Classification of Blood Volume Decompensation State via Machine Learning Analysis of Multi-Modal Wearable-Compatible Physiological Signals.

This paper presents a novel computational algorithm to estimate blood volume decompensation state based on machine learning (ML) analysis of multi-modal wearable-compatible physiological signals. To the best of our knowledge, our algorithm may be the first of its kind which can not only discriminate normovolemia from hypovolemia but also classify hypovolemia into absolute hypovolemia and relative hypovolemia. We realized our blood volume classification algorithm by (i) extracting a multitude of features from multi-modal physiological signals including the electrocardiogram (ECG), the seismocardiogram (SCG), the ballistocardiogram (BCG), and the photoplethysmogram (PPG), (ii) constructing two ML classifiers using the features, one to classify normovolemia vs. hypovolemia and the other to classify hypovolemia into absolute hypovolemia and relative hypovolemia, and (iii) sequentially integrating the two to enable multi-class classification (normovolemia, absolute hypovolemia, and relative hypovolemia). We developed the blood volume decompensation state classification algorithm using the experimental data collected from six animals undergoing normovolemia, relative hypovolemia, and absolute hypovolemia challenges. Leave-one-subject-out analysis showed that our classification algorithm achieved an F1 score and accuracy of (i) 0.93 and 0.89 in classifying normovolemia vs. hypovolemia, (ii) 0.88 and 0.89 in classifying hypovolemia into absolute hypovolemia and relative hypovolemia, and (iii) 0.77 and 0.81 in classifying the overall blood volume decompensation state. The analysis of the features embedded in the ML classifiers indicated that many features are physiologically plausible, and that multi-modal SCG-BCG fusion may play an important role in achieving good blood volume classification efficacy. Our work may complement existing computational algorithms to estimate blood volume compensatory reserve as a potential decision-support tool to provide guidance on context-sensitive hypovolemia therapeutic strategy.

Unifying the Estimation of Blood Volume Decompensation Status in a Porcine Model of Relative and Absolute Hypovolemia Via Wearable Sensing.

Hypovolemia remains the leading cause of preventable death in trauma cases. Recent research has demonstrated that using noninvasive continuous waveforms rather than traditional vital signs improves accuracy in early detection of hypovolemia to assist in triage and resuscitation. This work evaluates random forest models trained on different subsets of data from a pig model (n = 6) of absolute (bleeding) and relative (nitroglycerin-induced vasodilation) progressive hypovolemia (to 20% decrease in mean arterial pressure) and resuscitation. Features for the models were derived from a multi-modal set of wearable sensors, comprised of the electrocardiogram (ECG), seismocardiogram (SCG) and reflective photoplethysmogram (RPPG) and were normalized to each subject.s baseline. The median RMSE between predicted and actual percent progression towards cardiovascular decompensation for the best model was 30.5% during the relative period, 16.8% during absolute and 22.1% during resuscitation. The least squares best fit line over the mean aggregated predictions had a slope of 0.65 and intercept of 12.3, with an R2 value of 0.93. When transitioned to a binary classification problem to identify decompensation, this model achieved an AUROC of 0.80. This study: a) developed a global model incorporating ECG, SCG and RPPG features for estimating individual-specific decompensation from progressive relative and absolute hypovolemia and resuscitation; b) demonstrated SCG as the most important modality to predict decompensation; c) demonstrated efficacy of random forest models trained on different data subsets; and d) demonstrated adding training data from two discrete forms of hypovolemia increases prediction accuracy for the other form of hypovolemia and resuscitation.

Enabling the assessment of trauma-induced hemorrhage via smart wearable systems.

As the leading cause of trauma-related mortality, blood loss due to hemorrhage is notoriously difficult to triage and manage. To enable timely and appropriate care for patients with trauma, this work elucidates the externally measurable physiological features of exsanguination, which were used to develop a globalized model for assessing blood volume status (BVS) or the relative severity of blood loss. These features were captured via both a multimodal wearable system and a catheter-based reference and used to accurately infer BVS in a porcine model of hemorrhage (n = 6). Ultimately, high-level features of cardiomechanical function were shown to strongly predict progression toward cardiovascular collapse and used to estimate BVS with a median error of 15.17 and 18.17% for the catheter-based and wearable systems, respectively. Exploring the nexus of biomedical theory and practice, these findings lay the groundwork for digital biomarkers of hemorrhage severity and warrant further study in human subjects.

Testing the Efficacy of Pharmacological Agents in a Pericardial Target Delivery Model in the Swine.

To date, many pharmacological agents used to treat or prevent arrhythmias in open-heart cases create undesired systemic side effects. For example, antiarrhythmic drugs administered intravenously can produce drops in systemic pressure in the already compromised cardiac patient. While performing open-heart procedures, surgeons will often either create a small port or form a pericardial cradle to create suitable fields for operation. This access yields opportunities for target pharmacological delivery (antiarrhythmic or ischemic preconditioning agents) directly to the myocardial tissue without undesired side effects. We have developed a swine model for testing pharmacological agents for target delivery within the pericardial fluid. While fully anesthetized, each animal was instrumented with a Swan-Ganz catheter as well as left and right ventricle pressure catheters, and pacing leads were placed in the right atrial appendage and the right ventricle. A medial sternotomy was then performed and a pericardial access cradle was created; a plunge pacing lead was placed in the left atrial appendage and a bipolar pacing lead was placed in the left ventricle. Utilizing a programmer and a cardiac mapping system, the refractory period of the atrioventricular node (AVN), atria and ventricles was determined. In addition, atrial fibrillation (AF) induction was produced utilizing a Grass stimulator and time in AF was observed. These measurements were performed prior to treatment, as well as 30 min and 60 min after pericardial treatment. Additional time points were added for selected studies. The heart was then cardiopleged and reanimated in a four chamber working mode. Pressure measurements and function were recorded for 1 hr after reanimation. This treatment strategy model allowed us to observe the effects of pharmacological agents that may decrease the incidence of cardiac arrhythmias and/or ischemic damage, during and after open-heart surgery.

Tissue properties of the fossa ovalis as they relate to transseptal punctures: a translational approach.

OBJECTIVES: This study focused on how catheter size affects transseptal puncture, what transseptal indication means, and whether the swine model is predictive for humans. BACKGROUND: Transseptal puncture is a common procedure that gains access to the left atrium, allowing percutaneous mitral valve repair, left atrial appendage closure, and left-sided ablations. The basic approach has not changed in many years; however, the frequency of transseptal punctures and the size of devices are increasing with emerging treatments. METHODS: A broad range of devices (4 F to 18 F) were advanced through atrial septa of swine hearts; some devices were inserted in both swine and human hearts using 10 F catheters. RESULTS: Greater forces were required to puncture through the septa of human hearts compared to those of swine. Larger catheters used in swine hearts required greater force to advance them through the septa, causing greater dilation of tissue and sometimes tearing the floor of the fossa ovalis; analyses indicated an exponential increase in the size of the iatrogenic atrial septal defect. Specific tissue property testing of the septum primum showed that this tissue sheared at a lower exerted force in a superior to inferior direction. CONCLUSIONS: Results may provide physicians with important knowledge about what to expect when treating a possible iatrogenic atrial septal defect or help them understand the consequences of transseptal punctures. Comparative data between swine and human atrial septal tissue properties provide critical insights between the species and offer clinicians and device designers important information relative to differences in tissue behaviors.

Cardiac responses to the intrapericardial delivery of metoprolol: targeted delivery compared to intravenous administration.

Anti-arrhythmic drugs have narrow therapeutic ranges and typically can engender harmful side effects. The intrapericardial (IP) delivery of anti-arrhythmic agents proposes to achieve higher myocardial levels while minimizing plasma concentrations, thus diminishing systemic side effects. Furthermore, IP delivery enables concentrations at the target site to be more precisely controlled. Our study objective was to compare the relative cardiac effects of intrapericardial administration of metoprolol to standard intravenous (IV) delivery in a swine surgical model. In order to answer the question of how IP metoprolol affects sinus tachycardia, atrial electrophysiology, and pharmacokinetics compared with IV delivery, a medial sternotomy was performed on 21 swine that were divided into three groups: (1) After inducing sinus tachycardia, metoprolol boluses were delivered IP (n = 4) or IV (n = 4); (2) metoprolol was administered either IP (n = 3) or IV (n = 3) with saline controls (n = 3), and electrophysiologic data were collected; (3) metoprolol levels were tracked both in the blood (IV, n = 2) and pericardial (IP, n = 2) fluid. After either IP or IV delivery of metoprolol, heart rates were lowered significantly to 70% and 73% of control rate, respectively. The therapeutic effect of IV-administered metoprolol was considerably reduced after 1 h but was sustained longer in the IP group. Additionally, ventricular contractility and mean arterial pressure parameters were significantly lower in IV-treated animals but were nearly unaffected in IP-treated animals. With IP administration, the elimination half-life of metoprolol in pericardial fluid was 14.4 min with negligible accumulations in the plasma, whereas with IV delivery, the elimination half-life in plasma was 11.1 min with negligible amounts found in the pericardial fluid. The targeted intrapericardial delivery of metoprolol effectively lowers heart rates for sustained periods of time, with minimal effect on either ventricular contractility or mean arterial pressure. We did not observe dramatic changes in induced atrial fibrillation times or refractory periods using this model.

Cardiac remodeling as a consequence of atrial fibrillation: An anatomical study of perfusion-fixed human heart specimens.

BACKGROUND: Atrial fibrillation (AF) causes a continuum of atrial anatomical remodeling. METHODS: Using a library of perfusion-fixed human hearts, specimens with AF were compared to controls. During this preliminary assessment study, direct measurements were taken of atrial volume, pulmonary vein (PV) circumference, and left atrial (LA) wall thicknesses. RESULTS: Hearts with AF typically had larger atrial volumes, as well as a much larger variation in volume compared to controls (range of 59.6-227.1 mL in AF hearts compared to 65.1-115.9 mL in controls). For all hearts, right PVs were larger than left PVs (mean: 171.4 ± 84.6 mm([2]) for right and 118.2 ± 50.1 mm([2]) for left, P < 0.005). LA wall thicknesses ranged from 0.7 mm to 3.1 mm for both AF and control hearts. CONCLUSIONS: Hearts with AF had a large range of sizes which is consistent with the progression of atrial remodeling during AF. The large range of thicknesses will influence the amount of energy needed to create transmural lesions during ablation procedures.

MRI assessment of pacing induced ventricular dyssynchrony in an isolated human heart.

This study demonstrates the capabilities of MRI in the assessment of cardiac pacing induced ventricular dyssynchrony, and the findings support the need for employing more physiological pacing. A human donor heart deemed non-viable for transplantation, was reanimated using an MR compatible, four-chamber working perfusion system. The heart was imaged using a 1.5T MR scanner while being paced from the right ventricular apex (RVA) via an epicardial placed lead. Four-chamber, short-axis, and tagged short-axis cines were acquired in order to track wall motion and intramyocardial strain during pacing. The results of this study revealed that the activation patterns of the left ventricle (LV) during RVA pacing demonstrated intraventricular dyssynchrony; as the left ventricular mechanical activation proceeded from the septum and anterior wall to the lateral wall, with the posterior wall being activated last. As such, the time difference to peak contraction between the septum and lateral wall was approximately 125 msec. Likewise, interventricular dyssynchrony was demonstrated from the four-chamber cine as the time difference between the peak LV and RV free wall motion was 180 msec. With the ongoing development of MR safe and MR compatible pacing systems, we can expect MRI to be added to the list of imaging modalities used to optimize cardiac resynchronization therapy (CRT) and/or alternate site pacing.