This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in JMIR Medical Informatics, is properly cited. The complete bibliographic information, a link to the original publication on http://medinform.jmir.org/, as well as this copyright and license information must be included.
Accumulation of excess body fluid and autonomic dysregulation are clinically important characteristics of acute decompensated heart failure. We hypothesized that transthoracic bioimpedance, a noninvasive, simple method for measuring fluid retention in lungs, and heart rate variability, an assessment of autonomic function, can be used for detection of fluid accumulation in patients with acute decompensated heart failure.
We aimed to evaluate the performance of transthoracic bioimpedance and heart rate variability parameters obtained using a fluid accumulation vest with carbon black–polydimethylsiloxane dry electrodes in a prospective clinical study (System for Heart Failure Identification Using an External Lung Fluid Device; SHIELD).
We computed 15 parameters: 8 were calculated from the model to fit Cole-Cole plots from transthoracic bioimpedance measurements (extracellular, intracellular, intracellular-extracellular difference, and intracellular-extracellular parallel circuit resistances as well as fitting error, resonance frequency, tissue heterogeneity, and cellular membrane capacitance), and 7 were based on linear (mean heart rate, low-frequency components of heart rate variability, high-frequency components of heart rate variability, normalized low-frequency components of heart rate variability, normalized high-frequency components of heart rate variability) and nonlinear (principal dynamic mode index of sympathetic function, and principal dynamic mode index of parasympathetic function) analysis of heart rate variability. We compared the values of these parameters between 3 participant data sets: control (n=32, patients who did not have heart failure), baseline (n=23, patients with acute decompensated heart failure taken at the time of admittance to the hospital), and discharge (n=17, patients with acute decompensated heart failure taken at the time of discharge from hospital). We used several machine learning approaches to classify participants with fluid accumulation (baseline) and without fluid accumulation (control and discharge), termed
Among the 15 parameters, 3 transthoracic bioimpedance (extracellular resistance, R0; difference in extracellular-intracellular resistance, R0 – R∞, and tissue heterogeneity, α) and 3 heart rate variability (high-frequency, normalized low-frequency, and normalized high-frequency components) parameters were found to be the most discriminatory between groups (patients with and patients without heart failure). R0 and R0 – R∞ had significantly lower values for patients with heart failure than for those without heart failure (R0:
This is useful, for instance, for an in-home diagnostic wearable to detect fluid accumulation. Results suggest that fluid accumulation, and subsequently acute decompensated heart failure detection, could be performed using transthoracic bioimpedance and heart rate variability measurements acquired with a wearable vest.
Heart failure is estimated to affect more than 25 million people worldwide and over 6 million people in the United States [
Transthoracic bioimpedance can measure intrathoracic volume, a surrogate biomarker of pulmonary edema [
Traditionally, various types of electrodes have been used for transthoracic bioimpedance and ECG measurements using fluid accumulation vests: adhesive Ag-AgCl electrodes, which often result in skin irritation and are often misaligned when positioned; textile electrodes, which need to be wetted prior to every use; and recently proposed reusable carbon black–polydimethylsiloxane (PDMS) dry electrodes [
There are several studies [
In this prospective clinical study (System for Heart Failure Identification using an External Lung Fluid Device; SHIELD) to examine the performance of transthoracic bioimpedance and heart rate variability measured using carbon black–PDMS electrodes embedded in fluid accumulation vests for detection of acute decompensated heart failure, we hypothesized that (1) participants without acute decompensated heart failure should have resistance measurements that are higher than those of participants with acute decompensated heart failure at the time of admittance to the hospital; (2) participants with acute decompensated heart failure at the time of discharge from hospital should have smaller amount of accumulated lung fluid and therefore higher resistance measurements than those of participants with acute decompensated heart failure at the time of admission; and (3) autonomic function assessed by heart rate variability would provide additional information about the dysregulation of heart failure patients, hence, it would detect acute decompensated heart failure.
A total of 93 hospitalized individuals were prospectively enrolled in our observational study at the University of Massachusetts Medical Center. We acquired recordings from participants with acute decompensated heart failure taken within the first few hours of hospital arrival (baseline) and taken prior to discharge from hospital (discharge). We also acquired recordings from a group of patients without acute decompensated heart failure (control). All participants gave written informed consent before participating in the study, in accordance with the Declaration of Helsinki. The protocol was approved by the institutional review board of the University of Massachusetts Memorial Hospital (docket number H00014714).
The CONSORT diagram in
For this study, we used Philips prototype fluid accumulation vests [
CONSORT diagram. AF: atrial fibrillation; ESRD: end-stage renal disease; HF: heart failure; ICD: implantable cardioverter-defibrillator.
Transthoracic bioimpedance is a noninvasive method that measures the impedance of the tissue at a series of frequencies. A small alternating current, typically ranging from 100 µA to 10 mA, is injected into the tissue via electrodes, while the voltage drop is measured as the output. By applying Ohm’s law, the resistance of the body tissue can be calculated. Biological tissue is typically modeled with a resistance
where
If we measure impedance for frequencies between these two extreme cases, we obtain an arc-like Cole-Cole plot in the impedance plane [
The parameters of the model can be extrapolated from a set of measurements made at a predefined set of frequencies. The exponent α represents the heterogeneity of the tissue in the model. For each frequency, the real (resistance) and imaginary (reactance) part of the electrical impedance is estimated. The Taubin algorithm [
The sum of the square error is minimized in the fitting process. The
Illustrative example of the Cole-Cole plot of one patient.
To compute heart rate variability parameters, 4 minutes of clean ECG data were extracted from each 5-minute recording of ECG acquired simultaneously with transthoracic bioimpedance measurements. Noise and motion artifacts were removed from the ECG signals using a bandpass filter (0.05 Hz-40 Hz). The R peaks were detected using a validated algorithm [
We computed the indices of low frequencies of heart rate variability (low-frequency components of heart rate variability: 0.045 Hz to 0.15 Hz), high frequencies of heart rate variability (high-frequency components of heart rate variability: 0.15 Hz to 0.4 Hz), and the indices normalized to the total power of heart rate variability (normalized low-frequency components of heart rate variability, normalized high-frequency components of heart rate variability) [
Transthoracic bioimpedance and heart rate variability parameters computed in this study.
Parameter | Description | |
|
|
|
|
|
Model resistance of biological tissue—extracellular fluid or resistance when |
|
Model resistance of biological tissue—intracellular fluid | |
|
Resistance of biological tissue when |
|
|
Range of |
|
|
Characteristic frequency, ie, frequency corresponding to the upper point of Cole-Cole plot circle | |
|
Cell membrane capacitance | |
|
Exponent of the model representing tissue heterogeneity | |
Fitting error | Sum of squared error of the optimal Cole-Cole plot model | |
|
|
|
|
LFa HRVb | Low-frequency components of heart rate variability power |
Normalized LF HRV | Normalized low-frequency components of heart rate variability power | |
HFc HRV | High-frequency components of heart rate variability power | |
Normalized HF HRV | Normalized high-frequency components of heart rate variability power | |
PDMI sympatheticd | Sympathetic function heart rate variability dynamics | |
PDMI parasympathetice | Parasympathetic function heart rate variability dynamics |
aLF: low-frequency.
bHRV: heart rate variability.
cHF: high-frequency.
dPDMI sympathetic: principal dynamic mode index of sympathetic function.
ePDMI parasympathetic: principal dynamic mode index of parasympathetic function.
The normality of each parameter was tested using the Kolmogorov-Smirnov test [
Statistical analysis of the differences between groups provides insight into the suitability of the measures of transthoracic bioimpedance and heart rate variability to detect fluid accumulation, which is used as an indication of heart failure exacerbation. However, measurement results have nonlinear characteristics and cannot be completely described with linear statistical methods. Hence, we used nonlinear methods such as machine learning to examine 15 features derived from transthoracic bioimpedance and heart rate variability for classification between groups (control, baseline, and discharge). Furthermore, participants in the discharge group were partially recovered, so they could be considered similar to the participants in control group. We tested the feasibility of classifying participants without fluid accumulation in the lung, termed
For these classification analyses, 3 algorithms were used: support vector machines [
We approached 90 patients with heart failure who were eligible, and 43 were enrolled in this study. Out of the 43 enrolled participants, we were able to collect data from 28 participants with heart failure; 23 were included in the baseline group (mean 72, SD 10.7 years), and 17 were included in the discharge group (mean 72.4, SD 9 years). Only 12 participants were included in both baseline and discharge groups. There were several reasons for the lower number of participants in the discharge group: (1) in some cases, the recordings were of poor quality (n=14); (2) some participants (n=5) were lost to follow-up (ie, owing to a late night or weekend discharge); (3) some participants (n=7) could not provide the second recording owing to illness or refusal.
We enrolled 50 participants without acute decompensated heart failure (mean 71.5, SD 8.5 years) in the control group. Of the recordings taken on the 50 enrolled participants 32 of them were usable. It should be noted that participants from both groups were well matched with respect to age.
The demographic and medical characteristics of study participants are shown in
Demographic and clinical characteristics.
Characteristic | Control (n=32) | Acute decompensated heart failure (n=28) | |||||
Age, mean (SD) | 71.5 (8.5) | 72.4 (10.3) | .70 | ||||
|
|
|
|
||||
|
Male | 19 (59) | 18 (64) | .70 | |||
|
Female | 13 (41) | 10 (36) |
|
|||
|
|
|
.52 | ||||
|
White | 29 (91) | 26 (93) |
|
|||
|
Black | 1 (3) | 2 (7) |
|
|||
|
Othera | 2 (6) | 0 (0) |
|
|||
Chest circumference (cm), mean (SD) | 105.4 (14.1) | 107.8 (13.1) | .57 | ||||
|
27.7 (5.1) | 29.3 (6.6) | .28 | ||||
|
|
|
|
||||
|
Myocardial infarction | 3 (9) | 9 (32) | .03 | |||
|
Coronary artery disease | 7 (22) | 13 (46) | .04 | |||
|
Hypertension | 20 (63) | 23 (82) | .09 | |||
|
Stroke/transient ischemic attack | 2 (6) | 3 (11) | .50 | |||
|
Previous diagnosis of heart failure | 1 (3) | 17 (61) | <.001 | |||
|
Diabetes | 6 (19) | 7 (25) | .56 | |||
|
Dyslipidemia | 23 (72) | 20 (71) | .97 | |||
|
Chronic lung disease | 4 (13) | 9 (32) | .06 | |||
|
Renal failure | 2 (6) | 3 (11) | .53 | |||
|
Atrial fibrillation | 0 (0) | 13 (46) | <.001 | |||
|
|
|
|
||||
|
Heart rate (beats/min) | 75.4 (13.2) | 84.4 (25.1) | .09 | |||
|
Systolic blood pressure | 141.1 (28.6) | 146.1 (28.7) | .51 | |||
|
Diastolic blood pressure | 79.9 (13.1) | 81.3 (17.3) | .72 | |||
|
Respiratory rate (breaths/min) | 18.3 (2.2) | 20.7 (2.8) | <.001 | |||
|
Sodium (mg/dL) | 138.8 (2.4) | 138.9 (2.8) | .97 | |||
|
Potassium (mg/dL) | 4.1 (0.4) | 4.1 (0.8) | .79 | |||
|
Glucose (mg/dL) | 121.6 (45.4) | 143.5 (80.9) | .20 | |||
|
Blood urea nitrogen (mg/dL) | 19.2 (6.7) | 26.3 (18.9) | .06 | |||
|
Creatinine (mg/dL) | 1.1 (0.4) | 1.3 (0.6) | .15 | |||
|
B-type natriuretic peptideb | 112.0 (76.2) | 1013.9 (1004.5) | .14 | |||
|
Troponinb | 0.2 (1.0) | 0.2 (0.9) | .96 | |||
|
INR | 1.3 (0.7) | 1.4 (0.5) | .95 | |||
|
|
|
|
||||
|
Beta blocker | 2 (6) | 2 (7) | .89 | |||
|
Angiotensin converting enzyme inhibitor | 5 (16) | 1 (4) | .12 | |||
|
Diuretic | 2 (6) | 3 (11) | .53 | |||
|
Statin | 6 (19) | 3 (11) | .38 | |||
|
Oral anticoagulant | 2 (6) | 0 (0) | .18 |
aAsian; American Indian, or Alaska Native; Native Hawaiian or other Pacific Islander.
bData for the control group is for 6 patients only.
We compared values of 15 parameters from transthoracic bioimpedance and heart rate variability measurements between participants in control, baseline, and discharge groups (
Values of transthoracic bioimpedance and heart rate variability parameters.
Parameters | Control (n=32), mean (SD) | Baseline (n=23), mean (SD) | Discharge (n=17), mean (SD) | ||||||||
|
|
|
|
|
|
||||||
|
38.1 (10.8) | 26.5 (12.8)a | .006 | 34.2 (17.4) | .99 | ||||||
|
52.0 (17.0) | 52.0 (24.7) | >.999 | 54.3 (23.3) | >.999 | ||||||
|
4.08·10–8 (2.96·10–8) | 4.60·10–8 (1.71·10–8) | >.999 | 4.42·10–8 (1.85·10–8) | >.999 | ||||||
|
α | 0.609 (0.0881) | 0.716 (0.121)a | .003 | 0.646 (0.144) | .87 | |||||
|
6.11·10–4 ( 3.45·10–4) | 5.34·10–4 (1.51·10–4) | .83 | 5.07·10–4 (1.72·10–4) | .56 | ||||||
|
Fitting error (Hz) | 334 (669) | 232 (389) | .51 | 347 (374) | .35 | |||||
|
21.5 (6.0) | 17.0 (7.5) | .08 | 20.3 (9.1) | >.999 | ||||||
|
16.6 (6.1) | 9.54 (6.0)a | .001 | 13.9 (8.8) | .57 | ||||||
|
|
|
|
|
|
||||||
|
LFb HRVc | 3.5 (4.2) | 19.3 (43.4) | .06 | 19.2 (51.3) | .09 | |||||
|
Normalized LF HRV | 7.4 (14.4) | 32.9 (55.7)a | .02 | 34.6 (57.0)a | .01 | |||||
|
HFd HRV | 0.225 (0.134) | 0.178 (0.092) | .38 | 0.127 (0.085)a | .01 | |||||
|
Normalized HF HRV | 0.255 (0.154) | 0.391 (0.134)a | .003 | 0.371 (0.129)a | .02 | |||||
|
PDMI sympathetice | 11.8 (5.52) | 17.2 (12.4) | .06 | 15.3 (5.98) | .52 | |||||
|
PDMI parasympatheticf | 13.2 (5.47) | 17.1 (10.4) | .20 | 17.9 (7.56) | .14 | |||||
|
Mean heart rate | 72.3 (11.9) | 74.1 (18.0) | >.999 | 74.7 (15.9) | >.999 |
aDenotes a statistically significant difference with respect to control group.
bLF: low-frequency.
cHRV: heart rate variability.
dHF: high-frequency.
ePDMI sympathetic: principal dynamic mode index of sympathetic function.
fPDMI parasympathetic: principal dynamic mode index of parasympathetic function.
As for the heart rate variability parameters, for the baseline and discharge groups, high-frequency components of heart rate variability (baseline:
Highest accuracy and parameters included for control/baseline/discharge classification in each machine learning algorithm.
Type | Cubic SVMa | Quadratic SVM | Gaussian SVM | Decision tree | |||
|
|
|
|
|
|
||
|
Accuracy, % | 63 | 61 | 68 | 67 | 72 | |
|
|
|
|
|
|
|
|
|
|
|
x | x | x | x | x |
|
|
|
x |
|
x |
|
|
|
|
|
|
x |
|
x |
|
|
|
α |
|
|
x | x | x |
|
|
|
|
|
|
x | x |
|
|
Fitting error | x | x |
|
|
|
|
|
|
|
x |
|
x |
|
|
|
|
|
|
|
|
x |
|
|
|
|
|
|
||
|
Accuracy, % | 58 | 63 | 56 | 57 | 53 | |
|
|
|
|
|
|
|
|
|
|
LFb HRVc | x | x |
|
|
|
|
|
Normalized LF HRV | x | x | x | x | x |
|
|
HFd HRV | x | x |
|
x | x |
|
Normalized HF HRV |
|
x |
|
|
x | |
|
|
PDMI sympathetice |
|
x |
|
x |
|
|
|
PDMI parasympatheticf | x | x | x | x |
|
|
|
Mean heart rate |
|
x |
|
x | x |
|
|
|
|
|
|
||
|
Accuracy, % | 74 | 75 | 68 | 74 | 72 | |
|
|
|
|
|
|
|
|
|
|
|
x | x | x |
|
x |
|
|
|
x | x | x |
|
|
|
|
|
x |
|
|
x |
|
|
|
α | x |
|
x | x | x |
|
|
|
|
x |
|
|
x |
|
|
Fitting error | x | x |
|
x |
|
|
|
|
|
x |
|
|
|
|
|
|
|
x |
|
|
x |
|
|
LF HRV | x |
|
|
|
|
|
|
Normalized LF HRV |
|
|
|
|
|
|
|
HF HRV | x | x |
|
x |
|
|
|
Normalized HF HRV |
|
x |
|
|
|
|
|
PDMI sympathetic |
|
|
|
x |
|
|
|
PDMI parasympathetic |
|
|
|
|
|
|
|
Mean heart rate | x | x |
|
x |
|
aSVM: support vector machine.
bLF: low-frequency.
cHRV: heart rate variability.
dHF: high-frequency.
ePDMI sympathetic: principal dynamic mode index of sympathetic function.
fPDMI parasympathetic: principal dynamic mode index of parasympathetic function.
Highest accuracy and parameters included for patients without fluid/patients with fluid classification on each machine learning algorithm
Type | Cubic SVM | Quadratic SVM | Gaussian SVM | Decision tree | |||
|
|
|
|
|
|
||
|
Accuracy, % | 82 | 75 | 82 | 78 | 79 | |
|
|
|
|
|
|
|
|
|
|
|
x | x | x | x |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
x | x |
|
|
|
α | x |
|
|
|
|
|
|
|
|
|
x |
|
x |
|
|
Fitting error | x |
|
x |
|
|
|
|
|
x | x | x | x | x |
|
|
|
|
|
|
|
x |
|
|
|
|
|
|
||
|
Accuracy, % | 75 | 76 | 75 | 71 | 72 | |
|
|
|
|
|
|
|
|
|
|
LFb HRVc |
|
x |
|
|
|
|
|
Normalized LF HRV | x |
|
x |
|
x |
|
|
HFd HRV |
|
x | x | x |
|
|
Normalized HF HRV | x |
|
|
|
x | |
|
|
PDMI sympathetice | x |
|
x | x |
|
|
|
PDMI parasympatheticf |
|
x |
|
|
|
|
|
Mean heart rate |
|
x | x | x |
|
|
|
|
|
|
|
||
|
Accuracy, % | 92 | 88 | 83 | 85 | 81 | |
|
|
|
|
|
|
|
|
|
|
|
x |
|
x | x |
|
|
|
|
x |
|
|
|
x |
|
|
|
x | x |
|
x |
|
|
|
α |
|
|
|
|
|
|
|
|
|
x |
|
|
|
|
|
Fitting error |
|
|
x | x |
|
|
|
|
|
x | x |
|
x |
|
|
|
|
x |
|
|
|
|
|
LF HRV | x |
|
|
x |
|
|
|
Normalized LF HRV |
|
|
|
x | x |
|
|
HF HRV |
|
x | x |
|
|
|
|
Normalized HF HRV |
|
x |
|
x |
|
|
|
PDMI sympathetic |
|
|
|
|
|
|
|
PDMI parasympathetic | x |
|
|
x |
|
|
|
Mean heart rate | x | x |
|
x |
|
aSVM: support vector machine.
bLF: low-frequency.
cHRV: heart rate variability.
dHF: high-frequency.
ePDMI sympathetic: principal dynamic mode index of sympathetic function.
fPDMI parasympathetic: principal dynamic mode index of parasympathetic function.
Confusion matrix for quadratic support vector machine—the most accurate model for control/baseline/discharge classification.
Actual | Predicted, % | ||
|
Control | Baseline | Discharge |
Control | 78.1 | 6.3 | 15.6 |
Baseline | 13.0 | 82.6 | 4.3 |
Discharge | 29.4 | 11.8 | 58.8 |
Confusion matrix for cubic support vector machine—the most accurate model for patients without fluid/patients with fluid classification.
Actual | Predicted, % | |
|
Patients with fluid | Patients without fluid |
Patients with fluid | 82.6 | 17.4 |
Patients without fluid | 4.1 | 95.9 |
In this prospective observational study, we successfully trained machine learning models to classify participants with and without fluid accumulation using parameters obtained with a fluid accumulation vest, specifically transthoracic bioimpedance and heart rate variability parameters. We achieved a cross-validation accuracy of 92% using a cubic support vector machine model. The transthoracic bioimpedance parameters that contributed to this accuracy were related to intra- and extracellular fluid, whereas the heart rate variability parameters were mostly related to sympathetic activation. Our results suggest that the transthoracic bioimpedance and heart rate variability signals acquired with a wearable vest with carbon black–PDMS dry electrodes are suitable for detecting fluid accumulation and can potentially help with prediction and management of clinical worsening in heart failure patients.
In the past, transthoracic bioimpedance has been used for lung fluid abnormality detection [
Bioimpedance is a proven biomarker of acute decompensated heart failure. Our group previously performed a clinical study of 106 hospitalized patients discharged after an admission for acute decompensated heart failure. Participants were sent home with a fluid accumulation vests and we determined that it was feasible to measure transthoracic bioimpedance on a daily basis [
As for the heart rate variability, high-frequency components of heart rate variability (at admission:
In the machine learning classifications,
Statistical analysis and machine learning analysis showed similar results for a reduced set of features. For instance, extracellular resistance and low-frequency components of heart rate variability exhibited significant differences between non–heart failure (control) and heart failure groups (baseline and discharge), and these features were present in the most accurate model for fluid accumulation detection. However, other features including intracellular resistance, cell membrane capacitance, principal dynamic mode index of parasympathetic function, and mean heart rate did not exhibit significant differences between groups but were relevant for improving accuracy of the machine learning algorithms.
As for the limitations of the study, many recordings were not usable, mostly in the acute decompensated heart failure group. This is related to technical issues with the fluid accumulation vests, which can be partially attributed to the carbon black–PDMS electrodes. From the 28 participants with acute decompensated heart failure, we obtained reliable measures from only 23 participants at baseline and from 17 participants at discharge. We obtained data from both baseline and discharge for only 12 participants. Even in the control group, we collected usable data from only 32 out of the 50 participants. In some instances, applying a layer of hydrating lotion helped with data collection. This limitation could potentially diminish the clinical use of the device and must be addressed in the near future. A more robust hardware design, tailored to match the impedance of the carbon black–PDMS electrodes, is a potential improvement. Configurations that enable collection of transthoracic bioimpedance data from several locations on the thorax could help the quality and usability of the data, as accumulation of fluid does not occur always in the same location. Furthermore, given the limited data set, we have reported leave-one-subject-out cross-validation accuracy, and the results cannot be interpreted as conclusive concerning the efficacy of the transthoracic bioimpedance device and features derived from it. Instead, the results can be interpreted as promising, based on the validation of the transthoracic bioimpedance and its associated features and machine learning. A larger testing data set is required for further evaluation of transthoracic bioimpedance to allow for more definite conclusions about its efficacy.
There are several potential clinical applications of transthoracic bioimpedance measurements in patients with heart failure. Wearable technologies such as fluid accumulation vests could allow for rapid point-of-care diagnostics that could be used in the emergency setting to help identify heart failure decompensation. In addition, fluid accumulation vest measurements in different clinical states such as decompensated heart failure, predischarge, and in outpatient setting, could be used to establish a profile for a given patient that could improve diagnostic certainty and guide treatment. Moreover, triaging medical severity is a necessary and time-consuming step of the patient care process, but this is often difficult due to limitations in both the number of available medical personnel and individual provider time.
The device and algorithm in this study can be used in a longitudinal study with patients with heart failure, extending monitoring into the home. The system could be used to monitor a patient’s fluid accumulation daily and generate early warnings of heart failure decompensation, provide guidance on therapeutic changes to improve quality of life, and reduce heart failure readmissions. Alternatively, the system can be used to monitor either the discharge readiness of a patient from the hospital or the home treatment regime effectiveness on the patient. Wearable sensors such as the fluid accumulation vest can potentially provide an ideal avenue for patient monitoring over time, allowing for rapid action in response to acute decompensation. Garments integrating vital sign sensors have been utilized in acute medical settings to monitor patients with high medical risk profiles [
The main goal of this study was to evaluate the performance of biologically relevant parameters measured by a fluid accumulation vests with carbon black–PDMS dry electrodes. In our clinical study (SHIELD), transthoracic bioimpedance and heart rate variability parameters were considered for statistical analysis and discrimination between patients with nonacute decompensated heart failure and acute decompensated heart failure. As expected, our results show that among the 15 parameters, 2 (extracellular resistance and intracellular-extracellular difference in resistance) showed statistically significantly lower values (
atrial fibrillation
electrocardiography
end-stage renal disease
heart rate variability
implantable cardioverter-defibrillator
polydimethylsiloxane
This work was supported by the Smart and Connect Health of the National Science Foundation grants: 1522087, 1522084, and 1522052.
DDM has received honorary, speaking/consulting fee or grants from Flexcon, Rose Consulting, Bristol-Myers Squibb, Pfizer, Boston Biomedical Associates, Samsung, Phillips, Mobile Sense, Care Evolution, Flexcon Boehringer Ingelheim, Biotronik, Otsuka Pharmaceuticals, and Sanofi.