyal,et al.  Toward Ubiquitous Blood Pressure Monitoring via Pulse Transit Time: Theory and Practice , 2015, IEEE Transactions on Biomedical Engineering.

[9]  Christophe Ley,et al.  Detecting outliers: Do not use standard deviation around the mean, use absolute deviation around the median , 2013 .

[10]  M. Elgendi On the Analysis of Fingertip Photoplethysmogram Signals , 2012, Current cardiology reviews.

[11]  Ethel M. Frese,et al.  Blood Pressure Measurement Guidelines for Physical Therapists , 2011, Cardiopulmonary physical therapy journal.

[12]  Raúl Alcaraz,et al.  Application of the phasor transform for automatic delineation of single-lead ECG fiducial points , 2010, Physiological measurement.

[13]  Pablo Laguna,et al.  Bioelectrical Signal Processing in Cardiac and Neurological Applications , 2005 .

[14]  R. Vasan,et al.  The progression from hypertension to congestive heart failure. , 1996, JAMA.

[15]  S. Shalev-Shwartz,et al.  Understanding Machine Learning - From Theory to Algorithms , 2014 .

[16]  Jeffrey M. Hausdorff,et al.  Physionet: Components of a New Research Resource for Complex Physiologic Signals". Circu-lation Vol , 2000 .