Textrode-enabled transthoracic electrical bioimpedance measurements – towards wearable applications of impedance cardiography

The biophysical properties of blood, such as a conductivity twice that muscle tissue and several times higher than most other types of tissue [7, 8], produces cardio-related impedance changes that are easily recorded. Important parameters reflecting cardiac activity can be obtained from thoracic electrical bioimpedance measurements such as impedance change, ΔZ, and a first time derivative, dZ/dt. Using the correlation found by Lababidi et al [9] between the ICG waveforms and mechanical events of the cardiac muscle, characteristic points from the ECG and ICG signals can be used to estimate systolic time intervals and other clinically important hemodynamic indices.

ICG technology however, has not reached full acceptance among practitioners for the diagnosis and monitoring of hemodynamic activity in patients with cardiovascular conditions. Under clinical criteria, this technology has not been acknowledged as an established method, primarily because it has presented an unpredictable accuracy in the hemodynamic estimation for patients with severe heart disease or are critically ill [10] and secondly because of controversy about the physiological and anatomic origin of dZ/dt [11].

Nonetheless the United States Department of Health and Human Services (USHHS) endorses the use of TEB for specific cases of the following medical conditions [5]:

A set of two custom-made textile belts for neck and chest have been manufactured with neoprene and synthetic wrap knitted fabric at Swedish School of Textiles at the University of Borås. Four textrodes were embedded in the belts following a lateral spot electrode array [12] as is shown in Fig. 1.B. The textrodes had a contact area each of approximately 13.5 cm² and are made with conductive Velcro with loops made of silver coated fibers. Snap-buttons were used to enable the connection with the ICG measurement instrumentation. The adjustment of the belts and the correct positioning of the textrodes were achieved through Velcro fasteners and the stretchy properties of the belts materials.

Fig. 1

Red Dot repositionable electrodes manufactured by 3M were used in this study. The rectangular electrodes with a contact area of 10.1cm² and snap-button connectors, utilize conductive and adhesive hydro gel to ensure correct contact.

For this study a TEB-measurement one-minute in duration was recorded on three healthy subjects standing in a resting state and following the measurement protocol presented in Fig.2. Using the textrode belt set and conventional Ag/AgCl electrodes, single frequency measurements at 50 kHz were performed with the Respimon impedance cardiograph. The measurements were done using a tetra polar configuration and the lateral spot electrode array as suggested by Woltjer in [12]. The measurement set-up used is represented in Fig. 1A.

Fig. 2

Impedance Cardiograph (Respimon). - Contains an impedance plethysmography device that provides 2 EBI

analog outputs: base impedance Z₀ and impedance change ΔZ from a tetrapolar TEB impedance measurement performed at 50 kHz and a biopotential amplifier that produces and ECG signal.

LabView Data Acquisition Card NI USB-6218. - This card was used to digitize the analogue signals produced by the impedance cardiographer. The card was connected through USB to a PC that recorded and stored the measurements for further processing and analysis with MATLAB.

MATLAB Processing. To remove unwanted noise and offset such as 50 Hz noise and respiration, a digital filtering stage was implemented. Time delay estimation and compensation for the digital filters, signal synchronization, characteristic point detection, parameter estimation and statistical processing were all done in MATLAB.

Analyzed Parameters. – Heart rate estimated from the ICG, ECG, and peak first time derivative dZ/dt_max were some of the parameters analyzed in this study. In addition, a relevant hemodynamic parameter stroke volume, as well as the timing parameters left ventricular ejection time (LVET) and R-Z were compared.

Left Ventricular Ejection Time (LVET). The time elapsed from the opening to the closure of the aortic valve. In Figure 3 correspond to the time between points B and X.

Fig. 3

R to Z time (R-Z). Represents the time span between R-peak in the ECG and the maximal point in the dZ/dt curve. In Figure 3 correspond to the time between R and Z. R-Z time is used in the estimation of the Heather Index [13, 14], considered an index of the cardiac contractility.

Stroke Volume Formula. The estimation of stroke volume was done using the equation 1. In this equation proposed by Bernstein et al [15, 16], V_EPT represents the volume of electrically participating thoracic tissue (mL), dZ(t)/dt_max the peak of first time derivative(ohm/s²), Z₀ the measured transthoracic base impedance (ohms) and LVET the left ventricular ejection time (s).

Performed Comparison. The comparison in this work was based on the signal quality of the ICG and ECG curves and on consistencies in the estimation of relevant parameters. The signals were recorded during one minute and the parameter variation was analyzed by means of mean values and data dispersions.

In Fig. 4 the fundamental ICG signals impedance change ΔZ, dZ/dt, and ECG corresponding to Subject 1 are presented. From these signals no significant dissimilarities can be observed. The R-peak magnitudes acquired with the textile belts seem to be slightly larger, although this could be attributed to variations of the electrode placement with the two different types of electrodes.

Fig. 4

The estimated heart rate values obtained from the impedance and the ECG signals are presented in Fig. 5. In this figure, the HRs obtained from the electrolytic electrode were analyzed and compared against the HRs obtained from the textile belts. From Figure 5A and 5B high degree of similarity and minor differences were observed between the HRs derived from the ECG and ICG signals. The two types of electrode, textile and electrolytic, presented similar distributions with no significant differences in mean values and dispersions for any of the subjects.

Fig. 5

The results of the time interval LVET and the R to Z time are presented on Fig. 6. For the LVET in Fig. 6A, a clear overlap of the estimated values and similar distributions were observed for the three subjects with a maximal mean difference of 20.93 ms for Subject 2. The largest dispersion is observed in the textile-belt for subject 3. In the R to Z period shown in Fig. 6B, similar distributions were obtained with both types of electrodes and for the three subjects. Subject 2 presented the most evident mean difference with almost 9 ms whereas the largest data variance was observed in the Subject 3 when using the textile belt.

Fig. 6

The peak first time derivative dZ(t)/dt_max and the SV estimation are shown in Fig. 7. The SV estimation depicted in Fig. 7A presented overlapping zones in the distribution plots for the three subjects. Subject 3 showed a very close mean value but also the largest dispersion when using textile-belt. On the other hand Subject 2 exhibited the largest mean difference of 8.60 ml. The peak first time derivative dZ(t)/dt_max also presents evident overlapped zones of the distribution plots for all the subjects. In Fig. 7B) it can be observed how the largest mean difference is found in Subject 3 with 0.11 ohms/s² and also the largest data dispersion is observed in the same subject with the textrode belt.

Fig. 7

The ICG waveforms shown in Fig. 4, in general, exhibit good resemblance. Differences in the amplitude of the R-peaks depend more on the exact position of the electrodes than the electrode type. The time shifting observed are caused by the natural heart rate variability rather than by the electrode difference. The small differences observed in the HR analysis extracted from both sources (ECG and ICG) are most probably attributed to the natural HR variability of different measurement sessions and not necessarily to the type of electrode. Measurements obtained with both types of electrodes presented reasonable values of HR according to [17]. There were no missing beats and no need for special signal processing techniques for the signals acquired with belts. Thus the same algorithms for processing and detection were used for both types of electrodes.

Stroke volume is probably one of the most medical relevant hemodynamic indices obtained using thoracic electrical bioimpedance [18, 19]. Despite the clinically attractive attributes of SV estimation by means of TEB, the method has shown low consistency between the impedance wave forms and hemodynamic variables [20]. This inefficiency in the correlation between curves and estimated variables suggest a lack of robustness in the equation models to estimate SV and a low sensitivity and specificity of the impedance signal acquisition in some medical conditions. In addition to this, a necessary consensus on whether the physiological origin of dZ/dt comes predominantly from volume changes as traditionally believed or from blood resistivity changes in the vessels as more recent evidence presented by Bernstein et al. [16, 21] indicates must still be arrived at.

The peak value dZ(t)/dt_max is an important parameter in the estimation of stroke volume and other parameters. The peak first derivative measurements done with the belts presented a reasonable level of consistency. As can be seen from Table I, measurements performed with the textrode belts,

showed a mean variation from 85.9% to 114.8% of the values obtained from the Ag/AgCl reference electrode. Similar behavior was observed for the SV estimation where mean values of the textrode belts shown differences from 96.7% to 112.2% of the values acquired with the reference. No evident trends related with the type of electrode were found in the estimation of these two parameters. However, since measurements with the two types of electrodes were not taken simultaneously but rather sequentially in a different measurement session, variations in the SV estimation may not be entirely attributed to the type of electrode used but instead to the natural biological differences between two measurement sessions. Nonetheless SV measurements obtained with both type of electrodes can be considered as physiologically normal according to [22].

The timing parameters estimated with the textrode belts presented relative close resemblance. The LVET values obtained by the textrode belts and presented in Table I appear reasonable, ranging from 341 to 376 ms. Similar measurements were obtained and reported by Cokkinos et al [23]. Indeed the values obtained with the belt represent 96.2% to 106.3% of the value measured with the corresponding Ag/AgCl electrode.

A similar performance was observed in the estimation of R to Z time where measurements fluctuated from 97.2% to 105.4% of the corresponding value. This difference seems to be very small and although such timing parameters can be used to estimate hemodynamic information, the influence of observed differences on the estimation of such hemodynamic indices, was not analyzed in this study.