Maxim Blood Pressure Trending (MaximBPT) Algorithm Introduction and Performance Metrics
An optical sensor-based, cuffless blood pressure system is introduced in this application note. The most significant difference in the indications for the MaximBPT as compared to traditional noninvasive blood pressure (NIBP) devices is that MaximBPT requires calibration and is primarily designed to provide the user with trends in blood pressure, where successive estimations are relative to this calibration point. This application note provides a general description of blood pressure, along with the MaximBPT system, algorithms, and its operating principles. The protocols for calibration and measurement are specific to MaximBPT, yet the basic principles of clinical trials, reference measurements, and evaluation methodology are based on industry standards.
Blood pressure (BP) is a vital sign that provides critical information about cardiac output, cardiac pressure, and hemodynamic stability.
BP correlates well with the risk of advanced cardiovascular disease (ACVD). (Chockalingam) Also, hypertension, a medical condition in which systolic BP is above 140mmHg and/or diastolic BP is above 90mmHg persistently, is a risk factor for coronary heart disease and the single most important risk factor for stroke. Systolic BP is the maximum pressure in the arteries when the heart contracts. Diastolic BP is the minimum pressure in the arteries between the heart's contractions.
Therefore, continuously monitoring BP is valuable. However, the conventional noninvasive method—a cuff sphygmomanometer—usually allows only one reading every few minutes. Moreover, the inflation process disturbs the user and makes BP measurement inconvenient during sleep or exercise.
Maxim Blood Pressure Trending System
Maxim Blood Pressure Trending (MaximBPT) is a BP measurement system using a photoplethysmography (PPG) signal obtained from the fingertip. The system requires user-specific calibrations, and successive measurements are relative to this calibration point. The goal of the system is to make BP measurements readily available to the user through widely available consumer devices like smartphones or wellness watches. The system is most useful and accurate when measurements are taken in a resting state, as in traditional cuff-based BP systems. This application note contains evaluation methods and design specifications for the MaximBPT system.
In principle, changes in PPG shape are correlated to BP changes. The PPG pulse has two phases: the anacrotic phase is the rising edge of the pulse, and the catacrotic phase is the falling edge of the pulse (Figure 1). The first phase is associated with systole, and the second phase with diastole and wave reflections from the periphery. Some of the determinants of the PPG shape have been described in literature, such as systolic amplitude, pulse width, pulse area, peak-to-peak interval, pulse interval, and artery stiffness index. (Elgendi; Shin)
Figure 1. Waveform of the PPG and its characteristic parameters.
The MaximBPT system requires a PPG signal from the sensor and associated BP values from a traditional cuff-based BP device during calibration. Subsequent systolic and diastolic BP changes are calculated by PPG shape changes relative to the calibration instance.The MaximBPT system requires a PPG signal from the sensor and associated BP values from a traditional cuff-based BP device during calibration. Subsequent systolic and diastolic BP changes are calculated by PPG shape changes relative to the calibration instance.
Figure 2 depicts the main components of the MaximBPT system. It utilizes a Maxim wellness sensor, which can be adapted in mobile devices or wearables for wellness monitoring, and a MaximBPT software application (app) that uses the PPG signal of the wellness sensor.
Figure 2. Overview of the MaximBPT system.
Calibration is a critical step in which BP reference measurements are taken to calibrate the algorithm with the associated PPG signals. One of the following options is recommended to provide the BP reference values:
These calibration reference values are valid for 4 weeks, and thereafter recalibration is required to expect the accuracy figures to hold. The calibration protocol has three main sections:
Before calibration, the subject should:
During calibration, the subject should rest quietly for about five minutes in a seated position where subject's elbow rests on the table, slightly below the level of the heart. The palm should be turned upward, with the hand remaining open and relaxed. The arms, back, and feet should be supported and in a relaxed position. The legs should remain uncrossed.
BP Reference Measurement
The user will first create a unique user ID. MaximBPT calibration is required before the first use of the MaximBPT, and thereafter when the calibration expires. The app's graphical user interface (GUI) for calibration is depicted in Figure 3.
Figure 3. MaximBPT GUI and relevant instructions during the calibration procedure.
Once the user has created an ID or selected an existing ID, the user takes BP measurements with the reference device and records the systolic and diastolic BP measurements (in mmHg) within the MaximBPT application (Figure 3). It is important that the user follows the reference device instructions to ensure an accurate measurement.
PPG Data Collection by the MaximBPT Application
Once the user has collected the reference values, the user follows the MaximBPT application instructions for calibration.
After pressing the Measure BP button on the screen (Figure 3), the user places his or her finger onto the PPG sensor. The MaximBPT application collects the necessary BP values. Once the calibration PPG data collection is complete, the application returns the signal value success (Figure 4, left screenshot). The process takes approximately one minute.
If the calibration is not completed within a minute, the application will return a signal value of failure (Figure 4, right screenshot). Failure might occur if the PPG signal quality is low (due to low perfusion or motion artifacts).
Figure 4. Calibration progress state updates on the Android® application.
Figure 5. MaximBPT GUI and relevant instructions during regular measurement.
During PPG data collection (capture of user measurements), the collection progress will be displayed on the GUI (Figure 6). After PPG data collection, BP trending values for systolic and diastolic BP will be reported in the table and in graphical format, and the signal status will either change to Success or Failure as described in the PPG Data Collection by the MaximBPT Application section.
The user can observe the BP measurements per every heartbeat in the GUI. The display screen also provides the statistical values of the current session only, such as minimum/maximum BP values.
After the measurement is complete, the results will be added in the history log. When the BP HISTORY button on the GUI Welcome screen (Figure 5) is pressed, it will take the user to the history log page that lists the calibration BP, measurements and trending values, and time stamps (Figure 7).
Figure 6. GUI reporting the BP trending values.
Figure 7. Example of a BP HISTORY screen in the MaximBPT application.
The Association for the Advancement of Medical Instrumentation (AAMI) accuracy requirements (ANSI/AAMI/ISO 81060-2) are referenced to evaluate the algorithm. Criterion 1 of this standard requires a BP measurement such that the mean of the estimation error (MERR) must be within or equal to ±5.0mmHg, and that the standard deviation of the error (STDERR) must be no greater than 8.0mmHg, both for systolic and diastolic BP measurements.
where n is the number of determinations, and PDUT and PREF are the BP determination by the device under test and the reference device, respectively.
These requirements will be the accuracy target for MaximBPT within the calibration hold period, which is defined as 4 weeks.
The evaluation dataset is based on a data collection that was performed in an independent medical device clinical trial laboratory, and it is compliant with the normative references mentioned in ANSI/AAMI/ISO 81060-2 standards document. The reference device is a sphygmomanometer with a dual auscultator, and measurements were taken by two medical professionals concurrently. PPG data and reference measurements were collected from 136 subjects. Twelve subjects were selected for evaluation, and their data was not used for algorithm development.
MaximBPT accuracy was evaluated by running the MaximBPT executable file offline with the PPG samples from these subjects. The evaluation dataset subjects were selected such that age, gender, and BP levels are similar to the datasets left for model development. Evaluation results are summarized in Table 1.
Table 1. Accuracy Table for the Evaluation Dataset
This table depicts the mean (MERR) and standard deviation (STDERR) of the estimation error for post-calibration (a few minutes after calibration, 28 samples), post-induction (in a resting state, a few minutes after BP perturbations, 72 samples), and rest-to-rest (in a resting state, 3 to 4 weeks post-calibration, 216 samples) protocols. Each sample is a difference vector of two data packets, one from the calibration protocol and one from the test protocol.
Note that the Table 1 accuracy table is based on the most recent algorithm version and is subject to change.
Bland-Altman is a graphical method to compare two quantitative measurements. Scatter plots of the differences between the test device and reference measurement vs. the actual blood pressure difference relative to calibration values provides a good visualization for blood pressure estimates and error bounds. It is commonly used for BP evaluation, as depicted in Figure 8.
Figure 8. Scatter plot of measurement error vs. true BP change for MaximBPT evaluation in rest-to-rest (week-3/4) protocol is given.
Table 2 summarizes the major system specifications such as BP or heart rate bands, device- and sensor-related parameter intervals where the algorithm reports valid estimations; and use-case details. These measurement and environmental factors, as well as hardware and software specifications, should be taken into account during system evaluation.
Table 2. System Requirements and Specifications
Factors That Can Affect BP Readings
The following factors alter BP readings significantly for cuff-based traditional BP devices and/or Maxim's PPG-based BP measurement system.
Blood Pressure Cuff Is Too Small
It is important to make sure that the proper size BP cuff is used on the subject's upper arm. In fact, many measurement errors occur because of not taking the time to determine if the patient's arm circumference falls within the range of indicators on the cuff. Studies have shown that using a too small of a cuff can cause a patient's systolic BP measurement to increase up to 10mmHg. (Maxwell)
Blood Pressure Cuff Used over Clothing
During a measurement, the cuff should always be placed directly on the subject's arm. Studies have shown that measurement over clothing or with tight clothing pushed up can cause significant errors in the BP measurement. (Reeves; Handler)
Subject Is Not in a Resting State
To obtain an accurate measurement, it is important that the subject relaxes quietly in a comfortable chair for at least 5 minutes before a reading is taken. Activities such as exercising, talking, drinking, or eating can affect the systolic BP by 10mmHg to 20mmHg. (Zheng)
During the measurement, the subject should always be seated in a comfortable chair, legs uncrossed, with their back and arm supported. If the subject's back is not supported, the diastolic measurement may be increased by 5mmHg to 10mmHg. Crossing the legs has shown to raise the systolic BP by 2mmHg to 8mmHg. The positioning of the upper arm below the heart level will also result in higher measurement results, whereas positioning the upper arm above the heart level will result in lower BP values. These differences can increase/decrease the systolic BP 2mmHg for every inch above/below the heart level. (Perloff; O'Brien; Handler)
Stress or anxiety can cause large elevations in BP. If the BP measurement is taken while the subject is thinking about something that causes tension or stress, the BP levels could significantly increase. (Perloff)
Consumption of tobacco products (cigarettes, cigars, smokeless tobacco), alcohol, or caffeine (sodas, coffee, tea) causes BP levels to spike. Therefore, the subject should refrain from smoking/alcohol/caffeine intake at least 30 minutes before a measurement is taken. (James; Marmot; Groppelli)
BP tends to increase when the temperature is low in the measurement environment. (Alperovitch; Barnett)
BP is lower when the bladder is empty. As the bladder gradually fills, BP increases. Studies have shown that systolic BP measurements can increase 10mmHg to 15mmHg when the subject has a full bladder. (Scultéty)
Chockalingam, Arun; Campbell, Norman R.; and J. George Fodor. "Worldwide epidemic of hypertension." The Canadian Journal of Cardiology. vol. 22, no. 7, 2006, p. 553.
Elgendi, Mohamed. "On the Analysis of Fingertip Photoplethysmogram Signals." Current Cardiology Reviews. vol. 8, no. 1, 2012, pp. 14–25.
Shin, Hangsik and Se Dong Min. "Feasibility study for the non-invasive blood pressure estimation based on PPG morphology: normotensive subject study." Biomedical Engineering Online. vol. 16, no. 1, 2017, p. 10.
Maxwell, Morton H.; Schroth, Philip C.; Waks, Abraham U.; Karam, Maroun; and Leslie P. Dornfeld. "Error in blood-pressure measurement due to incorrect cuff size in obese patients." The Lancet. vol. 320, no. 8288, 1982, pp. 33–36.
Handler, Joel. "The importance of accurate blood pressure measurement." The Permanente Journal. vol. 13, no. 3, 2009, p. 51.
Reeves, Richard A. "Does this patient have hypertension? How to measure blood pressure." JAMA. 1995, vol. 273, no. 15, pp. 1211–8.
Zheng, Dingchang; Giovannini, Roberto; and Alan Murray. "Effect of respiration, talking and small body movements on blood pressure measurement." Journal of Human Hypertension. vol. 26, no. 7, 2012, p. 458.
Perloff, Dorothee; Grim, Carlene; Flack, John; Frohlich, Edward D; Hill, Martha; McDonald, Mary; and Bruce Z. Morgenstern. "Human Blood Pressure Determination by Sphygmomanometry." Circulation. vol. 88, no. 5, 1993, pp. 2460–2470.
O'Brien, Eoin, et al. "Practice guidelines of the European Society of Hypertension for clinic, ambulatory and self blood pressure measurement." Journal of Hypertension. vol. 23, no. 4, 2005, pp. 697–701.
James, Jack E. "Critical review of dietary caffeine and blood pressure: a relationship that should be taken more seriously." Psychosomatic Medicine. vol. 66, no. 1, 2004, pp. 63–71.
Marmot, Michael G., et al. "Alcohol and blood pressure: the INTERSALT study." BMJ Clinical Research. vol. 308, no. 6939, 1994, pp. 1263–1267.
Groppelli, Antonella; Giorgi, Dante M.; Omboni, Stefano; Parati, Gianfranco; and Giuseppe Mancia. "Persistent blood pressure increase induced by heavy smoking." Journal of Hypertension. vol. 10, no. 5, 1992, pp. 495–499.
Alpérovitch, Annick; Lacombe, Jean-Marc; Hanon, Olivier; Dartigues, Jean François; Ritchie, Karen; Ducimetière, Pierre; and Christophe Tzourio. "Relationship between blood pressure and outdoor temperature in a large sample of elderly individuals: The Three-City study." Archives of Internal Medicine. vol. 169, no. 1, 2009, pp. 75–80.
Barnett, Adrian; Sans, Susana; Salomaa, Veikko; Kuulasmaa, Kari; and Annette J. Dobson (WHO Monica Project). "The effect of temperature on systolic blood pressure." Blood Pressure Monitoring. vol. 12, no. 3, 2007, pp. 195–203.
Scultety, S.; Varga, B.; and D. Szabo. "Effect of bladder distension on blood pressure." International Urology and Nephrology. vol. 3, no. 1, 1971, pp. 11–19.