Continuous Monitoring of Health and Mobility Indicators in Patients with Cardiovascular Disease: A Review of Recent Technologies

Abstract

Cardiovascular diseases kill 18 million people each year. Currently, a patient's health is assessed only during clinical visits, which are often infrequent and provide little information on the person's health during daily life. Advances in mobile health technologies have allowed for the continuous monitoring of indicators of health and mobility during daily life by wearable and other devices. The ability to obtain such longitudinal, clinically relevant measurements could enhance the prevention, detection and treatment of cardiovascular diseases. This review discusses the advantages and disadvantages of various methods for monitoring patients with cardiovascular disease during daily life using wearable devices. We specifically discuss three distinct monitoring domains: physical activity monitoring, indoor home monitoring and physiological parameter monitoring.

Keywords: activity recognition; biomedical monitoring; cardiovascular disease; electrocardiography; indoor localisation; patient monitoring; photoplethysmography; prognosis and health management; remote monitoring; wearable devices.

Figure 1

Number of publications relating to health monitoring wearables that provide a user with information in the form of recommendations, monitoring and detection. From January 2010 to February 2019. Reprinted with permission from [6].

Figure 2

Signals measured on the y-axis of an accelerometer at 80 Hz in a 5 s window placed on the wrist of a single participant walking at different speeds. Filtered signal is obtained by applying a 4th-order bandpass filter with cut-off frequencies 0.25–2.5 Hz. ‘f’ denotes a forward arm swing. ‘b’ denotes a backward arm swing. Reprinted with permission from [20].

Figure 3

Results from study showing localisation accuracy increasing with increasing number of beacon nodes and decreasing with higher number of obstacles. Reprinted with permission from [59].

Figure 4

Variation in RSSI values from three access points using WiFi technology. Obtained from a person with a device standing in a fixed location. Each access point is at different distance from the user, with the orange, blue and green signals representing access points with increasing distances respectively. Reprinted with permission from [56].

Figure 5

Application of the Kalman Filter (KF) and unscented Kalman Filter (UKF) to reduce noise on RSSI values. Reprinted with permission from [60].

Figure 6

Example of geofencing through RSSI. The beacons in the middle of the rooms are the reference points. The red circle is the maximum radius of the beacon, green circle is the radius of the room (the geofence) and the purple circle is the radius where the devices have been detected.

Figure 7

Example of grid-based fingerprinting, where the map is divided into a grid and the location is estimated to a specific cell.

Figure 8

Location -of-interest-based fingerprinting using 6 areas of interest. Red circles denote the $Rx$. Reprinted with permission from [9].

Figure 9

Illustration of indoor positioning using visible light using LEDs and a photodiode (PD).

Figure 10

IoRL network architecture. Reprinted with permission from [96].

Figure 11

Map of the sub-acute rehabilitation facility: BLE beacon locations are represented by the red circles. Reprinted with permission from [26].

Figure 12

Energy intensity distribution amongst 2 outcome groups in the resident room and the therapy room. Reprinted with permission from [26].

Figure 13

Vesta platform overview. Reprinted with permission from [72].

Figure 14

Fingerprinting training phase to achieve room-level localisation using RSSI at 4 gateways. Reprinted with permission from [72].

Figure 15

Graph comparing the measurements obtained from activity recognition and room localisation during the preoperation, postoperation and follow-up phases. Each patient is represented as a coloured point. Linear regression line demonstrates the patient recovery trajectory. Reprinted with permission from [72].

Figure 16

2-day room localisation data for Patient A in the different phases of their operation. Reprinted with permission from [72].

Figure 17

Detailed information on the health indicators for Patient A during the different phases of their operation. Reprinted with permission from [72].

Figure 18

An ECG waveform, which is informative for the heart health conditions. Details about the PQRST can be found in [99].

Figure 19

Illustration of a participant taking ECGs using an Apple Watch to obtain different leads. (A) I, (B) II, (C) III, (D) V1, (E) V4, (F) V6. Reprinted with permission from [101].

Figure 20

Comparison of 6 leads between the Apple watch (red) and the standard 12-lead ECG (black). Reprinted with permission from [101].

Figure 21

PPG mechanism. The graph at the bottom gives an example of the raw signal obtained from PPG and how it corresponds to the flow of blood in the artery. In the systolic phase, there is less blood volume, so less of the light is absorbed, and hence, it gives a larger signal. Reprinted with permission from [107].

Figure 22

A single-lead ECG and PPG collected at the same time by a smartwatch on the wrist. The PPG waveform peak comes after the ECG waveform peak.

Figure 23

Peak detection algorithm used on a raw PPG signal to calculate HR. Reprinted with permission from [119].

Figure 24

Custom wearable with PPG and barometric pressure sensors providing values for heart rate and altitude climbed over time. An ECG is also acquired (left arm). The right arm uses a consumer device, Fitbit Charge 2, which provides the HR and floor count. Reprinted with permission from [123].

Figure 25

Detection of HR recovery onset and parameter extraction. (a) Altitude as measured by the barometric sensor, (b) detection of the steepest falling slope corresponding to the recovery onset, (c) estimation of HRR parameters. Reprinted with permission from [123].

Figure 26

Examples of exponential fittings. Slower HRR recoveries are displayed on the left column, which yield a higher coefficient of determinant. Faster HRR recoveries are displayed on the right column; there is a higher heart rate variability in the slower recovery phase, which results in a lower coefficient of determinant. Reprinted with permission from [123].

Figure 27

Hazard ratios for CVD-related outcomes based on RHR ranges compared to a reference group of individuals with RHR <65 bpm. Graph produced with data from a study by Woodward et al. [126].

Figure 28

Relationship between HR and VO2 during different daily work tasks. Reprinted with permission from [135].

Figure 29

Snapshot of an ECG recording with the 3 R-R intervals measured and labelled, displaying the variability between each heartbeat.

Figure 30

The variability of blood pressure caused by different factors. Reprinted with permission from [152].

Figure 31

Changes in arm angle over time, measured by an accelerometer. The significantly thinner line after midnight indicates sleep. Reprinted with permission from [162].