Hearables: feasibility of recording cardiac rhythms from single in-ear locations

Figures

• 

  Figure 1.. ECG volume conductor modelling. (a) Three-dimensional model structure: organs and tissue in the torso and head enclosed within the skin surface. (b) Heart vector: the orientation and magnitude of the current dipole p in px, py and pz directions; the heart vector at one point in time is shown in blue, and the trace of the tip of the heart vector from the start of the cycle until the current position (axes in 0.12 mA m) is shown in orange; the heart muscle is shown in pink in the background (axes in millimetres). The direction of heart vector change is indicated with arrows. (c) Head-/ear-ECG electrode positioning in Experiment A shown on the model: potential differences between electrodes on the upper and lower scalp (scalp-ECG), helix and concha (ear-ECG) and upper and lower neck (neck-ECG) were extracted. Measurements L and L/2 which were used during electrode positioning are also shown. (d) Ear-ECG electrode positioning in Experiment B: potential differences between electrodes on the left helix and left concha (left ear-ECG), right helix and right concha (right ear-ECG) and left helix and right concha (cross-ear-ECG) were extracted. (e) Reference wrist-ECG electrode positioning…

• 

  Figure 2.. Simulated and recorded cardiac electric potentials. (a) Simulated electric potential on the head surface and upper torso at the time of the R-wave peak (in millivolts) in the biophysics model. (b) Recorded ECG traces from the scalp-, ear- and neck-ECG channels from a single subject. The ECG potentials at the instance of the R-peaks in the reference (Lead I) channel are indicated (red circles). (c) Simulated cardiac rhythms in scalp-, ear- and neck-ECG based on volume conduction. (d) Measurements of cardiac rhythms in scalp-, ear- and neck-ECG. Results are based on averaging over a 10 min recording and exhibit clear cardiac rhythms in correspondence with the simulated predictions in (c). For both (c) and (d), ECG potentials at the timings of the P-, Q-, R-, S- and T-waves (from left to right) are circled. The timing of each eave is indicated.

• 

  Figure 3.. Cardiac rhythms from wrist-, scalp-, ear- and neck-ECG of five subjects (a–e). Reference Lead I cardiac rhythms are displayed along the first column. ECG acquired after N = 240 rhythms (blue) and N = 540 rhythms (green). ECG potentials at timings of P-, Q-, R-, S- and T-waves in each channel are circled for the N = 240 cycle ECG. Timings of the waves are indicated next to each wave.

• 

  Figure 4.. Performance metrics for median scalp-, ear- and neck-ECG cardiac rhythms for varying ensemble sizes (number of consecutive cardiac rhythms used during the calculation of the median), N. (a) Pearson correlation coefficient of the cardiac rhythms from a given channel with the grand-median cardiac rhythm from the reference channel. (b) Wave amplitude ratio between the R-peak and P-, Q-, S- and T-peaks from the cardiac rhythms from a given channel, normalized by equivalent ratios from the grand-median cardiac rhythm from the reference channel. (c) Wave timing error for the P-, Q-, S- and T-waves relative to those obtained from the grand-median cardiac rhythm from the reference channel. A vertical dashed line indicates the values at N = 240 rhythms.

• 

  Figure 5.. Simulated and recorded cardiac electric potentials. Simulations of (a) left ear and (b) right ear surfaces on the model at the timing of the R-peak. The outline of the mesh of the model is shown in black to improve the clarity of the shape of the ear surface. (c) Recorded ECG traces from left ear, right ear and cross ear-ECG channels from a single subject. The ECG potentials in each channel at the timings of the R-peaks in the reference (Lead I) channel are circled. (d) Simulated and (e) measured cardiac rhythms in wrist, left ear, right ear and cross ear-ECG. (e) Cardiac rhythms based on averaging over a 60 min recording. ECG potentials as the timings of the P-, Q-, R-, S- and T-waves (from left to right) are circled. The timing of each wave is indicated.

• 

  Figure 6.. Cardiac rhythms from wrist, left ear, right ear and cross ear-ECG of five subjects (a–e). Reference Lead I cardiac rhythms are displayed along the first column. The ECG acquired after N = 240 rhythms (blue) and N = 600 rhythms (green). The ECG potentials at timings of P-, Q-, R-, S- and T-waves in each channel are circled for the N = 240 cycle ECG. Timings of the waves are indicated.

• 

  Figure 7.. Performance metrics for median left-, right- and cross-ear ECG cardiac rhythms for varying values of cardiac rhythm ensemble size (number of consecutive cardiac rhythms used during the calculation of the median), N. (a) Correlation of the cardiac rhythms with the grand-median Lead I cardiac rhythm, (b) RMS amplitude ratio between the R-peak and P-, Q-, S- and T-peaks for a given channel, normalized by the values from Lead I and (c) RMSE of the timings of the P-, Q-, S- and T-waves relative to the R-wave between a given channel and the Lead I channel. A vertical dashed line indicates the values at N = 240, which are also displayed in table 3.