Hypertension Assessment via ECG and PPG Signals: An Evaluation Using MIMIC Database

1. Introduction

Hypertension is a major factor of many cardiovascular diseases (CVDs), which are a group of disorders of the heart and blood vessels, including coronary heart disease, cerebrovascular disease, peripheral arterial disease, rheumatic heart disease, etc. [1]. Although sometimes there are symptoms of headache, lack of breath, chest pain, and so on, for most people with hypertension, there are no symptoms at all. Therefore, it is also known as the “silent killer”, and 13% of global death is attributed to it [1]. With each heartbeat, blood is pumped via the contraction of the heart and flows through the whole body following the arterial system. Blood pressure is formed by the main propulsion of the heart’s pumped blood and blockage of the microcirculatory system. Therefore, the higher is the blood pressure, the more difficult it is for the heart to pump. This undoubtedly increases the burden of the heart and, in the long term, will lead to a series of CVDs and damage to the heart, blood vessels, brain, kidneys, and so on.Fortunately, blood pressure is the most important preventable factor of CVDs. Early prevention and management of hypertension are the major and most effective means of improving people’s health levels worldwide. Healthy lifestyles (healthy diet, non-alcohol consumption, non-tobacco use, and physical activity), early detection, evaluation of blood pressure levels, proper diagnosis, and treatment with low-cost medication are beneficial in the prevention and control of hypertension [2]. The seventh report of the Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure (JNC7) [3], which is funded and published by the US National Institutes of Health, is widely adopted. According to this report, different BP levels are divided into different hypertension categories, including normotension, prehypertension, stage 1 hypertension, and stage 2 hypertension. Due to the number of research participants, this study adopted the three BP categories of normotension, prehypertension, and hypertension, labeled according to the BP ranges of the JNC7 report [3].Clearly, earlier attention and treatment are more effective in preventing hypertension and other CVDs. However, as we know, most hypertension patients have no symptoms in the stage of elevated blood pressure and even in hypertension. Thus, many people miss the best time for treatment and experience some complications. However, some physiological signals change based on blood pressure level [4,5], such as electrocardiogram (ECG) and photoplethysmography (PPG). The morphological changes in physiological signals mainly reflect the change of function status of the heart and vascular system. Therefore, the morphological information of PPG could be used to assess hypertension [6]. For this purpose, the Medical Information Mart for Intensive Care (MIMIC) database [7,8] was used to collect the dataset for this study, which involves arterial blood pressure (ABP), ECG and PPG signals.Many researchers have used the MIMIC database assuming that all simultaneously collected signals were synchronized [9,10,11,12,13]. However, the creators of the MIMIC database have reported errors in the data matching and alignment in some recordings, as mentioned by Clifford et al. [14], confirming that not all signals were synchronized. This contradiction motivated our study, and we thought it would be useful to test the synchronicity-dependent features (features that rely on the time interval between ECG and PPG events) and asynchronicity-dependent features (features that rely only on features extracted from PPG events) to gain insights about the usability of the MIMIC database for evaluating hypertension either by using ECG and PPG signals or by using PPG alone.The rest of this paper is organized as follows: Section 2 explains the methods used in this study, including data collection, signal process, and feature extraction. Section 3 shows the comparison results of the different classification models and different feature sets. Finally, Section 4 and Section 5 discuss the results and conclusions on the differences and optimizations of arterial wave propagation theory and PPG morphological theory, respectively.

— Update: 02-01-2023 — cohaitungchi.com found an additional article Assessment of Hypertension Using Clinical Electrocardiogram Features: A First-Ever Review from the website www.ncbi.nlm.nih.gov for the keyword hypertensive changes in ecg.


Hypertension is a major cause of morbidity and mortality (1) and affects an estimated 1.4 billion people worldwide (2). Hypertension is defined as blood pressure (BP) that is elevated above 130 mmHg systolic or above 80 mmHg in stage 1 and above 140 mmHg or above 90 mmHg in stage 2 by the American College of Cardiology and American Heart Association (3). Long term consequences of hypertension can include ischemic heart disease, stroke, and end stage renal disease (1). Cardiac structural changes from prolonged hypertension can lead to arrhythmia and sudden cardiac death (4). Managing BP early, even at prehypertensive levels, may decrease the incidence of cardiovascular events later in life (5). Recognizing and intervening in acute BP emergencies can also decrease mortality (6). Hypertensive conditions in pregnancy, such as pre-existing hypertension, gestational hypertension, preeclampsia, and eclampsia, can lead to perinatal and maternal morbidity and mortality (7). All of these findings highlight the importance of having an effective BP surveillance system.

Prehypertension and primary hypertension can be difficult to identify, as their diagnosis relies on BP readings (3). Conventionally, BP is measured using an auscultatory sphygmomanometer or automated oscillometric devices in health care settings or with oscillometric devices at home (3). Obtaining accurate BP measurements are challenging as it is necessary to coordinate multiple factors with the patient such as ensuring the patient has emptied their bladder, has relaxed for at least 5 min and has avoided smoking, exercise, and caffeine 30 min prior (3). Furthermore, the BP cuff, the BP device type, the measurer’s technique, and the frequency of BP measurements also need to be considered (3). These measurements take time and training for health care providers and are inconvenient for patients who need to visit a doctor’s office or learn to manage their own BP monitoring at home. In addition, errors in BP readings are made by both physicians and patients (8, 9). Even if health care providers and patients meet all the measurement guidelines, the measurement methods only capture a few instantaneous readings of BP for analysis. Further, numerous daily variables can affect BP from moment to moment (3, 8, 10). This is concerning, as even a 5 mmHg difference to the measurement can change a patient’s BP category and management plan (10). Thus, there is an unmet need for an accurate, 24-h continuous and portable BP measurement system.

Read more  Hypertension and patients with acute coronary syndrome: Putting blood pressure levels into perspective

There has been ongoing work to develop alternative, more convenient ways of measuring BP using photoplethysmography (PPG) and electrocardiograms (ECGs or also known as EKGs) (14–18). These methods range from using a single biomedical signal to assess BP (17), to using multiple body sites with PPG input (15), to using PPG along with another type of input such as ECGs (18). However, measuring BP using only ECG signals has not been reviewed. The aim of our review is to identify if there are specific ECG features that could be used alone to measure BP or identify hypertension.

Note that the ECG waveform can provide valuable information about the heart status. The waveform features can be analyzed to determine, for example, heart rate variability and left ventricular hypertrophy. It can also be used to design machine-learning algorithms for other BP studies. Links between hypertension and such biomedical signals may show promise for monitoring BP.

It is hypothesized that elevated heart rates represent persistent sympathetic activation, which leads to increases in BP and cardiovascular complications (19). Heart rate variability (HRV) can be estimated from ECG signals and can reveal whether there is dysregulation of the autonomic system leading to a decrease in the ability to return to lower heart rates (19, 20). HRV may show some promise in monitoring hypertension and poor pregnancy outcomes as HRV has been found to decrease as BP increases (20–24). However, HRV can be measured in different ways and requires its own review outside of this paper.

Left ventricular hypertrophy (LVH) is thought to be an adaptive response to hypertension and is found in a small portion of hypertensive patients at 5–18% using ECG signals (23, 25). It has been used to predict dangerous cardiac arrhythmia and other cardiovascular risks in hypertensive patients (23, 25). Note that LVH is generally a long term complication for elevated BP over a long period of time (4), which may make it less promising finding to identify new onset hypertension.

Some early work has looked at using machine-learning to identify ECG signal features that can potentially categorize individuals as high-risk hypertension (SBP ≥ 130, DBP ≥ 80) or low-risk hypertension (SBP < 130, DBP < 80) (26). Note that authors did not mention the logical operation between the SBP and DBP, in other words it was not clear if it is an “OR” or “AND” between them. ECG signal components may be mapped (or interpreted) into a quantitative number based on machine-learned algorithms and considered for the identification of and monitoring of chronic conditions (i.e., pulmonary artery hypertension, hypertension, coronary artery disease, etc.) (27). Additionally, ECG signal components may be categorized into hypertension (SBP ≥ 140 OR DBP ≥ 90), prehypertension (SBP = 120–139 OR DBP = 80–89) or normal BP (SBP < 120 AND DBP < 80) and then machine learning with this categorization may predict specific BP values for individuals (28). However, there are few studies in this field to draw conclusions from at this time.

This review focuses on analyzing ECG wave morphology without HRV, LVH or machine learning. ECG wave morphology has been a promising area of investigation related to hypertensive individuals. In a review of ECG parameters in the context of hypertension and arrhythmia risk, P wave duration and QT dispersion were noted to have decreased after BP was controlled with medication (23). In addition, T wave amplitudes have been noted to differ between hypertensive patients and normotensive individuals (23) However, there have been no detailed reviews specifically looking at ECG wave morphology differences between hypertensive and normotensive individuals. Interestingly, ECG wave morphology changes can happen over a short period of time, such as from before to after treatment of a hypertensive crisis (29). Therefore, ECGs may provide 24-h continuous opportunity for monitoring of hypertension and for alerting patients and their health care providers to acute conditions, such as hypertensive emergencies or preeclampsia.

In this current review, we conducted a search of literature published from the last decade on the use of only ECG wave morphology to either measure BP or identify hypertension. The aim of this review is to summarize the current understanding of monitoring BP, identifying hypertension using only ECG wave morphology and provide an effective BP surveillance system. We set our ultimate review question to be: can we use only ECG wave morphology to measure BP or identify hypertension in humans?

Read more  Brain Injury: Complications and Medical Problems

— Update: 03-01-2023 — cohaitungchi.com found an additional article The Diagnostic Yield of Routine Electrocardiography in Hypertension and Implications for Care in a Southwestern Nigerian Practice from the website www.dovepress.com for the keyword hypertensive changes in ecg.


Hypertension presents a growing and significant public health challenge, being a primary contributor to heart and kidney diseases, stroke and premature death.1–3 It is estimated that over a billion people in the world have hypertension with two-thirds living in low and middle income countries (LMIC). This figure has been projected to expand further by 29% by the year 2025 with the majority of this increase expected to be driven by the developing nations.3

Yet these LMICs like Nigeria, have limited availability and affordability of diagnostic and therapeutic services.4,5 Hence there is often a pragmatic need for the critical evaluation of routine medical interventions and investigations for their roles in clinical decision making. One of the routine tests recommended in the evaluation of the hypertensive patient is the electrocardiogram (ECG), a simple, noninvasive, largely accessible and relatively inexpensive electrophysiological test that evaluates cardiac function.6 Its low sensitivity for the detection of left ventricular hypertrophy (LVH) has been described,7 yet it remains an important guide for risk stratification in hypertension.

The aim of this study was to determine the diagnostic yield of a routine ECG in subjects with hypertension seen at the cardiac electrophysiology unit at the University of Medical Sciences Teaching Hospital Complex, Akure, over a six-year period.

Patients and Methods

A six-year study of the ECGs performed in the cardiac electrophysiology unit of University of Medical Sciences Teaching Hospital Complex, Akure, Ondo state (formerly State specialist hospital, Akure) was conducted. This unit is the major electrocardiography services provider to a population of about 642,000 people residing in Akure, the capital city of oil and cocoa-rich Ondo state, southwest Nigeria. The unit made use of a General Electric MAC 5500 HD electrocardiography machine.

Anonymized data were extracted from the cardiac electrophysiology records and were collated into a spreadsheet. Demographic variables such as age and gender, referral, indication for ECG and medical diagnoses were retrieved. Electrocardiographic variables included the cardiac rhythm and rate, LVH, left atrial enlargement (LAE), T wave abnormalities, cardiac axis, blocks, complexes, QT segment abnormalities, atrial fibrillation (AF)/flutter, supraventricular tachycardia and pre-excitation syndrome.

A composite score was also derived to primarily assess the presence and severity of ECG abnormalities in this cohort of hypertensive patients and secondarily, to determine the diagnostic yield. Twelve ECG parameters were assigned a score of 0 and 1 for normal and abnormal finding respectively, thus each ECG entry had a minimum score of 0 and a maximum score of 12.

Statistical Analysis


Voltage criteria for LVH was met in only a fifth of ECGs, least prevalent in the young, intermediate in the middle aged and most prevalent in the elderly.

There was also a similar and significant increment in prevalence of electrocardiographic LAE with age. There was a 16.7% prevalence of sinus arrhythmia, significantly higher in females (18.8 females: 13.2% males) but with no significant differences across age groups.

T wave abnormalities were described in 9.2% of tests with inferior T–wave anomalies least prevalent. Atrial and ventricular complexes were observed in 5.5% of the study population and were rare under the age of 45years (1.6%), intermediate in ages 45–65years (3.5%), and commonest in subjects older than 65years (12.2%).

Long and short QT abnormalities accounted for 3.8% and 0.2% respectively in this cohort.

Left anterior fascicular block was present in only 4.8% of all ECGs and was twice as prevalent in males (2.3:1 M/F ratio) and in the elderly (8.4%). First degree AVB and RBBB were reported in 2% of all ECGs while complete heart block was documented in only 1 subject.

Rare electrocardiographic abnormalities included paroxysmal SVT (0.9%), AF (0.7%), atrial flutter (0.2%) and WPW pattern pre-excitation (0.2%). WPW had a male predilection (sex ratio of 4:1) while paroxysmal supraventricular tachycardia, AF and flutter had similar occurrence in both sexes. AF and WPW pattern pre-excitation in individuals older than 65 years were at least eight and four times commoner respectively than in the middle aged and the young.


The majority of the subjects were older than 45 years; a finding that reflects the fact that hypertension increases with advancing age.10 The female predominance in this population may be due to the general better health seeking behavior, better hypertension awareness and early identification of hypertensive disorders in pregnancy during antenatal care, as well as longer life expectancy.11–13

AF was an uncommon finding in this study population unlike what obtains in previous studies in which AF was observed to be the most common arrhythmia found in hypertension.14 AF has been reported to be lower in Africans than people of other regions.15 A study in the USA showed that the prevalence of AF in Africans was notably lower among the African American population compared to Caucasians.16 AF prevalence of 0.7% was found in Kenyans and Tanzanians.17,18 A possibility of genetic factors being responsible for the low prevalence of AF in the African population has been suggested. The population distribution of the elderly accounting for the minority of the cohort studied may also explain the lower incidence of AF in this study, which contrasts with the preponderance of AF among the elderly population in other studies.19

Read more  The Effects of Beetroot Juice on Blood Pressure, Microvascular Function and Large-Vessel Endothelial Function: A Randomized, Double-Blind, Placebo-Controlled Pilot Study in Healthy Older Adults

Axis deviation, in particular LAD, was the most frequent electrocardiographic finding in this study and a wide array of reasons have been posited for its occurrence in many hypertensive subjects. At one end is that LAD may be inconsequential, probably as a result of the left anatomic position of the heart while at the other is that it may be due significant LVH and left bundle branch block.20–22 In this study, male sex and advancing age were determinants of the LAD, an observation that has been similarly established.23

Isolated LAFB was the commonest block observed in this cohort but it has been traditionally considered a benign electrocardiographic finding in individuals without structural heart disease, commoner in males and with advancing age.24,25 However, its existence in the presence of other cardiac pathologies portends a poor prognosis.26

First degree atrioventricular block (AVB) and right bundle branch block (RBBB) were reported in 2% of all ECGs of the study population, while complete heart block was present in only one of the subjects. Researchers have documented an association between various degrees of heart block and hypertension, which was attributed to possible infiltration of the atrioventricular conduction system by fibrosis as a result of high blood pressure.27,28

Atrial and ventricular complexes were quite rare in the young (1.6%) but relatively common in the elderly (12.2%). Supra ventricular and ventricular ectopics are frequent findings in hypertensive subjects.14

=LAE was also shown to be of high prevalence in this cohort of subjects, being highest in subjects older than 65 years. LAE is a notable morphological adaptation found in hypertensive patients that occurs early in hypertension, influenced by age and an independent predictor of cardiovascular events.29–32

LVH was found in 20% of the study population. The sensitivity of ECG LVH is low and echocardiography together with cardiac magnetic resonance imaging (CMRI) have been shown to give better LVH yield. A study found electrocardiographic LVH of 9% in contrast to echocardiographic LVH prevalence of 32% in the same group of patients.33 Yet, electrocardiographic LVH correlates fairly well with increased cardiovascular morbidity and mortality in the hypertensive population and deserves due attention in clinical decision making. In this study the higher proportion of younger population may also contribute to the relatively low prevalence of LVH QT abnormalities, noted in this study has been demonstrated repeatedly in hypertensive patients and its significant contribution to arrhythmias and premature cardiac death has been established.34,35 QT interval is a measure of myocardial depolarization. Cardiomyocyte hypertrophy and consequent increased left ventricular mass with accompanying changes in left ventricular transmural dispersion of depolarization, changes in the tone of the autonomic nervous system of some hypertensive patients, antihypertensive medications and rarely mechano-electrical feedback are some of the possible mechanisms that has been advanced as being responsible for the QT interval alteration observed in hypertensive subjects.36

The only significant ECG abnormality that was not prevalent in the elderly was heart-rate disorder, which was most frequent in subjects younger than 45 years.

T wave abnormalities were described in the ECGs of the study group. T wave abnormalities in hypertensive individuals have reportedly varied widely, from non-specific T wave abnormality to strain T wave abnormality is a marker of LVH and it portends poor prognosis in hypertensive patients.37,38

The diagnostic yield of ECG in the youngest age group was below 10% for most ECG abnormalities with the notable exceptions of LVH (14.5%), abnormal rhythm (18.9%), rate (17.3%) and axis deviation (15%). This may be attributable to shorter duration of hypertension and lower prevalence of hypertensive heart disease in this group.

The increased prevalence of ECG abnormalities in the elderly population may be explained by the higher prevalence and severity of hypertension in the elderly population as established by previous studies.11,39,40

The composite scores show that half of all ECGs performed on a young hypertension subject will show no abnormality in contrast to middle age and the elderly in which two-thirds and three quarters respectively of completed ECGs will have at least one electrocardiographic anomaly.

ECG is an essential investigation in hypertensive patients as ECG markers may predict sudden cardiac death, especially P wave duration, interatrial block, and deep terminal negativity of the P wave in V1, prolonged QT and T peak–T end intervals, QRS duration and between QRS and T vectors, premature ventricular contractions, and ECG hypertrophy criteria.41


This study comprised of a large amount of ECGs and is likely to enhance its robustness and generalizability of its findings.


This was a retrospective descriptive study that explored associations rather than causality. Also the roles of blood pressure control, medications, comorbidities such as renal disease, lifestyle factors such as smoking and lipid panels were not explored due to the retrospective nature of the study and non-inclusion of these important data in electrophysiology records of the institution.


The diagnostic yield of ECG in hypertension is highest in the elderly, intermediate in the middle aged and lowest in the young. These findings are likely to be linked to the duration of hypertension as well as age-related electrocardiographic changes. This increasing burden of ECG abnormalities across the age groups for hypertensive individuals calls for improved awareness of hypertension and vigorous interventions to limit the devastating cardiovascular consequences that may ensue.


Recommended For You

About the Author: Tung Chi