Machine Learning Model Based on Transthoracic Bioimpedance and Heart Rate Variability for Lung Fluid Accumulation Detection: Prospective Clinical Study.

BACKGROUND: 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. OBJECTIVE: 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). METHODS: 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 with fluid and without fluid groups, respectively. RESULTS: 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: P=.006; R0 - R∞: P=.001), indicating that a higher volume of fluids accumulated in the lungs of patients with heart failure. A cubic support vector machine model using the 5 parameters achieved an accuracy of 92% for with fluid and without fluid group classification. The transthoracic bioimpedance parameters were related to intra- and extracellular fluid, whereas the heart rate variability parameters were mostly related to sympathetic activation. CONCLUSIONS: 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.

Novel Method of Atrial Fibrillation Case Identification and Burden Estimation Using the MIMIC-III Electronic Health Data Set.

BACKGROUND: Atrial fibrillation (AF) portends poor prognoses in intensive care unit patients with sepsis. However, AF research is challenging: Previous studies demonstrate that International Classification of Disease (ICD) codes may underestimate the incidence of AF, but chart review is expensive and often not feasible. We aim to examine the accuracy of nurse-charted AF and its temporal precision in critical care patients with sepsis. METHODS: Patients with sepsis with continuous electrocardiogram (ECG) waveforms were identified from the Medical Information Mart for Intensive Care (MIMIC-III) database, a de-identified, single-center intensive care unit electronic health record (EHR) source. We selected a random sample of ECGs of 6 to 50 hours' duration for manual review. Nurse-charted AF occurrence and onset time and ICD-9-coded AF were compared to gold-standard ECG adjudication by a board-certified cardiac electrophysiologist blinded to AF status. Descriptive statistics were calculated for all variables in patients diagnosed with AF by nurse charting, ICD-9 code, or both. RESULTS: From 142 ECG waveforms (58 AF and 84 sinus rhythm), nurse charting identified AF events with 93% sensitivity (95% confidence interval [CI]: 87%-100%) and 87% specificity (95% CI: 80%-94%) compared to the gold standard manual ECG review. Furthermore, nurse-charted AF onset time was within 1 hour of expert reader onset time for 85% of the reviewed tracings. The ICD-9 codes were 97% sensitive (95% CI: 88-100%) and 82% specific (95% CI: 74-90%) for incident AF during admission but unable to identify AF time of onset. CONCLUSION: Nurse documentation of AF in EHR is accurate and has high precision for determining AF onset to within 1 hour. Our study suggests that nurse-charted AF in the EHR represents a potentially novel method for AF case identification, timing, and burden estimation.

Atrial Fibrillation Detection in ICU Patients: A Pilot Study on MIMIC III Data.

Atrial fibrillation (AF) is the most prevalent arrhythmia, resulting in varying and irregular heartbeats. AF increases risk for numerous cardiovascular diseases including stroke, heart failure and as a result, computer aided efficient monitoring of AF is crucial, especially for intensive care unit (ICU) patients. In this paper, we present an automated and robust algorithm to detect AF from ICU patients using electrocardiogram (ECG) signals. Several statistical parameters including root mean square of successive differences, Shannon entropy, Sample entropy and turning point ratio are calculated from the heart rate. A subset of the Medical Information Mart for Intensive Care (MIMIC) III database containing 36 subjects is used in this study. We compare the AF detection performance of several classifiers for both the training and blinded test data. Using the support vector machine classifier with radial basis kernel, the proposed method achieves 99.95% cross-validation accuracy on the training data and 99.88% sensitivity, 99.65% specificity and 99.75% accuracy on the blinded test data.