Combining 24-Hour Continuous Monitoring of Time-Locked Heart Rate, Physical Activity and Gait in Older Adults: Preliminary Findings

Figures

• 

  Figure 1. Illustration of the ANNE device. The system includes a chest-worn sensor that captures ECG and inertial measurement unit (IMU) data and a finger-worn sensor that records photoplethys- mography (PPG) and pulse oximetry data.

• 

  Figure 2. Defining body positions and activities over a 4 h window. (a) Example of accelerometer and gyroscope data, derived activity, and HR over a 4 h window. Accelerometer (vertical in blue, anterior–posterior in red) and gyroscope (green) data were smoothed with a 10-point moving average to highlight trends. (b) Barcode summary of different activities detected over the same 4 h period, with color-coded segments indicating lying (light orange), sitting (light green), standing (light purple), and walking (darker blue). The x-axis shows the actual time of day, rounded to the nearest half hour, for context. (c) Heart rate (light pastel red) measured during the same 4 h window, also smoothed with a 10-point moving average to illustrate fluctuations in beats per minute (bpm). Overall, this visualization demonstrates how raw sensor data (from the accelerometer, gyroscope, and ECG) correlates with inferred postures and activities, providing insights into the subject’s transitions and physiological responses.

• 

  Figure 3. Age–activity correlations. Scatter plots showing the relationship between age and the relative time spent doing four different activities over a 24 h period: (a) lying, (b) sitting, (c) standing, and (d) walking. Each dot represents an individual participant. The dashed lines represent the fit of the linear regression. Pearson correlation coefficients (r) and corresponding p-values are displayed in each subplot. Notably, significant correlations were observed for sitting (positive correlation, r = 0.22, p = 0.030) and walking (negative correlation, r = −0.33, p = 0.001), indicating that older participants tend to spend more time sitting and less time walking. No significant correlations were found for lying (r = −0.08, p = 0.463) or standing (r = −0.07, p = 0.481).

• 

  Figure 4. Average heart rate and standard deviation for the four main activities across all study subjects. The plot shows the mean HR (bpm) for each activity—lying, sitting, standing, and walking—with each represented by different markers and colors. Error bars indicate the standard deviation of the HR values for each activity. The data are aggregated from multiple subjects, with a minimum segment duration of 30 s per activity. The mean HR values are as follows: lying, 69.83 ± 13.72 bpm; sitting, 78.13 ± 14.25 bpm; standing: 80.79 ± 14.33 bpm; and walking, 92.04 ± 14.17 bpm. The Welch’s t-test p-values for consecutive activities are all smaller than 0.001, indicating significant differences between each pair of activities. The average HR increases, as expected, with the vigor of the activity, from lying to walking.

• 

  Figure 5. Heart rate responses to activity transitions in subjects with differing HRRSs. (a) Subject with a moderate-to-high HRRS of 0.58, demonstrating appropriate directional HR changes during transitions. The lying-to-sitting transitions (pastel blue) and sitting-to-standing transitions (pastel orange) show consistent increases in heart rate, as expected with more demanding postures. (b) Subject with a low HRRS of 0.22, exhibiting inconsistent directional HR changes during the same types of transitions. Despite transitions to more demanding postures, their HR sometimes fails to increase appropriately or even decreases, indicating less optimal cardiovascular adaptation. The gray dashed line at time zero represents the moment of transition. This figure illustrates how the HRRS captures the physiological appropriateness of directional HR changes, with higher HRRSs reflecting more consistent increases in HR with more demanding activities.

• 

  Figure 6. Heart rate distributions for different activities and accompanying HRRS distribution. (a) The upper left subplot shows the probability density functions (PDFs) for the HR of the subject with the lowest HRRS of −0.11, where their median HR for walking is lower than that for standing and sitting. (b–g) The next subplots display PDFs for the HR of subjects with progressively higher HRRSs, with increasingly separated distributions seen for different activities, with scores ranging from 0.09 to 0.73. Each subplot includes fitted distributions for HR values smoothed using a kernel density estimation with a bandwidth of 3 bpm. (h) Histogram of the HRRSs’ distribution for all subjects, which appears to follow a normal distribution. This figure demonstrates the variability in HR responses across different activities, with the HRRSs indicating the degree of separation between activity-specific HR distributions.