Background of invention

According to the World Health Organization (WHO) and Imperial College London (ICL), the number of adults in age range from 30 to 79 suffering hypertension has increased from 650 million to 1.28 billion since 1990 given the results from the global analysis on hypertension prevalence, detection, treatment and control. Only 1 in 5 women, and lof 4 men have the problem under control, meaning that nearly half of the people are unaware of their elevated blood pressure. The comprehensive study conducted by WHO and ICL, and published in the Lancet, showed that since 1990 the rate of hypertension has decreased in high-income countries, but has increased in many low- or middle-income countries, i.e., nearly 82% of people with hypertension live in low- or middle- income countries. Even though the diagnosis is pretty much straightforward, the research has shown that nearly 580 million people were never diagnosed, and 720 million people were not treated with medicines at all [1]. Hypertension is considered as number one “silent killer”, asymptomatically causing nearly 1000 heart disease- and stroke- related deaths daily worldwide. It is responsible for 7 in 10 first heart attacks, 8 in 10 first strokes, and 7 in 10 cases of chronic heart failures. As it is investigated in a study, the lifetime risk of stroke for men without hypertension at the age of 45 is 17.21% and for men with hypertension is almost double - 32.79% [2],

Besides the health issues, hypertension also affects the budget of the countries. Research investigating the hypertension costs in 15 countries (high-, low-, and middle-income), showed an average total cost of hypertension measured in international dollars (Int$) is 630. 14 Int$ per person. USA is among the most affected with 316 billion Int$ total costs. When calculating the total costs, the researchers consider the costs for health care services, the medications to treat hypertension, and the loss of productivity from premature death [3], New recommendations are needed to help countries improve the management of hypertension, among which is the regular check of the blood pressure systolic and diastolic values and treating the hypertension with medications.

The regular checks of blood pressure systolic and diastolic values can be done both invasively and non-invasively. The invasive measurements are obtained in hospital environments by intra-arterial pressure sensing catheter. The non-invasive measurements can be performed in both ambulatory and home environments by using either brachial cuff-based oscillometric devices, or electronic sphygmomanometers. The non-invasive measurements are performed in discrete time intervals specified by the physician, or the person itself. Such devices require human-action to perform the measurements and cannot be considered for usage when a person is sleeping or is prevented to take the measurements from other factors. Continuous measurements up to our knowledge can still be performed only invasively and in hospital environments.

Given the advancements in the wearable technology, there are some methodologies published in papers and patents, that attempt to predict blood pressure from physiological signals and as such are candidates to be integrated in the state-of-the-art wearable devices, mostly smartwatches and patch-like devices. The mostly used physiological signal is photoplethysmogram (PPG) [4, 5, 6, 7, 8, 9, 10, 11] whose relationship with blood pressure is still not clear in the research literature [12, 13, 14], or a combination of both PPG and ECG [15],

The following patented inventions are known to the applicant of the current patent application: US1051748B2 (Wrist worn accelerometer for pulse transit time (PTT) measurements of blood pressure,

US10786161B1 (Method for collection of blood pressure measurement),

US 10959681B2 (Noninvasive blood pressure measurement and monitoring), and are considered closest to the current patent application at least to an extent known to the applicant.

The above-mentioned devices and methods from the cited patent applications are used for measuring and detection of high blood pressure, detection of hypertension as well as alarming of high blood pressure. They all differ from the invention described in the current patent application since the data for parameters of blood pressure are differently forecasted and are not based on the ECG results.

Up to the knowledge of the applicant there is still no method patented that described the relationship between the blood pressure hypertension and the ECG only, without using additional physiological signals. Therefore, inventing a method that can be integrated in a wearable device, or can reside as a Cloud solution which is able to continuously monitor the blood pressure given only the ECG of a person, anytime, and without human interaction, is still of interest for achieving healthcare continuum. Brief summary of invention

The present invention is directed to a novel method for blood pressure hypertension prediction from ECG. The method is implemented in a novel biomedical wearable device that measures single-lead ECG, extract features, and predicts blood pressure hypertension. The same method is also implemented as a Cloud solution that can be used by any other wearable device or application that is able to transmit ECG signals to the Cloud. The following presents a simplified summary of the embodiments of the invention along with some critical elements needed to provide a basic understanding of the invention which is comprehensively described in the following sections.

In the first embodiment of the present invention, a method based on Artificial Intelligence (Al) is developed that models the relationship between the blood pressure hypertension and the ECG. The methodology is built upon deep analysis of the relationship between the aVR ECG lead and the parallel continuous arterial blood pressure waveforms.

In the second embodiment of the present invention, a wearable device is created that measures ECG augmented limb lead aVR. The hardware solution consists of microcontroller unit (MCU), Analog Front END (AFE) for ECG Applications, notification components, and a battery. The wearable device measures raw ECG signal which is further processed by the microcontroller that integrates the method for blood pressure hypertension prediction described in the first embodiment of the present invention.

In the third embodiment of the present invention, the method for blood pressure hypertension prediction resides as independent method in the Cloud and can be accessed by any wearable device or any application that is able to transmit ECG signals to the Cloud.

Brief description of the drawings

FIG. 1 illustrates the Einthoven triangle and the look from the aVR lead.

FIG. 2 illustrates the placement of the device on the human chest.

FIG. 3 illustrates the scheme of the hardware solution.

FIG. 4 illustrates the methodology flow of the software solution.

FIG. 5 illustrates the alignment between the aVR and the arterial blood pressure (ABP).

FIG. 6 illustrates the cardiac cycle phases.

FIG. 7 illustrates the fiducial points of ECG cycle.

FIG. 8 illustrates the features extracted given the fiducial points of the ECG cycle.

FIG. 9 illustrates the importance of the features for the blood pressure hypertension prediction.

FIG. 10 illustrates the novel method in the Cloud independently from any device. Detailed description of the invention

Considering the ECG power in hypertension diagnosis, even though not a particular lead specified, a summarized comprehensive review of articles confirms promising associations among the P wave, QT, QRS, R wave amplitude, and the T wave, with an elevated systolic and diastolic blood pressure [30], The methodology in this patent is based on the long forgotten aVR lead as this augmented unipolar lead provides an insight into the cavity of the heart from the right shoulder perspective, specifically targeting the right ventricle outflow tract and basal part of the septum [16], As the aVR lead obtains information from the right upper side of the heart, shown in FIG. 1, it provides specific P wave, PR interval, ST segment, and Q wave according to which various diagnosis can be done. aVR is characterized with deepest S wave and as such the ST segment elevation is proven to be a good predictor for left anterior descending coronary artery (LAD) occlusion [17, 18], acute left main coronary artery (LMCA) occlusion [19], stress-induced cardiomyopathy [20] as well as for in-hospital deaths [21], Considering the Q wave in aVR lead, it is found to be related to severe regional wall motion abnormality in the apical and inferior regions [22], The P wave in aVR lead can be used to differentiate atrial tachyarrhythmias [23], The R wave is shown to be useful in predicting Brugada syndrome [24], A PR segment depression in aVR lead have been reported in left-sided pneumothorax [25], The ST segment depression and PR segment elevation in aVR lead are proven to be helpful in a diagnosis of acute pericarditis [26], Given the literature and the research, even though neglected with years, the aVR lead has shown powerful diagnosis capabilities for many coronary diseases [27, 28], including the identification of intra- and interatrial conduction defect that is associated with atrial fibrillation in hypertensive patients [29],

The signal in the solution described in this patent is obtained by proper placement of the electrodes on the person’s body as presented in FIG. 2. The augmented unipolar lead aVR is measured with four electrodes all placed in the frontal plane of the human body:

RA = right arm

LA = left arm

LL = left leg

RL = right leg (neutral, or, ground)

Einthoven triangle in FIG. 1 describes the relations between the limb leads and the electrodes. Each lead is obtained by adding or subtracting voltages from the recording electrodes, i.e., each lead has a specific quantity and direction. The aVR lead of our interest is directed at -150 degrees towards the RA electrode, and is calculated as: aVR = RA - (LA + LL)/2.

The internal architecture of the hardware is presented in FIG. 3. The input is obtained from the four electrodes and transferred to the two-channel low-power Analog Front END (AFE) unit for ECG Applications. The aVR signal is buffered in the MCU unit at which resides the method for hypertension prediction. The aVR portions are taken according to sliding window principle to achieve continuous hypertension predictions. In a case of positive hypertension prediction, the MCU sends control signals to the components that further alert the person in visual, sound and/or vibrational manner. The system is powered by a battery.

FIG. 4 presents the complete methodology comprised of five steps that led to the method for hypertension prediction, the embodiment of the present invention. In the first step a database of aVR signals and aligned arterial blood pressure (ABP) waveforms is established. FIG. 5 presents a sample of an aVR signal and aligned ABP, with marked systole and diastole regions. The blood pressure is defined as the amount of power that the blood is exerting on the artery walls. The systole is considered the period when the blood flows into the aorta, and the diastole is the period when the heart is relaxing, i.e., when the heart expands while receiving blood intro the ventricles. Herefrom, the systolic blood pressure (SBP) is the pressure achieved on the artery wall when the heart beats (ventricular contraction), and the diastolic blood pressure (DBP) is the pressure on the artery walls when the heart relaxes between beats (ventricular relaxation). The mean arterial pressure (MAP) is the average arterial pressure during one cardiac cycle and is calculated as:

MAP = (SBP + 2*DBP)/3

A blood pressure of above 140/80 mmHg (SBP/DBP mmHg) is considered to be hypertension. However, a MAP is contemplated to be a better indicator of perfusion to vital organs than SBP and must be maintained of a minimum 60 mmHg [31], Given the formula for MAP calculation and the hypertension boundaries, an elevation of MAP of above 100 mmHg is considered a hypertension. Our method is further learned to recognize MAP hypertension.

The second step of the methodology is the segmentation of the signals at cardiac cycles levels and observe the aVR segments labeled with the SBP and DBP values derived from the ABP waveform at each corresponding cycle. A cardiac cycle presented in FIG. 6, distinguishes two basic phases, the ventricular systole and the ventricular diastole, or only referred to as systole and diastole. The diastole presents the time when the ventricles are in relaxation mode. The diastole itself also encompasses atrial diastole and atrial systole. Atrial diastole is the time at which the atria are receiving blood from the superior vena cava and inferior vena cava in the right atrium, and from the four pulmonary veins into the left atrium. At that time the blood is passively flowing into the ventricles through the atrioventricular valves. At the end of the atrial diastole, both atria contract pushing additional amount of blood into the ventricles. This process represents the atrial systole. Afterwards the atrioventricular valves close and the ventricles contract to eject the blood through the opened aortic valves into the aorta and pulmonary artery. In a meanwhile, the atria enter the relaxation (diastole) mode. After the ventricle’s systole, the ventricles also enter the relaxation mode and wait for the atria to deliver blood for the next systole. During one cardiac cycle, those phases cause remarkable morphological characteristics on the ECG waveform. The P wave is created when the atrioventricular valves open to let the blood flow from the atria into the ventricles. The interventricular septum is the first area of the ventricular muscle to be activated (from left to right), and this generates the Q wave representing the initial phase of the ventricular systole. The QRS complex represents the spread of a stimulus through the ventricles. And the T wave is generated by the repolarization of the ventricles entering the relaxation mode.

As depicted in FIG. 4 the segmented aVR segments are further preprocessed to delineate the fiducial points from each cycle. The fiducial points locations are depicted in FIG. 7. After fiducial points delineation, next is the features extraction step that defines the features to be analyzed and whose relations to MAP values are going to be explored. The features analyzed are shown in FIG. 8. Two of the features, PP interval and TP segment, are depended on the consecutive cycle. To achieve binary prediction, the MAP values are mapped into 0 (no-hypertension) and 1 (hypertension) class according to the rule MAP >= 100 mmHg as a criterium for hypertension. Having established a database of labeled features we were able to train the Al model and extract the decisive and most important features related to hypertension and test on aVR signals from patients not included in the training dataset.

FIG. 9 presents the most important features according to which hypertension decision is made and those are: PQ interval, PR segment, ST segment and TP segment. Aligned with the systole and diastole phases, it can be easily perceived that those ECG features represent each of the subphases of the whole cardiac cycle. The PQ interval starts at the end of the atrial diastole and lasts until the end of the atrial systole. It represents how fast the action potential is transmitted from the atria to the ventricles. At that time the ventricles are in diastole mode and the arterial pressure decreases as the aorta is not receiving any blood. The PR segment represents the electrical depolarization of the atria. When analyzing segments, we are not interested only in duration, but what is important is the elevation, or the depression. For example, the PR segment elevation occurs in lead aVR in the setting of pericarditis [32] - a condition that is characterized with low blood pressure [33], The PR segment depression might occur as a result from atrial fibrillation that might have emerged from hypertension, as the hypertension is associated with left ventricular hypertrophy, impaired ventricular filling, left atrial enlargement, and slowing of atrial conduction velocity. All those changes are in favor to the development of atrial fibrillation [34], During the PR segment, the ventricles are filled with blood as the atria are in systole mode. The ST segment occurs during the ventricular systole while the atria are in relaxation, diastole mode. At that time the blood flows into the aorta increasing the arterial blood pressure. The ST segment abnormality in the aVR lead and the associated conditions/diseases that led to its elevation or depression are already previously discussed. The relation between the ST segment elevation at aVR and the hypertension is already confirmed in few studies [35, 36], Eventually, the most interesting is the role of the TP segment as it represents the end of the T wave and the beginning of the P wave in the next cycle. During this segment, both the atria and the ventricles are in relaxation mode. There are no specific conditions/disease that change elevate or depress this segment and it usually serves as a baseline for determining the depression or the elevation of the ST segment. Even though its inclination is assumed as not important, our algorithm has shown that both its duration and inclination contribute to the prediction of hypertension.

Hypertension is already proved to be associated with right ventricular morphological and functional abnormalities. Even though studies on the importance of the right ventricle in hypertension are rarely conducted, a clinical study has shown that the right ventricular diastolic dysfunction may be an early clue to hypertensive heart disease [37],

The novel method not necessarily needs to be integrated in the wearable device described in FIG. 2 and FIG. 3. The method can also reside in the Cloud and can be accessed by any wearable or non-wearable device that is able to transmit ECG signals to the Cloud, or by any other application that is able to send ECG signals to the Cloud. FIG. 10 depicts the possible interactions between the novel method and the external devices. As shown in the figure, as soon as the method receives the ECG aVR stream, it preprocesses the signal, extract the features and do the predictions which are returned to the sourcing device or application.

As an independent unit, the method can reside in any technology that collects ECG aVR signals.

References

[1] Zhou, B., Carrillo-Larco, R.M., Danaei, G., Riley, L.M., Paciorek, C.J., Stevens, G.A., Gregg, E.W., Bennett, J.E., Solomon, B., Singleton, R.K. and Sophiea, M.K., 2021. Worldwide trends in hypertension prevalence and progress in treatment and control from 1990 to 2019: a pooled analysis of 1201 population-representative studies with 104 million participants. The Lancet, 398( 10304), pp.957- 980.

[2] Turin, T.C., Okamura, T., Afzal, A.R., Rumana, N., Watanabe, M., Higashiyama, A., Nakao, Y., Nakai, M., Takegami, M., Nishimura, K. and Kokubo, Y., 2016. Hypertension and lifetime risk of stroke. Journal of hypertension, 34(1), pp.116-122.

[3] Wierzejska, E., Giemas, B., Lipiak, A., Karasiewicz, M., Cofta, M. and Staszewski, R., 2020. A global perspective on the costs of hypertension: a systematic review. Archives of Medical Science, 16(f).

[4] Nye, R.; Zhang, Z.; Fang, Q. Continuous non-invasive blood pressure monitoring using photoplethysmography: A review. In Proceedings of the 2015 International Symposium on Bioelectronics and Bioinformatics (ISBB), Beijing, China, 14-17 October 2015; pp. 176-179.

[5] Goli, S.; Jayanthi, T. Cuff less continuous non-invasive blood pressure measurement using pulse transit time measurement. Int. J. Recent Dev. Eng. Technol. 2014, 2, 87.

[6] Choi, Y.; Zhang, Q.; Ko, S. Noninvasive cuffless blood pressure estimation using pulse transit time and Hilbert — Huang transform. Comput. Electr. Eng. 2013, 39, 103-111.

[7] Narasimhan, R., Kimoto, R.C., Sullivan, T.J., Whitehurst, T.K., Young, D.P.S., Zeng, Z. and Klaassen, E., Apple Inc, 2019. Wrist worn accelerometer for pulse transit time (PIT) measurements of blood pressure . U.S. Patent 10,517,489.

[8] Archdeacon, R.P., Bodymatter Inc, 2020. Method for collection of blood pressure measurement. U.S. Patent 10,786,161.

[9] Zengyou, Z., Edan Instruments Inc, 2019. Medical equipment and PTT-based non-invasive blood pressure measurement method and device. CN112022128A, CN Patent.

[10] Guo S., Zhang L., Tian H., Zhao Y., Hebei University of Technology, 2018. Real-time blood pressure monitor system and method based on heart impact signal and photosignal. CN 108186000A, CN Patent.

[11] Colburn, D. and Sia, S.K., Columbia University of New York, 2021. Self-Calibrating, Cuffless, and Non-invasive Blood Pressure Monitor. U.S. Patent Application 17/204,352.

[12] I Caros, J.M.S. Continuous Non-invasive Blood Pressure Estimation. Ph.D. Thesis, Polytechnic University of Catalonia, Barcelona, Spain, 2011.

[13] Wong, M.Y.M.; Poon, C.C.Y.; Zhang, Y.T. An evaluation of the cuffless blood pressure estimation based on pulse transit time technique: A half year study on normotensive subjects. Cardiovasc. Eng. 2009, 9, 32-38.

[14] Payne, R.; Symeonides, C.; Webb, D.; Maxwell, S. Pulse transit time measured from the ECG: An unreliable marker of beat-to-beat blood pressure. J. Appl. Physiol. 2006, 100, 136-141.

[15] Selvaraj, N., Vital Connect Inc, 2021. Noninvasive blood pressure measurement and monitoring. U.S. Patent 10,959,681.

[16] George, A., Arumugham, P.S. and Figueredo, V.M., 2010. aVR-the forgotten lead. Experimental & Clinical Cardiology, 75(2), p.e36.

[17] Engelen, D.J., Gorgels, A.P., Cheriex, E.C., De Muinck, E.D., Oude Ophuis, A.J., Dassen, W.R., Vainer, J., van Ommen, V.G. and Wellens, H.J., 1999. Value of the electrocardiogram in localizing the occlusion site in the left anterior descending coronary artery in acute anterior myocardial infarction. Journal of the American College of Cardiology, 34(2), pp.389-395.

[18] Kotoku, M., Tamura, A., Abe, Y. and Kadota, J., 2009. Determinants of ST-segment level in lead aVR in anterior wall acute myocardial infarction with ST-segment elevation. Journal of Electrocardiology, 42(2), pp.112-117.

[19] Yamaji, H., Iwasaki, K., Kusachi, S., Murakami, T., Hirami, R., Hamamoto, H., Hina, K., Kita, T., Sakakibara, N. and Tsuji, T., 2001. Prediction of acute left main coronary artery obstruction by 12- lead electrocardiography: ST segment elevation in lead aVR with less ST segment elevation in lead VI. Journal of the American College of Cardiology, 38(5), pp.1348-1354.

[20] Rostoff, P., Latacz, P., Piwowarska, W., Konduracka, E., Bolech, A. and Zmudka, K., 2009.

Transient ST-segment elevation in lead aVR associated with tako-tsubo cardiomyopathy . International Journal of Cardiology, 134(f), pp.e97-el00.

[21] Barrabes, J.A., Figueras, J., Moure, C., Cortadellas, J. and Soler-Soler, J., 2003. Prognostic value of lead aVR in patients with a first non-ST-segment elevation acute myocardial infarction. Circulation, 108(7), pp.814-819.

[22] Kotoku, M., Tamura, A., Abe, Y. and Kadota, J., 2010. Significance of a prominent Q wave in lead negative aVR (- aVR) in acute anterior myocardial infarction. Journal of Electrocardiology, 43(f), pp.215-219.

[23] Gorgels, A.P., Engelen, D.J.M. and Wellens, H.J., 2001. Lead aVR, a mostly ignored but very valuable lead in clinical electrocardiography. Journal of the American College of Cardiology, 38(5), pp.1355-1356.

[24] Bigi, M.A.B., Aslani, A. and Shahrzad, S., 2007. aVR sign as a risk factor for life-threatening arrhythmic events in patients with Brugada syndrome. Heart Rhythm, 4(8), pp.1009-1012.

[25] Strizik, B. and Forman, R., 1999. New ECG changes associated with a tension pneumothorax: a case report. Chest, 115(6), pp.1742-1744. [26] Spodick, D.H., 2003. Acute pericarditis: current concepts and practice. Jama, 289(9), pp.1150- 1153.

[27] Tamura, A., 2014. Significance of lead aVR in acute coronary syndrome. World Journal of Cardiology, 6(7), p.630.

[28] Williamson, K., Mattu, A., Plautz, C.U., Binder, A. and Brady, W.J., 2006. Electrocardiographic applications of lead aVR. The American journal of emergency medicine, 24(7), pp.864-874.

[29] Francia, P., Ricotta, A., Balia, C., Adduci, C., Semprini, L., Frattari, A., Modestino, A., Mercanti, F., Sensini, I., Caprinozzi, M. and Tocci, G., 2015. P-wave duration in lead aVR and the risk of atrial fibrillation in hypertension. Annals ofNoninvasive Electrocardiology, 20(2), pp.167-174.

[30] Bird, K., Chan, G., Lu, H., Greeff, H., Allen, J., Abbott, D., Menon, C., Lovell, N.H., Howard, N., Chan, W.S. and Fletcher, R.R., 2020. Assessment of hypertension using clinical electrocardiogram features: a first-ever review . Frontiers in medicine, 7, p.583331.

[31] DeMers, D. and Wachs, D., 2021. Physiology, mean arterial pressure. In StatPearls [Internet], StatPearls Publishing.

[32] Spodick, D.H., 1973. Diagnostic electrocardiographic sequences in acute pericarditis: significance of PR segment and PR vector changes. Circulation, 48(3), pp.575-580.

[33] Rahman, A. and Liu, D., 2011. Pericarditis: Clinical features and management. Australian Family Physician, 40(10), pp.791-796.

[34] Healey, J.S. and Connolly, S.J., 2003. Atrial fibrillation: hypertension as a causative agent, risk factor for complications, and potential therapeutic target. The American journal of cardiology, 91(10), pp.9-14.

[35] Abbase, A.H. and ALjubawii, A.A., 2010. The significance of ST segment elevation in lead aVR in acute anterior myocardial infarction. History.

[36] Adar, A., Onalan, O. and Cakan, F., 2019. Relationship between ST-segment shifts in lead aVR and coronary complexity in patients with acute coronary syndrome. Acta Cardiologica Sinica, 35(1), p.l l.

[37] Akintunde, A.A., Akinwusi, P.O., Familoni, O.B. and Opadijo, O.G., 2010. Effect of systemic hypertension on right ventricular morphology and function: an echocardiographic study: cardiovascular topics. Cardiovascular Journal of Africa, 21(5), pp.252-256.