Physiological condition monitoring based on biovibration and radio frequency data analysis

Abstract

The patient monitoring device includes an ECG sensor coupled to the patient, a sensor coupled to the patient and configured to generate a bio-oscillation signal, and a radio frequency monitoring device configured to generate information corresponding to electromagnetic energy reflected from the patient's thoracic cavity. The processor processes the ECG signal, the bio-oscillation signal, and the radio frequency information to generate a plurality of physiological parameters of the patient. The processor further performs at least one of a predictive analysis and a trend analysis of the plurality of physiological parameters to determine a current clinical condition of the patient. The trend analysis includes determining a substantial relationship between changes in the plurality of physiological parameters. The processor may further compare the patient's current clinical condition with predetermined clinically actionable criteria to determine one or more clinically actionable events and provide an output related to the one or more clinically actionable events.

Description

[Background technology]

[0001] Related Applications This application is a continuation of U.S. Provisional Application No. 62 / 644,216, entitled "Monitoring Physiological Status Based on Bio-Vibrational and Radio Frequency Data Analysis," filed March 16, 2018, and claims priority to that provisional application under 35 U.S.C. § 119. The entire contents of all subject matter described in the above application are hereby incorporated by reference into this application as if fully set forth herein.

[0002] The present disclosure relates to monitoring a patient's physiological condition based on patient data obtained from multiple sensor sources, such as bio-vibration sensors and radio frequency sensors.

[0003] Patients with cardiac disease are prone to frequent recurrence of acute episodes due to congestive heart failure (CHF). For example, pulmonary congestion is a leading cause of hospitalization and rehospitalization among CHF patients. Similarly, patients with chronic kidney disease undergoing hemodialysis also suffer from acute adverse events. For example, some patients with end-stage renal disease may have moderate to severe pulmonary congestion before hemodialysis. These patients have a high prevalence of heart failure and a poor prognosis. Various electronic and mechanical devices exist for monitoring and treating a patient's heart failure. In some instances, medical devices such as cardiac monitors or defibrillators may be surgically implanted or externally connected to the patient, depending on the underlying condition being monitored or treated. In some cases, physicians may treat heart failure using medical devices alone or in combination with pharmaceutical therapy.

[0004] Left untreated, heart failure can lead to certain life-threatening arrhythmias. Both atrial and ventricular arrhythmias are common in patients with heart failure. One of the most life-threatening cardiac arrhythmias is ventricular fibrillation, which occurs when normal, regular electrical impulses are replaced by irregular, rapid impulses, causing the heart muscle to cease normal contractions. Because patients receive no perceptible warning of the impending fibrillation, death often occurs before necessary medical assistance can arrive. Other cardiac arrhythmias can include an extremely slow heart rate, known as bradycardia, or an extremely fast heart rate, known as tachycardia. Cardiac arrest can occur when a patient's heart fails to deliver sufficient blood flow to the brain and other vital organs due to various cardiac arrhythmias, including ventricular fibrillation, ventricular tachycardia, excitation-contraction dissociation (PEA), and asystole (when the heart ceases all electrical activity). It is generally helpful to monitor patients with heart failure to diagnose symptoms of heart failure early and provide interventional therapy as soon as possible. Summary of the Invention

[0005] In certain implementations, the patient monitoring system includes an ECG sensor coupled to the patient and configured to detect one or more ECG signals of the patient, a vibration sensor coupled to the patient and configured to detect one or more cardiac vibration signals of the patient, an ultra-wideband radio frequency transceiver circuit including one or more radio frequency antennas and coupled to the patient, and one or more processors. In some examples, the ultra-wideband radio frequency transceiver is configured to cause the one or more radio frequency antennas to direct radio frequency electromagnetic energy toward the patient's thoracic cavity and receive, via the one or more radio frequency antennas, radio frequency information corresponding to the reflected radio frequency electromagnetic energy reflected from within the patient's thoracic cavity. In some examples, the one or more processors are configured to process the one or more ECG signals, the one or more cardiac vibration signals, and the radio frequency information to generate multiple physiological parameters of the patient, including one or more combined physiological parameters. The one or more processors may perform at least one of predictive analysis and trend analysis of the multiple physiological parameters, including the combined physiological parameters, to determine a current clinical condition of the patient. In implementations described herein, the trend analysis includes determining the existence of a substantial relationship between changes in the plurality of physiological parameters. The one or more processors may compare the patient's current clinical symptoms to predetermined clinically actionable criteria to determine one or more clinically actionable events and cause an output device to provide an output related to the one or more clinically actionable events. Implementations of the patient monitoring systems described herein may include one or more of the following features.

[0006] In certain implementations of the above-described patient monitoring system, the trend analysis includes a correlation analysis.

[0007] Certain implementations of the above-described patient monitoring system perform at least one of predictive analysis and trend analysis on physiological parameters collected over a previous period including one or more of at least 24 hours, at least 48 hours, at least one week, at least two weeks, at least one month, at least six weeks, at least two months, at least four months, at least six months, at least one year, and at least two years.

[0008] Certain implementations of the above-described patient monitoring systems perform at least one of predictive and trend analysis on physiological parameters collected over a previous period corresponding to at least available patient history.

[0009] In certain implementations of the above-described patient monitoring system, the one or more clinically actionable events include at least one of an automatic event that is triggered without user input and a manual event that is triggered based on a user response to an output and comprises one or more instructions to perform one or more actions.

[0010] In certain implementations of the above-described patient monitoring system, the output regarding one or more clinically actionable events is based on exceeding one or more thresholds defined for a plurality of physiological parameters or results of at least one of predictive and trending analyses.

[0011] In a particular implementation of the above-described patient monitoring system, performing the predictive analysis includes inputting the plurality of physiological parameters into an artificial neural network and determining a current clinical condition of the patient based on an output of the artificial neural network.

[0012] In a particular implementation of the above-described patient monitoring system, performing the predictive analysis includes inputting the plurality of physiological parameters into a deep learning process and determining a current clinical condition of the patient based on the output of the deep learning process.

[0013] In certain implementations of the patient monitoring system described above, performing the predictive analysis includes inputting the plurality of physiological parameters into a machine learning process and determining a current clinical condition of the patient based on an output of the machine learning process. In some examples, the current clinical condition of the patient includes a prediction score based on the output of the machine learning process, the prediction score indicating a likelihood of an adverse event occurring. In some examples, the adverse event includes one or more of an arrhythmia event, a seizure event, a syncope event, and a hospitalization event.

[0014] In certain implementations of the above-described patient monitoring system, performing the trend analysis includes determining whether at least one of the plurality of physiological parameters exceeds a threshold, correlating the at least one of the plurality of physiological parameters that exceeds the threshold against at least one additional physiological parameter selected from the plurality of physiological parameters to generate a correlation score, and determining a current clinical condition of the patient based on the correlation score. In some examples, the correlation score indicates a change in the patient's condition.

[0015] In certain implementations of the above-described patient monitoring system, the vibration sensor is further configured to sense one or more pulmonary vibrations of the patient, the one or more pulmonary vibrations including at least one of bronchial vibrations, wheezing, crackles, wheezing, bullae, pleural rubs, squawks, glottal sounds, pharyngeal sounds, or other vibrations.

[0016] In certain implementations of the above-described patient monitoring system, the one or more ECG signals may include heart rate, heart rate variability, PVC load or PVC count, atrial fibrillation load, Pose , heart rate irregularity, QRS height, QRS width, changes in size or shape of one or more ECG signal morphologies, cosine RT, artificial pacing, QT interval, QT variation, T wave width, T wave alternans, T wave variation, and ST segment changes.

[0017] In a particular implementation of the above-described patient monitoring system, the one or more cardiac vibration signals include at least one of an S1 vibration, an S2 vibration, an S3 vibration, an S4 vibration, and a cardiac murmur vibration.

[0018] In a particular implementation of the above-described patient monitoring system, the radio frequency information includes a measurement of fluid content within the patient's thoracic cavity.

[0019] In a particular implementation of the above-described patient monitoring system, the plurality of physiological parameters includes one or more of left ventricular contraction time (LVST), electromechanical activation time (EMAT), %LVST, and left ventricular end-diastolic pressure (LVEDP).

[0020] In a particular implementation of the above-described patient monitoring system, one or more processors are incorporated within a cardiac monitoring device adapted to be worn by the patient.

[0021] In certain implementations of the above-described patient monitoring system, one or more processors are incorporated into a remote processing device.

[0022] In a particular implementation of the above-described patient monitoring system, one or more processors are incorporated into a wearable defibrillation device adapted to be worn by the patient.

[0023] In certain implementations, the second patient monitoring system includes at least one vibration sensor coupled to the patient and configured to detect one or more cardiac vibration signals; at least one ultra-wideband radio frequency transceiver coupled to the patient; and one or more processors. In some examples, the at least one ultra-wideband radio frequency transceiver is configured to direct radio frequency electromagnetic waves toward the patient's lungs and detect radio frequency information responsive to the radio frequency electromagnetic waves passing through the patient's lungs. In some examples, the one or more processors are configured to process the detected one or more cardiac vibration signals over a predetermined period of time to determine at least one cardiac vibration metric of the patient, process the patient's radio frequency information over a predetermined period of time to determine at least one lung fluid metric of the patient, determine an output related to one or more clinically actionable events based on the determined at least one cardiac vibration metric and the determined at least one lung fluid metric, and cause the output device to provide the output. Implementations of the second patient monitoring system described herein may include one or more of the following features.

[0024] In a particular implementation of the second patient monitoring system described above, the one or more clinically actionable events include at least one of an automatic event that is triggered without user input and a manual event that is triggered based on a user response to an output and comprises one or more instructions to perform one or more actions.

[0025] In a particular implementation of the second patient monitoring system described above, determining the output includes performing predictive analysis of the determined value or trend of at least one cardiac vibration metric and the determined value or trend of at least one lung fluid metric.

[0026] In a particular implementation of the second patient monitoring system described above, performing predictive analysis includes inputting the determined value or trend of at least one cardiac vibration metric and the determined value or trend of at least one lung fluid metric into a machine learning process; determining a predictive score based on an output of the machine learning process, the predictive score indicating a likelihood of an adverse event occurring; and determining an output based on the predictive score.

[0027] In certain implementations of the second patient monitoring system described above, determining the output includes determining the existence of a substantial relationship between changes in the at least one cardiac vibration metric and the at least one lung fluid metric by performing a trend analysis of changes in the at least one cardiac vibration metric and the at least one lung fluid metric. In some examples, detecting the existence of a substantial relationship between changes in the at least one cardiac vibration metric and the at least one lung fluid metric by performing a trend analysis of the changes includes performing a correlation analysis.

[0028] In a particular implementation of the above-described second patient monitoring system, determining the output includes performing a correlation analysis, the correlation analysis including processing the one or more cardiac vibration signals and the radio frequency information to generate a plurality of physiological parameters of the patient including one or more combined physiological parameters; correlating at least one of the plurality of physiological parameters that exceeds a threshold against at least one additional physiological parameter selected from the plurality of physiological parameters to generate a correlation score, the correlation score indicating a change in the patient's condition; and determining the output based on the correlation score.

[0029] In a particular implementation of the second patient monitoring system described above, the one or more cardiac vibration signals include at least one of an S1 vibration, an S2 vibration, an S3 vibration, an S4 vibration, a ventricular wall motion, and a cardiac murmur vibration.

[0030] In a particular implementation of the second patient monitoring system described above, the radio frequency information includes a measurement of fluid content within the patient's thoracic cavity.

[0031] In certain implementations, the third patient monitoring system includes an ECG sensor coupled to the patient and configured to detect one or more ECG signals of the patient, a vibration sensor coupled to the patient and configured to detect one or more cardiac vibration signals of the patient, an ultra-wideband radio frequency transceiver circuit including one or more radio frequency antennas and coupled to the patient, and one or more processors. In some examples, the ultra-wideband radio frequency transceiver circuit is configured to cause the one or more radio frequency antennas to direct radio frequency electromagnetic energy toward the patient's thoracic cavity and receive via the one or more radio frequency antennas radio frequency information corresponding to reflected radio frequency electromagnetic energy reflected from within the patient's thoracic cavity. In some examples, the one or more processors are configured to process the one or more ECG signals, the one or more cardiac vibration signals, and the radio frequency information to generate multiple physiological parameters of the patient, including one or more combined physiological parameters. The one or more processors may perform a trend analysis of the multiple physiological parameters, including the combined physiological parameter, to generate trend results, update the patient's monitoring schedule based on the trend results, determine a current clinical condition of the patient based on the trend results, and compare the patient's current clinical condition to predetermined clinically actionable criteria to determine one or more clinically actionable events. The one or more processors may cause an output device to provide output related to the one or more clinically actionable events. Implementations of the third patient monitoring system described herein may include one or more of the following features.

[0032] In a particular implementation of the above-mentioned third patient monitoring system, the one or more processors are further configured to process the one or more ECG signals, the one or more cardiac vibration signals, and the radio frequency information in accordance with the updated monitoring schedule to generate updated physiological parameters of the patient, perform a trend analysis of the updated physiological parameters including the one or more combined physiological parameters to generate updated trend results, and determine an updated clinical condition of the patient based on the updated trend results.

[0033] In a particular implementation of the third patient monitoring system described above, the one or more clinically usable events include at least one of an automatic event that is triggered without user input and a manual event that is triggered based on a user response to an output and comprises one or more instructions to perform one or more actions.

[0034] In a particular implementation of the third patient monitoring system described above, the one or more ECG signals may include heart rate, heart rate variability, PVC load or PVC count, atrial fibrillation load, Pose , heart rate irregularity, QRS height, QRS width, changes in size or shape of one or more ECG signal morphologies, cosine RT, artificial pacing, QT interval, QT variation, T wave width, T wave alternans, T wave variation, and ST segment changes.

[0035] In a particular implementation of the third patient monitoring system described above, the one or more cardiac vibration signals include at least one of an S1 vibration, an S2 vibration, an S3 vibration, an S4 vibration, and a cardiac murmur vibration.

[0036] In a particular implementation of the third patient monitoring system described above, the radio frequency information includes a measurement of fluid content in the patient's thoracic cavity.

[0037] In certain implementations, a fourth patient monitoring system includes an ECG sensor coupled to the patient and configured to detect one or more ECG signals of the patient, a vibration sensor coupled to the patient and configured to detect one or more cardiac vibration signals of the patient, an ultra-wideband radio frequency transceiver circuit including one or more radio frequency antennas and positioned on at least one aorta of the patient, and one or more processors. In some examples, the ultra-wideband radio frequency transceiver circuit is configured to cause the one or more radio frequency antennas to direct radio frequency electromagnetic energy toward at least a portion of the patient and receive, via the one or more radio frequency antennas, radio frequency information corresponding to reflected radio frequency electromagnetic energy reflected from within the at least a portion of the patient. In some examples, the one or more processors are configured to process the one or more ECG signals, the one or more cardiac vibration signals, and the radio frequency information to generate multiple physiological parameters of the patient, including one or more combined physiological parameters. The one or more processors may perform at least one of predictive analysis and trend analysis of the plurality of physiological parameters, including the combined physiological parameter, to determine a current clinical condition of the patient, compare the current clinical condition of the patient with predetermined clinically actionable criteria to determine one or more clinically actionable events, and cause an output device to provide an output related to the one or more clinically actionable events. Implementations of the fourth patient monitoring system described herein may include one or more of the following features.

[0038] In a particular implementation of the above fourth patient monitoring system, the at least one aorta includes at least one of the radial artery, the brachial artery, the aorta, and one or more pulmonary arteries.

[0039] In a particular implementation of the fourth patient monitoring system described above, the radio frequency information includes at least one of blood pressure information, cardiac wall motion information, blood flow information, cardiac rhythm information, and fluid content information. [Brief explanation of the drawings]

[0040] Various aspects of at least one example are discussed below with reference to the accompanying drawings, which are not intended to be drawn to scale. The drawings include illustrations and further understanding of various aspects and examples, and are incorporated in and constitute a part of this specification, but are not intended to limit the scope of the present disclosure. The drawings, together with the remainder of the specification, serve to explain the principles and operation of the described and claimed aspects and examples. In the drawings, each identical or nearly identical component shown in the various figures is designated by a similar numeral. For clarity, not all components may be labeled in every figure.

[0041] [Figure 1] 1 illustrates a wearable medical device according to an example of the present disclosure.

[0042] [Figure 2] 2 shows a schematic diagram of a sample controller of a wearable medical device, such as that shown in FIG. 1, according to an example of the present disclosure.

[0043] [Figure 3] 1 illustrates an overview of a sample network according to an example of the present disclosure.

[0044] [Figure 4A] 1 illustrates a sample radio frequency pathway unit attached to a patient, according to an example of the present disclosure.

[0045] [Figure 4B] 1 shows a schematic diagram of a radio frequency patch antenna unit according to an example of the present disclosure.

[0046] [Figure 5A] 1 shows a schematic plot of propagation delay and amplitude of radio frequencies reflected from tissue of a patient, according to an example of the present disclosure. [Figure 5B] 1 shows a schematic plot of propagation delay and amplitude of radio frequencies reflected from tissue of a patient, according to an example of the present disclosure.

[0047] [Figure 6] 1 illustrates an exemplary arterial pulse waveform, according to an example of the present disclosure.

[0048] [Figure 7] 1 shows a schematic diagram of a vibration sensor according to an example of the present disclosure.

[0049] [Figure 8] 1 shows a chart of lung vibrations and associated time amplitude plots.

[0050] [Figure 9] 1 illustrates a sample overview of a process utilizing predictive analytics, according to an example of the present disclosure.

[0051] [Figure 10] 1 illustrates a sample overview of a process for utilizing trend analysis, according to an example of the present disclosure.

[0052] [Figure 11] 1 shows a diagram of a patient monitoring system according to an example of the present disclosure. [Figure 12] 1 shows a diagram of a patient monitoring system according to an example of the present disclosure.

[0053] [Figure 13] 1 shows two sample ECG lead signals according to an example of the present disclosure.

[0054] [Figure 14] 1 illustrates a sample process for navigating a machine learning process according to an example of the present disclosure.

[0055] [Figure 15] 1 illustrates a sample architecture of an artificial neural network, according to an example of the present disclosure.

[0056] [Figure 16A] 10 illustrates a sample graph of trend analysis results with a Y-axis based on normalized units, according to an example of the present disclosure. [Figure 16B]10 illustrates a sample graph of trend analysis results with a Y-axis based on normalized units, according to an example of the present disclosure. [Figure 16C] 10 illustrates a sample graph of trend analysis results with a Y-axis based on normalized units, according to an example of the present disclosure. [Figure 16D] 10 illustrates a sample graph of trend analysis results with a Y-axis based on normalized units, according to an example of the present disclosure.

[0057] [Figure 17] 1 illustrates a sample process flow for monitoring a patient's condition using the techniques described herein, according to an example of the present disclosure. DETAILED DESCRIPTION OF THE INVENTION

[0058] Wearable medical devices, such as cardiac event monitoring devices, are used in clinical or ambulatory settings to monitor and record various physiological signals of a patient. These signals can be used to determine a patient's current symptoms and to predict, plan for, and prepare for future adverse events, such as potential cardiac events, or other adverse changes to the patient's medical health. For example, ECG monitoring combined with a patient's medical history can be utilized, e.g., via a machine learning process, to calculate a patient's risk score and predict, for example, the future occurrence of a potential cardiac arrhythmia event. An example of such a machine learning process is described in U.S. Patent Application Publication No. 2016 / 0135706, entitled "Medical Premonitory Event Estimation," the contents of which are incorporated herein by reference.

[0059] In some implementations, the patient monitoring system may include sensors configured to collect patient physiological signals other than ECG signals. For example, the vibration sensor may be configured to collect biomedical vibration signals, such as cardiac vibration signals, e.g., S1, S2, S3, and S4 signals, and pulmonary vibration signals, e.g., vibration signals from the lungs, particularly, or other airway-related anatomical regions, particularly the trachea and pharynx. Additionally, a radio frequency sensor, such as an ultra-wideband transceiver circuit, may be configured to collect information corresponding to radio frequency electromagnetic energy reflected from within the patient's thoracic cavity and / or heart. For example, the radio frequency sensor may be configured to monitor transthoracic (e.g., pulmonary) fluid volume and changes therein. In the implementations described herein, the processing device may further process the collected signals and associated information from the various underlying physiological sensors to generate one or more combined physiological parameters, e.g., a physiological parameter determined based on two or more underlying physiological sensors and associated information. At least one of predictive analysis (e.g., utilizing machine learning processes such as artificial neural networks) and trend analysis (e.g., correlation analysis) may be performed on the collected signals and received information, and the combined metrics, to determine the patient's current condition and one or more clinically actionable events.

[0060] For example, the present disclosure relates to a patient monitoring device and system. The device or system may include an ECG sensor coupled to the patient and configured to detect one or more ECG signals from the patient; a sensor coupled to the patient and configured to detect one or more bio-oscillation signals (including cardiac or pulmonary vibration signals) from the patient; and a radio-frequency-based monitoring device coupled to the patient and including a radio-frequency antenna and associated transceiver circuitry configured to direct radio-frequency electromagnetic energy toward the patient's thoracic cavity and / or heart. The radio-frequency transceiver circuitry produces a plurality of measurements corresponding to the radio-frequency electromagnetic energy reflected from the patient's thoracic cavity and / or heart and received via the receiving radio-frequency antenna. One or more processors process the one or more ECG signals, the one or more cardiac or pulmonary vibration signals, and the radio-frequency measurements to generate a plurality of physiological parameters for the patient, including one or more combined physiological parameters. Furthermore, the processor may be further configured to perform at least one of predictive analysis and trend analysis of the plurality of physiological parameters, including the combined physiological parameters, to determine a current clinical condition of the patient. In some implementations, a patient monitoring device coupled to a patient may transmit measured physiological information from the patient to a remote server for analysis. At the remote server, one or more processors may be configured to perform at least one of predictive analysis and trend analysis of the plurality of physiological parameters according to the principles described herein. For example, the trend analysis may include determining the existence of a substantial relationship between changes in the plurality of physiological parameters. The one or more processors may further compare the patient's current clinical symptoms with predetermined clinically actionable criteria to determine one or more clinically actionable events and cause an output device to provide an output related to the one or more clinically actionable events.

[0061] As an example, a patient may be examined by a healthcare provider at an outpatient facility and be required to use one or more wearable patient monitoring devices described herein to monitor the patient's physiological parameter set. For example, monitoring may begin with only one of the physiological parameter set, such as ECG-based metrics. Over time or in response to a specific symptom described herein, additional physiological parameter sets other than ECG parameters, such as bio-oscillation parameters or radio frequency-based parameters (or a combination thereof), may be initiated. Based on the collected physiological parameter set, the device may perform processing to assess the patient's current clinical condition. Upon a follow-up visit (e.g., weekly, biweekly, or monthly), the device may provide updated information regarding the patient's clinical condition. The device may further be configured to determine whether the patient is at high risk for developing heart failure symptoms that may ultimately require hospitalization. The patient and / or the patient's physician may be notified of decisions and appropriate intervention actions that may be taken. For example, the patient's medications may be adjusted, the physician may initiate close outpatient monitoring, and a follow-up visit may be scheduled. Exploring the above example, once a patient is identified as being at high risk for heart failure, they may be upgraded to a wearable monitoring and therapy device, such as a wearable cardioverter-defibrillator (WCD), for extended periods of time. Such a device can monitor the patient's cardiac condition and, if warranted, deliver one or more defibrillation and / or pacing pulse therapies to the patient. During patient use, the wearable cardioverter-defibrillator can continue to collect updated physiological information. This updated information can then be provided to a physician, who can continue to monitor changes in the patient's health by analyzing changes in trends associated with the updated physiological information. For example, this analysis can guide further therapy changes and provide information regarding when the risk of acute heart failure is reduced.

[0062] Many patients with cardiac disease also struggle with managing symptoms related to pleural effusion, resulting in frequent recurrence of acute episodes. These patients may develop congestive heart failure (CHF). Pulmonary congestion is a leading cause of hospitalization and rehospitalization among CHF patients. Therefore, it is desirable to measure and monitor the degree of pulmonary congestion. The radio frequency sensors described herein can enable direct and accurate fluid measurement, for example, by monitoring changes in a patient's thoracic impedance. Such information can be utilized by caregivers in formulating a patient's treatment plan and potentially shorten hospital stays by enabling effective medication balancing. For example, this technology can enable early detection of edema to avoid rehospitalization.

[0063] Similarly, patients with chronic kidney disease (CKD) undergoing hemodialysis also require fluid management strategies. Approximately 60 percent of patients with end-stage renal disease (ESRD) may have moderate to severe pulmonary congestion before hemodialysis. Furthermore, patients diagnosed with CKD and undergoing continuous hemodialysis tend to have a high prevalence of heart failure (HF) and a poor overall prognosis. Thus, congestive heart failure is a frequent clinical symptom in dialysis patients. For these reasons, fluid assessment and management can be useful in addressing CKD and HF. In the present disclosure, a noninvasive wearable monitoring system may be used to integrate and / or analyze this information with biofrequency information, ECG information, and other combined metrics, and to assess thoracic impedance and / or fluid changes.

[0064] Such methods offer several advantages over existing monitoring and predictive techniques. By combining physiological measurements from multiple sensor sources and utilizing various types of analytics to measure as well as verify changes in physiological signals, the present disclosure more robustly and accurately analyzes a patient's current condition and the likelihood that the patient's condition will change. Based on this information, a personalized treatment plan can be developed for the patient and periodically updated, while the patient's condition is periodically monitored for changes, either positive or negative.

[0065] The teachings of the present disclosure may generally be applied to extracting physiological metrics for further processing from external medical monitoring and / or therapy devices (e.g., devices that are not fully implanted within a patient's body). External medical devices may include, for example, ambulatory medical devices that may and are designed to travel with a patient as they go about their daily lives. Exemplary ambulatory medical devices may be wearable medical devices such as wearable cardioverter-defibrillators (WCDs), wearable cardiac monitoring devices, in-hospital devices such as in-hospital wearable cardioverter-defibrillators (HWDs), short-term wearable cardiac monitoring and / or therapy devices, mobile cardiac monitoring devices, and other similar wearable medical devices.

[0066] A wearable medical device may allow for continuous use by a patient. In some implementations, continuous use can be in fact, substantially, or nearly continuous. That is, the wearable medical device may be used continuously, except for sporadic periods during which use is temporarily suspended (e.g., while the patient is bathing, while the patient is changing into new and / or different clothing, while charging / replacing batteries, while washing clothing, etc.). However, such substantially or nearly continuous use as described herein may qualify as continuous use. For example, a wearable medical device may be configured to be worn by a patient for as many as 24 hours a day. In some implementations, a patient may remove the wearable medical device for short periods of time each day (e.g., for 30 minutes to take a bath).

[0067] Furthermore, wearable medical devices can be configured for extended or prolonged use. Such devices can be configured for use by patients for extended periods of days, weeks, months, or years. In some examples, a wearable medical device can be used by a patient for an extended period of at least one week. In some examples, a wearable medical device can be used by a patient for an extended period of at least 30 days. In some examples, a wearable medical device can be used by a patient for an extended period of at least one month. In some examples, a wearable medical device can be used by a patient for an extended period of at least two months. In some examples, a wearable medical device can be used by a patient for an extended period of at least three months. In some examples, a wearable medical device can be used by a patient for an extended period of at least six months. In some examples, a wearable medical device can be used by a patient for an extended period of at least one year. In some implementations, extended use can continue until a physician or other caregiver gives the patient explicit instructions to discontinue use of the wearable medical device.

[0068] Regardless of long-term wear, use of a wearable medical device may include continuous or near-continuous wear by a patient, as described above. For example, continuous use may include continuous wearing or attachment of a wearable medical device to a patient, e.g., through one or more electrodes described herein, during both monitoring periods and periods when the device may not be monitoring the patient but is otherwise still worn by or attached to the patient. A wearable medical device may be configured to continuously monitor a patient for cardiac-related information (e.g., electrocardiogram (ECG) information, including arrhythmia information, cardiac vibrations, pulmonary vibrations, etc.) and / or non-cardiac information (e.g., blood oxygen, the patient's temperature, blood glucose levels, tissue fluid volume, and / or pulmonary vibrations). A wearable medical device may perform its monitoring at periodic or aperiodic time intervals or on a periodic or aperiodic basis. For example, monitoring during such intervals or times may be triggered by a user action or another event.

[0069] As noted above, the wearable medical device may be configured to monitor other physiological parameters of the patient in addition to cardiac-related parameters. For example, the wearable medical device may be configured to monitor, among other things, pulmonary vibrations (e.g., using a microphone and / or accelerometer), respiratory vibrations, sleep-related parameters (e.g., snoring, sleep apnea), and tissue fluids (e.g., using a radio frequency transmitter and radio frequency sensor).

[0070] Other exemplary wearable medical devices include automatic cardiac monitors and / or defibrillators for use in certain special situations and / or environments, such as in a combat zone or inside an emergency vehicle. Such devices may be configured for rapid (or substantially rapid) use in emergency medical care. In some examples, the wearable medical devices described herein may be pacing-enabled, e.g., capable of delivering pacing pulse therapy to a patient.

[0071] In implementations, exemplary therapeutic medical devices may include an in-hospital continuous monitoring defibrillator and / or an in-hospital continuous monitoring pacing device, such as an in-hospital wearable defibrillator. In such examples, electrodes may be adhesively attached to the patient's skin. For example, the electrodes may include disposable adhesive electrodes. For example, the electrodes may include a sensing component and a therapy component disposed in separate adhesive patches for the sensing and therapy electrodes. In some implementations, both the sensing component and the therapy component may be integrated and disposed in the same electrode adhesive patch that is then attached to the patient. In exemplary implementations, the electrodes may include an adhesively attachable anterior therapy electrode, an adhesively attachable posterior therapy electrode, and multiple adhesively attachable sensing electrodes. For example, the adhesively attachable anterior therapy electrode is attached to the front of the patient's torso to deliver pacing or defibrillation therapy. Similarly, the adhesively attachable posterior therapy electrode is attached to the back of the patient's torso. In an exemplary scenario, at least three adhesively attachable ECG sensing electrodes may be attached to at least the upper part of the patient's chest adjacent to the patient's right arm, the upper part of the chest adjacent to the patient's left arm, and toward the patient's lower chest in a manner directed by a trained professional.

[0072] Patients monitored by in-hospital defibrillators and / or in-hospital pacing devices may be confined to their beds or rooms for a significant amount of time (e.g., 90 percent or more of patients stay in the hospital). As a result, the user interface may be configured to interface with users other than the patient, such as nurses, for device-related functions, such as initial device baselines, setting and adjusting patient parameters, and changing device batteries.

[0073] In implementations, examples of therapeutic medical devices may include short-term continuous monitoring defibrillators and / or short-term continuous monitoring pacing devices, such as short-term outpatient wearable defibrillators. For example, such short-term outpatient wearable defibrillators may be prescribed by a physician for patients exhibiting syncope. The wearable defibrillators may be configured to monitor patients exhibiting syncope by, for example, analyzing the patient's cardiac activity for abnormal patterns that may indicate abnormal physiological function. For example, such abnormal patterns may occur before, during, or after the onset of symptoms. In such exemplary implementations of such short-term wearable defibrillators, an electrode assembly may be adhesively attached to the patient's skin, and the electrode assembly may have a similar configuration as the in-hospital defibrillators described above.

[0074] In some implementations, the medical device may be a patient monitoring device without a therapy or treatment function. For example, such a patient monitoring device may include a cardiac monitoring device or cardiac monitor configured to monitor one or more cardiac physiological parameters of a patient, e.g., to remotely monitor and / or diagnose the patient's condition. For example, the cardiac physiological parameters may include the patient's ECG information, cardiac vibrations (e.g., using an accelerometer or microphone), and other related cardiac information. The cardiac monitoring device is a portable device that the patient may carry with them as they go about their daily lives. The cardiac monitor may be configured to detect the patient's ECG via multiple cardiac sensing electrodes. For example, the cardiac monitor may be attached to the patient via at least three adhesive cardiac sensing electrodes positioned around the patient's torso. Such cardiac monitors are used for cardiac event monitoring, e.g., in patient populations reporting irregular cardiac symptoms and / or cardiac diseases. Exemplary cardiac disorders may include atrial fibrillation, bradycardia, tachycardia, atrioventricular block, Launghanon-Levine syndrome, atrial flutter, sinus node dysfunction, cerebral ischemia, syncope, atrial asystole, and / or palpitations. For example, the patient may be prescribed a long-term cardiac monitor, e.g., for 10 to 30 days or longer. In some mobile cardiac event monitoring applications, the portable cardiac monitor may be configured to monitor a patient with a cardiac anomaly substantially continuously, and when such an abnormality is detected, the monitor may automatically transmit data regarding the abnormality to a remote server. The remote server may be located in a 24 / 7 monitoring center, where data is interpreted by trained reviewers and / or caregivers for qualified cardiac conditions, and feedback is provided to the patient and / or designated caregiver via periodic detailed or event-triggered reports. In certain cardiac event monitoring applications, the cardiac monitor may be configured to allow the patient to manually press a button on the cardiac monitor to report symptoms. For example, patients may report symptoms such as congestion, shortness of breath, lightheadedness, racing heartbeat, fatigue, fainting, chest discomfort, weakness, dizziness, and / or vertigo.The cardiac monitor may record predetermined physiological parameters of the patient (e.g., ECG information) for a period of time (e.g., 1 to 30 minutes before and 1 to 30 minutes after the reported symptom). The cardiac monitor may be configured to monitor physiological parameters of the patient other than cardiac-related parameters. For example, the cardiac monitor may be configured to monitor cardiac vibrations (e.g., using an accelerometer or microphone), pulmonary vibrations, respiratory vibrations, sleep-related parameters (e.g., snoring, sleep apnea), and tissue fluid, among others.

[0075] FIG. 1 illustrates an exemplary medical device 100 that is external to a patient 102, portable, and wearable, and configured to implement one or more features described herein. For example, the medical device 100 can be a non-invasive medical device configured to be positioned substantially outside the patient. The medical device 100 can be, for example, a portable medical device that can and is designed to travel with the patient as the patient goes about their daily lives. For example, the medical device 100 described herein, such as the LifeVest® wearable defibrillator available from ZOLL® Medical, Inc., can be attached to the patient's body. The wearable defibrillator is typically worn nearly or substantially continuously for two to three months at a time. While the patient wears the wearable defibrillator, the wearable defibrillator may be configured to continuously or substantially continuously monitor the patient's vital signs, and when a need for treatment is determined, the wearable defibrillator may be configured to deliver one or more therapeutic electrical pulses to the patient. For example, such therapeutic shocks may be pacing, defibrillation, or transcutaneous electrical nerve stimulation (TENS) pulses.

[0076] The medical device 100 may include one or more of the following: a garment 110, one or more sensing electrodes 112 (e.g., ECG electrodes), one or more treatment electrodes 114a and 114b (collectively referred to herein as treatment electrodes 114), a medical device controller 120, a connectivity pod 130, a patient interface pod 140, a belt 150, or any combination thereof. In some examples, at least some of the components of the medical device 100 may be configured to attach to (or, in some examples, be permanently integrated into) the garment 110 that can be worn around the patient's torso.

[0077] The medical device controller 120 can be operably coupled to the sensing electrodes 112, which can be attached to the garment 110, e.g., integrated into the garment 110, or removably attached to the garment, e.g., using hook-and-loop fasteners. In some implementations, the sensing electrodes 112 can be permanently integrated into the garment 110. The medical device controller 120 can be operably coupled to the treatment electrodes 114. For example, the treatment electrodes 114 can also be integrated into the garment 110, or in some implementations, the treatment electrodes 114 can be permanently integrated into the garment 110.

[0078] Component configurations other than that shown in Figure 1 are possible. For example, the sensing electrodes 112 may be configured for attachment to various locations around the body of the patient 102. The sensing electrodes 112 may be operably coupled to the medical device controller 120 via a connecting pod 130. In some implementations, the sensing electrodes 112 may be adhesively attached to the patient 102. In some implementations, at least one of the sensing electrodes 112 and the treatment electrodes 114 may be included in a single integrated patch and may be adhesively affixed to the patient's body.

[0079] The sensing electrodes 112 may be configured to detect one or more cardiac signals. Examples of such signals include ECG signals and / or other sensed cardiac physiological signals from the patient. In certain implementations, the sensing electrodes 112 may include additional components, such as accelerometers, vibration signal detection devices, and other measurement devices that record additional parameters. For example, the sensing electrodes 112 may be further configured to detect other types of patient physiological parameters and vibration signals, such as tissue fluid volume, cardiac vibrations, pulmonary vibrations, respiration-related vibrations, and patient motion, which are anatomical characteristics within the airways. Exemplary sensing electrodes 112 include metal electrodes with an oxide coating, such as tantalum pentoxide electrodes, as described in U.S. Pat. No. 6,253,099, entitled "Cardiac Monitoring Electrode Apparatus and Method," the contents of which are incorporated herein by reference.

[0080] In some examples, the therapy electrodes 114 may be further configured to include sensors configured to detect ECG signals as well as other physiological signals of the patient. In some examples, the connectivity pod 130 may include a signal processor configured to amplify, filter, and digitize cardiac signals before transmitting them to the medical device controller 120. When the medical device 100 determines that such therapy is warranted based on signals detected by the sensing electrodes 112 and processed by the medical device controller 120, one or more of the therapy electrodes 114 may be configured to deliver one or more therapeutic defibrillation shocks to the body of the patient 102. Exemplary therapy electrodes 114, in certain implementations, may include conductive metal electrodes, such as stainless steel electrodes, including one or more conductive gel placement devices configured to supply conductive gel to the metal electrodes prior to delivering the therapeutic shock.

[0081] In some implementations, the medical devices described herein may be configured to switch between a therapeutic medical device and a monitoring medical device configured to only monitor a patient (e.g., not provide or perform any therapeutic functions). For example, therapeutic components, such as the therapeutic electrodes 114, and associated circuitry may be disconnected (or connected) from the medical device or switched off (or on) as needed. For example, the medical device may have optional therapeutic elements (e.g., defibrillation and / or pacing electrodes, components, and associated circuitry) configured to operate in a therapeutic mode. The optional therapeutic elements may be physically disconnected from the medical device as a means of converting the therapeutic medical device into a monitoring medical device for a specific application (e.g., operating only in a monitoring mode) or patient. Alternatively, the optional therapeutic elements may be deactivated (e.g., by physical or software switching), essentially turning the therapeutic medical device into a monitoring medical device for a specific physiological purpose or for a specific patient. As an example of a software switch, to deactivate a therapy element on a medical device, an authorized person may access a protected user interface on the medical device and select a pre-configured option or perform some other user action via the user interface.

[0082] Figure 2 shows a sample component-level diagram of a medical device controller 120. As shown in Figure 2, the medical device controller 120 may include a therapy delivery circuit 202, data storage 204, a network interface 206, a user interface 208, at least one battery 210, a sensor interface 212, an alarm manager 214, and at least one processor 218. A patient monitoring medical device may include a medical device controller 120 that includes similar components as those described above, but does not include the therapy delivery circuit 202 (shown in dashed lines).

[0083] The therapy delivery circuit 202 may be coupled to one or more electrodes 220 (e.g., therapy electrodes 114 described above in connection with FIG. 1 ) configured to deliver therapy to a patient. For example, the therapy delivery circuit 202 may include or be operably connected to circuit components configured to generate and deliver a therapeutic shock. The circuit components may include, for example, resistors, capacitors, relays and / or switches, an electrical bridge such as an h-bridge (e.g., including multiple insulated gate bipolar transistors or IGBTs), voltage and / or current measurement components, and other similar circuit components arranged and connected such that the circuit components cooperate with the therapy delivery circuit and operate under the control of one or more processors (e.g., processor 218) to provide, for example, one or more pacing or defibrillation therapy pulses.

[0084] Pacing pulses may be used to treat cardiac arrhythmias such as bradycardia (e.g., less than 30 beats per minute) and tachycardia (e.g., greater than or equal to 150 beats per minute) using, for example, fixed rate pacing, demand pacing, antitachycardia pacing, etc. Defibrillation pulses may be used to treat ventricular tachycardia and / or ventricular fibrillation.

[0085] The capacitor may include a parallel-connected capacitor bank of multiple capacitors (e.g., two, three, four, or more capacitors) that can be switched into a series connection during the discharge of a defibrillation pulse. For example, four capacitors of approximately 650 microfarads each may be used. The capacitors may have a surge rating in the range of 350 to 500 volts and can be charged from the battery pack in approximately 15 to 30 seconds.

[0086] For example, each defibrillation pulse may deliver energy in the range of 60 to 180 joules. In some implementations, the defibrillation pulse may be a biphasic truncated exponential waveform, whereby the signal may switch (e.g., in the charging direction) between positive and negative portions. This type of waveform may be effective at defibrillating a patient at lower energy levels compared to other types of defibrillation pulses (e.g., monophasic pulses, etc.). For example, the amplitude and width of the two phases of the energy waveform may be automatically adjusted to deliver a precise amount of energy (e.g., 150 joules) regardless of the patient's body impedance. The therapy delivery circuit 202 may be configured to perform the switching and pulse delivery operations, for example, under the control of the processor 218. As energy is delivered to the patient, the amount of energy delivered may be tracked. For example, the amount of energy may be held at a predetermined constant value even when the pulse waveform is dynamically controlled based on factors such as the body impedance of the patient to whom the pulse is being delivered.

[0087] The data storage 204 may include, for example, one or more of flash memory, solid-state memory, magnetic memory, optical memory, cache memory, combinations thereof, and other non-transitory computer-readable media. The data storage 204 may be configured to store executable instructions and data used in the operation of the medical device controller 120. In particular implementations, the data storage may include executable instructions that, when executed, are configured to cause the processor 218 to perform one or more functions.

[0088] In some examples, network interface 206 may facilitate communication of information between medical device controller 120 and one or more other devices or entities over a communications network. For example, if medical device controller 120 is included in a portable medical device (such as medical device 100), network interface 206 may be configured to communicate with a remote computing device, such as a remote server or other similar computing device. Network interface 206 may include communications circuitry that transmits data according to the Bluetooth® wireless standard for exchanging such data over short distances to an intermediary device, such as a base station, a “hotspot” device, a smartphone, a tablet, a portable computing device, and / or other device near the wearable medical device. The intermediary device may then communicate the data to a remote server over a broadband cellular network communications link. This communication link may implement high-speed wireless broadband cellular technologies (e.g., 2.5G, 2.75G, 3G, 4G, 5G cellular standards) and / or Long Term Evolution (LTE) technologies, or GSM / EDGE® and UMTS / HSPA® technologies. In some implementations, the intermediary device may communicate with the remote server via a Wi-Fi® communication link based on the IEEE 802.11 standard.

[0089] In certain implementations, the user interface 208 may include one or more physical interface devices, such as input devices, output devices, and combination input / output devices and software stacks configured to drive the operation of the device. These user interface elements may display visual, auditory, and / or tactile content. Thus, the user interface 208 may receive input or provide output that allows a user to interact with the medical device controller 120.

[0090] The medical device controller 120 may further include at least one battery 210 configured to power one or more components integrated into the medical device controller 120. The battery 210 may include a rechargeable multi-cell battery pack. In one exemplary implementation, the battery 210 may include three or more 2200 mAh lithium-ion cells that power other device components within the medical device controller 120. For example, the battery 210 may provide its power output within a range of 20 mA to 1000 mA (e.g., 40 mA) and may support 24, 48, 72, or more hours of run time between charges. In particular implementations, the battery capacity, run time, and type (e.g., lithium-ion, nickel-cadmium, or nickel-metal hydride) may be varied to best suit a particular application of the medical device controller 120.

[0091] The sensor interface 212 may be coupled to one or more sensors configured to monitor one or more physiological parameters of the patient. As shown, the sensors may be coupled to the medical device controller 120 via a wired or wireless connection. The sensors may include one or more electrocardiogram (ECG) electrodes 222 (e.g., similar to the sensing electrodes 112 described above in connection with FIG. 1), a bio-vibration sensor 224, a tissue fluid monitor 226 (e.g., based on an ultra-wideband radio frequency device), and a patient motion sensor 228.

[0092] The ECG electrodes 222 may monitor the patient's ECG information. For example, the ECG electrodes 222 may be galvanic (e.g., conductive) electrodes and / or capacitive electrodes configured to measure changes in the patient's electrophysiology to determine the patient's ECG information. The ECG electrodes 222 may transmit information describing the ECG signal to the sensor interface 212 for subsequent analysis.

[0093] The biovibration sensor 224 may detect patient vibrations associated with cardiac and pulmonary activity, for example. For example, the biovibration sensor 224 may be configured to detect cardiac or thoracic vibration values, including any one or all of S1, S2, S3, and S4. From these cardiac vibration values, specific cardiac vibration metrics or combination metrics, including any one or more of electromechanical activation time (EMAT), left ventricular contraction time (LVST), or percentage left ventricular contraction time (%LVST), may be calculated. The biovibration sensor 224 may include a vibration sensor configured to detect vibrations from the patient's cardiac system and provide an output signal corresponding to the detected cardiac vibrations. The biovibration sensor 224 may further include a multi-channel accelerometer, for example, a three-channel accelerometer, configured to sense motion in each of three orthogonal axes so that patient motion / body position can be detected and associated with the detected cardiac vibration information. The biovibration sensor 224 may transmit information describing the cardiac vibration information to the sensor interface 212 for subsequent analysis.

[0094] The interstitial fluid monitor 226 may use radio frequency-based technology to assess fluid volume and fluid accumulation within the patient's bodily tissues. For example, the interstitial fluid monitor 226 may be configured to measure fluid content in the lungs, which is typical for diagnosing and following up on pulmonary edema or congestion in patients with heart failure. The interstitial fluid monitor 226 may include one or more antennas configured to direct radio frequencies at the patient's tissue and measure an output radio frequency signal responsive to the radio frequencies transmitted through the tissue. In certain implementations, the output radio frequency signal includes a parameter indicative of the fluid volume within the patient's tissue. The interstitial fluid monitor 226 may transmit information indicative of the interstitial fluid volume to the sensor interface 212 for subsequent analysis. Examples of radio frequency sensors are further described below. Figures 4A and 4B It is written as:

[0095] The patient motion sensor 228 may include one or more accelerometers configured to measure motion data related to patient motion. In certain implementations, the patient motion sensor 228 may be configured to measure the number of steps taken by the patient over a certain period of time. For example, the patient may be instructed to perform a certain exercise, such as a walking test. The patient motion sensor 228 may be configured to measure step count and pace information during the certain exercise. Note, however, that the patient motion sensor 228 is shown as a separate component by way of example only. In certain implementations, the one or more accelerometers included within the patient motion sensor 228 may be integrated into other components, such as the bio-vibration sensor 224 or the interstitial fluid monitor 226.

[0096] The sensor interface 212 may be coupled to any one or combination of sensing electrodes / other sensors to receive other patient data indicative of patient parameters. Once data from the sensors is received by the sensor interface 212, the data may be sent by the processor 218 to appropriate components within the medical device controller 120. For example, if cardiac data is collected by the biovibration sensor 224 and transmitted to the sensor interface 212, the sensor interface 212 may transmit the data to the processor 218, which then relays the data to a cardiac event detector. The cardiac event data may also be stored in the data storage 204.

[0097] In particular implementations, the alarm manager 214 may be configured to manage alarm profiles and notify one or more intended recipients of events specified in the alarm profiles as being of interest to the intended recipients. These intended recipients may include users (patients, physicians, and monitoring personnel) and external entities such as computer systems (monitoring systems or emergency response systems). The alarm manager 214 may be implemented using hardware or a combination of hardware and software. For example, in some examples, the alarm manager 214 may be implemented as a software component stored in the data storage 204 and executed by the processor 218. In this example, instructions included in the alarm manager 214 may cause the processor 218 to configure alarm profiles and notify intended recipients using the alarm profiles. In other examples, the alarm manager 214 may be an application-specific integrated circuit (ASIC) coupled to the processor 218 and configured to manage alarm profiles and notify intended recipients using alarms specified in the alarm profiles. Thus, examples of the alarm manager 214 are not limited to a particular hardware or software implementation.

[0098] In some implementations, processor 218 includes one or more processors (or one or more processor cores), each configured to execute a sequence of instructions that result in manipulated data and / or to control the operation of other components of medical device controller 120. In some implementations, when performing a particular process (e.g., cardiac monitoring), processor 218 may be configured to make particular logic-based decisions based on received input data, and the processor may be further configured to provide one or more outputs that may be used to control or otherwise inform subsequent processing to be performed by processor 218 and / or other processors or circuits to which processor 218 is communicatively coupled. Thus, processor 218 reacts in a particular manner to particular input stimuli and generates corresponding outputs based on the input stimuli. In some example cases, processor 218 may undergo a series of logical transitions that may set various internal register states and / or other bit cell states internal or external to processor 218 to logic high or logic low. The processor 218 referred to herein may be configured to perform functions for which software is stored in a data store coupled to the processor 218, the software being configured to cause the processor 218 to undergo a series of various logical decisions that result in the functions being performed. The various components described herein as executable by the processor 218 may be implemented in various forms, such as specialized hardware, software, or a combination thereof. For example, the processor may be a digital signal processor (DSP), such as a 24-bit DSP processor. The processor may be a multi-core processor, for example, having two or more processing cores. The processor may be an advanced RISC machine (ARM) processor, such as a 32-bit ARM processor or a 64-bit ARM processor.This processor may run an embedded operating system and may include services provided by the operating system that may be used for file system operations, display and audio generation, basic networking functions, firewall functions, data encryption, and communications.

[0099] FIG. 3 illustrates a sample network 300 showing a sample topology illustrating how a wearable medical device (e.g., the medical device 100 described above) may be operatively connected to a remote server. As described above in connection with FIG. 2, the wearable medical device controller 120 includes a network interface 206 for transmitting data via a wireless link, such as a Bluetooth® wireless link (e.g., via a “hotspot” or other base station or intermediary device), a broadband cellular link, or a Wi-Fi® communication link based on the IEEE 802.11 standard. As shown in FIG. 3, the physician computer 302 and the wearable medical device 304 may be operatively connected to a monitoring server 308 via the network 306. In certain implementations, while worn, the wearable medical device 304 may collect information about the patient, such as various patient metrics and parameters described herein. Depending on the connection to the network 306 and the programming of the wearable medical device 304, the wearable medical device may be configured to periodically transmit collected information to the monitoring server 308 for further processing. For example, the monitoring server 308 may be configured to monitor the patient's physiological condition based on patient data obtained from multiple sensor sources, including bio-vibration sensors and radio frequency sensors, and transmitted to the monitoring server by, for example, one or more wearable medical devices 304. In some examples, one or more physicians may use the physician computer 302 to access the patient's condition information to review changes in the patient's health condition, receive instructions / recommendations to modify the patient's treatment plan, and perform other similar functions.

[0100] In some examples, as described above, the patient may wear a radio frequency device, e.g., a radio frequency transceiver such as an ultra-wideband transceiver circuit configured to direct radio frequency electromagnetic energy toward the patient via a radio frequency antenna and produce radio frequency information corresponding to the reflected electromagnetic energy. In certain implementations, the radio frequency device (transceiver and associated antenna) may be integrated into a patch worn by the patient in addition to the wearable medical device described above. For example, as shown in FIG. 4A , a patch 402 may be adhesively attached to the patient 400 at a predetermined location on the patient's body.

[0101] In other implementations, the radio frequency device can be incorporated into a wearable defibrillator garment. In some implementations, the radio frequency device can be incorporated into a wearable continuous event monitoring device. In some implementations, the radio frequency device can be incorporated into one or more treatment electrodes and / or patches of a hospital wearable defibrillator. In the above implementations, the radio frequency device (e.g., incorporated into patch 402) can be positioned on the patient 400's lower left abdominal region as shown in FIG. 4A and configured to direct radio frequency electromagnetic energy into the patient's thoracic cavity, e.g., toward the patient's lungs and / or heart. In another example, the radio frequency device can be placed over at least one major artery, such as the radial artery, brachial artery, aorta, and one or more pulmonary arteries, to monitor the patient's arterial pulse. Information about the arterial pulse can be used to monitor the patient's blood pressure measurements according to implementations described below.

[0102] Radio frequency information corresponding to reflected electromagnetic energy from the patient's thoracic cavity may indicate lung fluid volume (absolute value and volume change), thoracic impedance, blood pressure measurements, heart rate measurements, and / or specific cardiac conditions based on monitoring the patient's cardiac wall motion. Examples of radio frequency transceivers are described, for example, in U.S. Pat. No. 8,989,837, entitled "Methods and Systems for Determining Fluid Content of Tissue," U.S. Pat. No. 7,122,012, entitled "Detection of Fluids in Tissues," U.S. Patent Application Publication No. 2010 / 0256,462, entitled "Method and System for Monitoring Thoracic Tissue Fluid," U.S. Pat. No. 9,675,251, entitled "Electromagnetic Probes, Methods for Fabricating Thereof, and Systems Which Use Such Electromagnetic Probes," U.S. Pat. No. 7,725,150, entitled "System and Method for Extracting Physiological Data Using Ultra-Wideband Radar and Improved Signal Processing Techniques," and U.S. Pat. No. 9,002,427, entitled "Apparatus and Method for Continuous Noninvasive Measurement of Respiratory Function and Events," the contents of which are incorporated herein by reference. As noted above, radio frequency information corresponding to reflected electromagnetic energy from the patient's thoracic cavity may similarly indicate changes in the patient's thoracic impedance values.

[0103] 4B shows a schematic diagram of an exemplary radio frequency device 430, such as the radio frequency device shown in FIG. 4A. The radio frequency device 430 may be configured to function as a radio frequency transceiver using components contained within an integrated package 442, which may have the form of a patch, for example, or be integrated into a wearable medical device as described above. In some implementations, the radio frequency device 430 may be packaged within a package 442 having a form factor of approximately 20 mm x 50 mm. In other examples, the form factor may be smaller, for example, within a range of approximately 10 mm x 25 mm to approximately 20 mm x 50 mm. In some examples, the package 442 may include an adhesive layer, such as by which the radio frequency device 430 may be attached to a patient's skin.

[0104] The radio frequency device 430 includes at least two antennas, a transmitting antenna 432a and a receiving antenna 432b, that transmit radio frequency energy to and receive radio frequency energy from the patient's thoracic cavity. In some implementations, the antennas 432a and 432b may have flat and / or flexible contours to more closely conform to the shape and contours of the patient's body. The transceiver 436 is configured to generate drive signals that transmit and / or receive radio frequency energy at the antennas 432a and 432b. The transceiver 436 is further configured to receive and process reflected electromagnetic energy received from the patient's body by the antennas 432a and 432b. In certain implementations, the active background cancellation circuit 434 cancels background components from reflected signals, e.g., signals from depths and / or tissues of no interest. The processor 438 controls the operation of the components of the radio frequency device 430, as described below. The processor 438 may monitor one or more radio frequency signal path characteristics and generate radio frequency information corresponding to a selected tissue depth (e.g., within a range of 1 cm to 20 cm) and / or tissue of interest, such as the effective path length of the radio frequency path and one or more phase and / or amplitude changes of the reflected radio frequency relative to the transmitted radio frequency. The processor 438 may process this radio frequency information and convert it into interpretable physiological metrics, such as thoracic fluid content information (TFC), cardiac wall motion data, and / or arterial pulse information. In some implementations, the radio frequency device 430 may be configured to transmit the radio frequency information to a remote server via the communication interface 448. In such implementations, the remote server may be configured to process the received radio frequency information and perform the conversion into the physiological metrics described above. A power module 444, such as a thin battery, provides power to the components of the patch unit.

[0105] In some implementations, the radio frequency device 430 may further include ECG electrodes 440 in electrical contact with the patient's skin and ECG acquisition circuitry 450 that filters and digitizes the ECG signal for input to the processor 438. The ECG signal may be used to gate the transmission of radio frequencies to and reception of radio frequencies from the patient's thoracic cavity. For example, the transceiver 436 may trigger radio frequency transmissions based on the timing of the R waves of the ECG signal.

[0106] In certain implementations, the radio frequency device 430 includes a user interface, such as one or more indicator LEDs 446, that indicate the operational status of the patch (on / off, and possibly parameters such as battery level, skin contact characteristics, or signal strength). Alternatively or additionally, the user interface may include a more informative display, such as an LCD, and user controls, such as an on / off button and an adjustment button.

[0107] The communication interface 448 communicates with a remote console to transmit the radio frequency data and ECG measurement data and, in some cases, receive operational commands. For example, the communication interface 448 typically includes a wireless link such as a Bluetooth® or WiFi link. For example, the remote console may be located near the patient's location and thus directly receive and process data from the communication interface 448. Alternatively, the communication interface 448 may communicate with a local gateway, such as a personal computer, smartphone device, or dedicated "hotspot" device, that communicates with a remote server over a network, such as the Internet or a telephone network. In these implementations, the console may be a network server connected to one or more databases and configured to store the radio frequency data and ECG data for subsequent reference, processing, and analysis. For example, a physician or other specialist may be provided with access and / or tools to perform data analysis. In some situations, the data may be available to a separate diagnostic computer system via a wired or wireless communication link. These system configurations are particularly useful for ambulatory monitoring of multiple patients over extended periods of time in various geographic locations.

[0108] As noted above, radio frequency devices such as device 430 may be used to monitor various patient metrics and parameters. For example, the radio frequency device may be configured to monitor tissue fluid metrics, blood pressure, chest wall motion, and other similar patient parameters.

[0109] To measure tissue fluid metrics, a radio frequency device may be configured to direct electromagnetic energy into a tissue region (e.g., the lung or thoracic cavity) and measure the amplitude and delay of the returned / reflected electromagnetic energy. For example, FIG. 5A illustrates a schematic plot 500 showing the amplitude, and FIG. 5B illustrates a schematic plot 505 showing the propagation delay of radio frequency waves reflected from a patient's heart. Note that the scales of plots 500 and 505, as shown in FIGS. 5A and 5B, are arbitrary. The delay and amplitude vary periodically, albeit to a lesser extent, with each heart cycle, as particularly shown by the sharp peak in line 507 shown in FIG. 5B and the general downward shift in amplitude in line 502 shown in FIG. 5A.

[0110] The depression in the horizontal scale from 250 to 260 on both lines 502 and 507 corresponds to the inhalation period during the respiratory cycle. This depression in FIG. 5B indicates that when the lungs fill with air, the effective radio frequency path length through the lungs narrows because the average dielectric constant along that path decreases while the physical distance between the antenna and the heart remains approximately the same. Exhalation empties the lungs of air, widening the effective radio frequency path length. Furthermore, the amplitude of the reflected wave in FIG. 5A decreases during inhalation, likely due to greater variation in dielectric constant and therefore more reflections along the radio frequency path through the lungs when the lungs fill with air.

[0111] Lungs with a large liquid content generally have a higher average dielectric constant than healthy lungs, resulting in a larger radio frequency path delay across the lungs. Because radio frequencies traverse the lungs with fewer reflections, the overall amplitude may also be larger. Meanwhile, during a respiratory cycle, the difference between an air-filled lung and an airless lung is expected to be smaller in both amplitude and delay than the difference shown in FIGS. 5A and 5B. Therefore, to monitor tissue fluid content, the processing device may, for example, compare the delay and possibly the amplitude of the reflected wave with a reference provided by healthy and unhealthy lungs or with previous measurements taken on the same patient. Additionally or alternatively, the processing device may assess the amount of fluid in the lungs by analyzing changes in the amplitude and / or delay of the reflected wave during one or more respiratory cycles. In certain implementations, to quantify the assessment of fluid accumulation, the actual physical distance traversed by the radio frequency passing through the lungs may be measured, and a relationship (e.g., ratio) between the effective radio frequency path length and that physical distance may be calculated.

[0112] Additionally, the radio frequency device may be used to measure a patient's blood pressure. For example, information determined from the patient's arterial pulse waveform may be used by the processing device to determine the patient's blood pressure. As shown in FIG. 6, a patient's arterial pulse waveform 600 may be obtained by directing a radio frequency wave into the patient's artery and measuring the reflected radio frequency wave. Based on the reflected wave, the processing device may measure changes in the artery's radar cross section (RCS).

[0113] For example, during a cardiac cycle, a radio frequency device may generate and transmit radio frequencies toward an artery that may be located at a particular depth from the radio frequency device. In some examples, some or all of the transmitted radio frequencies may bounce back to the radio frequency device. In some cases, the radio frequency device may transmit radio frequencies continuously or non-continuously. During a cardiac cycle, the diameter of the artery may vary over time, and the RCS of the artery obtained by the radio frequency device may similarly change over time. In some implementations, an arterial pulse waveform (e.g., pulse waveform 600) indicative of pulse waves propagating through the artery may be determined from various RCS measurements. Various clinical information can then be obtained from the arterial pulse waveform, such as, but not limited to, measurements (continuous or non-continuous) of arterial stiffness, pulse wave velocity, cardiac output, and blood pressure. In some embodiments, the reflected echoes may be modulated by the artery during the cardiac cycle, and information from the reflected echoes may be utilized to determine / evaluate the arterial pulse waveform. For example, the measurement range may change during the cardiac cycle, resulting in a change in the phase of the reflected waves. In such cases, the information can be used to determine / assess the arterial pulse waveform.

[0114] In certain implementations, to more accurately determine a patient's blood pressure, a radio frequency device can be calibrated to include baseline measurements of the patient's blood pressure and pulse transit time (PTT). PTT can be obtained using multiple sensors attached at known locations on the patient and measuring the pulse arrival time (PAT) at each location. By calculating the difference in PAT at each location and knowing the distance between the sensors, the processing device can calculate PTT. This information can then be used to determine the patient's pulse wave velocity (PWV).

[0115] In a particular implementation, a linear transformation relating systolic blood pressure (SBP) and diastolic blood pressure (DBP) to PTT may be expressed as: SBP = (a × PTT) + b, DBP = (c × PTT) + d The coefficients a, b, c, and d may be calibrated for each patient. In some embodiments, other types of transformations may be used to calculate blood pressure. For example, in a model assuming a constant thickness and constant radius of the arteries, the blood pressure P may be expressed as P = a × ln(PTT) + b, where a and b are again constants that are calibrated for each patient. In some embodiments, obtaining the PTT of the pulse in the artery, or the PWV, in either case, may be used to determine the blood pressure value in the artery. Additional examples of determining blood pressure using radio frequency devices are described, for example, in U.S. Patent Application Publication No. 2016 / 0345845, entitled "Systems, Apparatuses and Methods for Determining Blood Pressure," the contents of which are incorporated herein by reference.

[0116] From information about the collected parameters and metrics described above (e.g., tissue fluid volume and blood pressure), additional metrics may be determined. For example, the processing device may extract respiratory information from the fluid measurement information to determine information such as chest wall movement and respiratory rate. For example, the processing device may be configured to associate chest wall expansion with periods when the patient is inhaling and chest wall contraction with periods when the patient is exhaling.

[0117] Cardiac vibrations are noises generated by the heartbeat and the resulting blood flow caused by the heartbeat. Specifically, the vibrations reflect the turbulence created when the heart valves close. During cardiac monitoring, for example, one or more cardiac vibration sensors may be used to detect these unique and distinct vibrations, which provide important auditory data regarding cardiac conditions.

[0118] In healthy adults, there are at least two normal cardiac oscillations, often described as a thump or thump, that occur sequentially with each heartbeat. For example, the first cardiac oscillation (S1) and the second cardiac oscillation (S2) are produced by the closure of the atrioventricular (AV) and semilunar (SL) valves, respectively. More specifically, the S1 oscillation represents the closure of the AV valves, including the tricuspid valve located between the right atrium and the right ventricle and the mitral valve located between the left atrium and the left ventricle. The S2 oscillation represents the closure of the SL valves, including the pulmonary valve located between the right ventricle and the pulmonary artery, which pumps blood to the lungs to obtain oxygen, and the aortic valve located between the left ventricle and the aorta, which pumps oxygenated blood to the body.

[0119] Systole refers to the portion of the cardiac cycle when the ventricles contract. Diastole is the portion of the cardiac cycle when the ventricles relax and refill with blood after systole. Similarly, atrial diastole is the period when the atria relax. During ventricular diastole, the pressure in the left and right ventricles decreases from the peak reached during systole (e.g., 120 mmHg in a normal heart). When the pressure in the left ventricle decreases below the pressure in the left atrium, the mitral valve opens, allowing pooled blood from the atrium to flow into the ventricle.

[0120] When the smaller upper atrial chamber contracts during late diastole, it pumps blood into the larger lower ventricular chamber. When the lower ventricular chamber fills and the valves to the atria close, the ventricles undergo isovolumic contraction (contraction of the ventricles while all valves are closed), marking the first phase of systole. The second phase of systole pumps blood from the left ventricle to the aorta and the body's extremities, and from the right ventricle to the lungs. Thus, the atria and ventricles contract in alternating order. The left and right atria simultaneously pump blood into the ventricles. The left and right ventricles then similarly contract simultaneously.

[0121] Table 1 below provides an overview of the cardiac cycle. Table 1 [Table 1]

[0122] In addition to S1 and S2 vibrations, S3 and S4 vibrations may be detected with a sensitive sensor, such as a vibration sensor. Typically, an examining physician performing cardiac auscultation (e.g., listening to internal heart vibrations with a stethoscope) will not hear S3 vibrations in a healthy adult.

[0123] The third cardiac oscillation, S3, occurs immediately after the two normal cardiac oscillations (i.e., S1 and S2). S3 oscillations typically occur at the beginning of the middle third of diastole, approximately 0.12 to 0.18 seconds after S2. This produces a classic rhythm that has been compared to the cadence of the word "Kentucky," whose final syllable indicates S3 oscillation. S3 oscillations are lower in pitch than S1 or S2 oscillations because they do not originate from the heart valves. While S3 oscillations are generally harmless in young people, some trained athletes, and occasionally pregnant women, their reappearance later in life may signal a cardiac disorder, such as left ventricular failure, as seen in diastolic congestive heart failure (CHF). S3 oscillations are thought to be caused by the back-and-forth movement of blood between the ventricular walls and the atria as blood rushes through them. The reason the third heartbeat doesn't occur until the middle third of diastole is probably because the ventricles aren't filling enough to generate enough pressure to resonate during early diastole. Generally, S3 is associated with heart failure on indirect stethoscope testing because its detection indicates that the ventricles are stiffer than those of a normal heart, where the vibrations of blood beating against the ventricular walls are audible.

[0124] The intensity of the third cardiac vibration (S3 intensity) is based on the strength and duration of the vibration. For example, a sensitive vibration sensor configured to detect S3 vibrations may provide an S3 vibration intensity value in the range of 0 to 10. In some implementations, if this intensity value is equal to or exceeds 5.0, the reporting process may indicate that S3 is present.

[0125] The fourth cardiac oscillation, S4, is produced by the oscillation of blood rushing through stiffened or hypertrophied ventricles. When heard with a stethoscope, the S4 oscillation is a sign of pathological conditions, typically left ventricular failure or hypertrophy, as seen in systemic hypertension, severe aortic stenosis, and hypertrophic cardiomyopathy. The oscillation occurs just after atrial contraction at end-diastole and just before the S1 oscillation, producing a rhythm sometimes referred to as the "Tennessee" gallop.

[0126] As described above, wearable medical devices may be configured to monitor other physiological parameters of a patient in addition to cardiac-related parameters. In certain implementations, the wearable medical device may be configured to monitor using a vibration sensor, such as a microphone and / or an accelerometer, located on the patient's chest region. In some implementations, the vibration sensor may be configured to detect cardiac vibrations (S1, S2, S3, and S4 vibrations, heart murmurs), pulmonary vibrations, respiratory / chest wall motion, sleep-related parameters (e.g., snoring, sleep apnea), and other similar vibration-based parameters.

[0127] 7 shows a sample schematic including a vibration sensor 700. In implementations, the vibration sensor 700 may include a high-fidelity diaphragm 705, such as a dynamic electret capacitor, ribbon-based, or piezoelectric crystal-based diaphragm. Multiple motion sensors 710 (e.g., at least two, four, six, or more) may be separately positioned around the periphery of the diaphragm 705. Vibration signals from the diaphragm 605 and the multiple motion sensors 710 may be digitized by a series of analog-to-digital converters (ADCs) 715 and processed via a digital signal processing unit 720. For example, the digital signal processing unit may include a series of digital filters.

[0128] For example, the motion sensor 710 may include a multi-axis accelerometer that produces a digitized signal that is input to the digital signal processing unit 720 along with a digitized diaphragm signal.

[0129] Signals from one or more of the motion sensors 710 may be used to monitor low frequency vibrations, including, for example, respiratory and / or chest wall motion.

[0130] When monitoring vibration signals in the high frequency range (e.g., specific lung vibrations, heart murmurs, etc.), the diaphragm signal may be analyzed as described below. To allow better isolation of the high frequencies of interest, signals from peripheral motion sensors 710 may be summed to determine a common mode signal. The common mode signal may indicate low frequency components that may then be filtered out from the diaphragm signal.

[0131] Tables such as Table 2 shown below may be stored in the memory of the sensor 700. A typical sampling rate ranging from 44.1 KHz to approximately 60 KHz may be implemented in the ADC to convert the acquired diaphragm vibration signal into a digital vibration signal. In some examples, the ADC 715 may implement a sample size ranging from 12 bits to 16 bits, which allows for greater dynamic distance resolution. Various digital filters may be performed on the digital input signal to reject interference signals, such as the 60 Hz component and common-mode rejection signals described above. The digitized vibration signal may then be analyzed to determine various vibrations based on the table. For example, frequency-selective filters and frequency-selective circuits may operate in the digital domain to isolate frequency ranges of interest according to the following table: Exemplary digital filter techniques may include fast Fourier transform (FFT), discrete cosine transform (DCT), and infinite impulse response (IIR) filters, among others. The processor may be configured to receive an indication indicating the type of symptom to be monitored. Based on the indicated symptoms, the processor 725 may retrieve relevant frequency ranges from memory and instruct the digital signal processing unit 720 to isolate the relevant signals for analysis and input to the next stage. Table 2 [Table 2]

[0132] In certain implementations, vibration analysis of vibrations detected via a chest vibration sensor, such as sensor 700 described above, can provide information about characteristic patterns of vibration. Vibration analysis can include monitoring vibrations ranging from one-tenth of a hertz to approximately one hertz to monitor low-frequency chest cavity motion, such as respiratory motion, chest wall motion, and possibly cardiac wall motion. For example, if the sensor is substantially aligned with the patient's ventricular apex, a sensor implementing vibration analysis can detect and monitor ventricular wall motion. Other vibration patterns can be monitored similarly.

[0133] In certain implementations, multiple frequency ranges of lung vibration may be monitored at once. For example, obstruction can produce chest vibration movements, such as paradoxical breathing. During inhalation, the diaphragm descends to its limit, pushing abdominal contents down and out, creating negative pressure. This negative pressure pulls the chest wall inward, resulting in abdominal expansion and chest sagging during inhalation, which is the opposite of normal breathing. During exhalation in paradoxical breathing, the chest expands and the abdomen sags, again the opposite of normal breathing. The more compliant the chest wall, as seen in infants, the more visible the chest movement. During paradoxical breathing, when the tongue is retracted into the airway, the lower jaw is pulled further back with each breath, often resulting in noisy breathing. The patient may exhibit, for example, wheezing, snoring, and / or grunting. Thus, wheezing-related frequencies (>500 Hz) can be monitored along with very low frequency vibrations (0.2 Hz) of the chest wall and / or abdomen that may indicate paradoxical respiratory movements.

[0134] For example, certain lung vibrations range from approximately 100 Hz to 5000 Hz (e.g., tracheal vibrations), >500 Hz (e.g., wheezing), >100 Hz to 5000 Hz (e.g., wheezing), ~Breath sounds 802 have characteristic patterns at various frequencies, including 150 Hz (e.g., blisters), and <350 Hz (e.g., pleural rubs). FIG. 8 illustrates various characteristic patterns of specific lung vibrations. As shown in FIG. 8, a list of breath sounds 802 may have one or more associated amplitude-time plots 804. Each breath sound 802 may include a unique set of acoustic characteristics. For example, as shown in FIG. 8, normal lung vibrations may be characterized as a low-pass filtered noise with a general frequency of 100 Hz-1000 Hz, with measured energy falling to 200 Hz. Various other lung vibrations, such as wheezes, blisters, and pleural rubs, have similar characteristics, as shown in FIG. 8. Similarly, each breath sound 802 may have an associated amplitude-time plot 804, shown in FIG. 8 as two non-extended time plots 806 and extended time plots 808. 8, the extended time plot 808 of a particular breath sound 802 may correspond to a portion of the non-extended time plot 806 indicated by the dashed box. Additional pulmonary symptoms, such as crackles, squawks, glottal sounds, pharyngeal vibrations, and other similar lung vibrations, and their corresponding frequencies, may be cited from "Fundamentals of Lung Auscultation," by A. Bohadana, G. Izbicki, and S. Kraman, in the New England Journal of Medicine (2014).

[0135] The frequencies associated with cardiac vibrations and murmurs generally range from approximately 20 Hz to 500 Hz. Low-frequency cardiac vibrations are those whose dominant frequency is less than approximately 100 Hz, such as the S3, S4, and diastolic murmurs of mitral stenosis. Certain cardiac murmurs have higher frequency components, such as aortic regurgitation, whose dominant frequency is approximately 400 Hz. Patient metric analytics to track patient health

[0136] As described above, in various scenarios relevant to the present disclosure, a patient monitoring system may include sensors configured to collect patient physiological signals other than ECG signals. For example, a vibration sensor may be configured to collect bio-vibration signals, such as cardiac and pulmonary vibration signals. Furthermore, a radio-frequency sensor, such as an ultra-wideband transceiver circuit, may be configured to collect information corresponding to radio-frequency electromagnetic energy reflected from within the patient's chest cavity and / or heart. In implementations described herein, the processing device may further process the collected signals and associated information from the underlying physiological sensors to generate one or more combined physiological parameters, such as a physiological parameter determined based on two or more underlying physiological sensors and associated information. At least one of predictive analysis (e.g., utilizing machine learning processes or artificial neural networks) and trend analysis (e.g., correlation analysis) may be performed on the collected signals and received information, as well as the combined metrics, to determine the patient's current condition and one or more clinically actionable events.

[0137] More specifically, in certain implementations, a patient monitoring system may include an ECG sensor coupled to a patient and configured to detect one or more ECG signals of the patient, a vibration sensor coupled to the patient and configured to detect one or more cardiac vibration signals of the patient, an ultra-wideband radio frequency transceiver circuit including one or more radio frequency antennas coupled to the patient and configured to cause the one or more radio frequency antennas to direct radio frequency electromagnetic energy toward and receive via the one or more radio frequency antennas radio frequency information corresponding to reflected radio frequency electromagnetic energy reflected from within the patient's thoracic cavity, and one or more processors. The one or more processors may be configured to process a) the one or more ECG signals, b) the one or more cardiac vibration signals, and c) the radio frequency information to generate a plurality of physiological parameters of the patient, including one or more combined physiological parameters. The one or more processors may further be configured to perform at least one of predictive analysis and trend analysis of the plurality of physiological parameters, including the combined physiological parameter, to determine a current clinical condition of the patient, wherein the trend analysis includes determining the existence of a substantial relationship among changes in the plurality of physiological parameters, compare the current clinical condition of the patient with predetermined clinically actionable criteria to determine one or more clinically actionable events, and cause an output regarding the one or more clinically actionable events to be provided to an output device.

[0138] In another implementation, a patient monitoring system may include at least one vibration sensor coupled to a patient and configured to detect one or more cardiac vibration signals, at least one ultra-wideband radio frequency transceiver coupled to the patient, and one or more processors. The at least one ultra-wideband radio frequency transceiver may be configured to direct radio frequency electromagnetic waves toward the patient's lungs and detect radio frequency information responsive to the radio frequency electromagnetic waves passing through the lungs. The one or more processors may be configured to process the detected one or more cardiac vibration signals over a predetermined period of time to determine at least one cardiac vibration metric for the patient, process the patient's radio frequency information over a predetermined period of time to determine at least one lung fluid metric for the patient, determine an output related to one or more clinically actionable events based on the determined at least one cardiac vibration metric and the determined at least one lung fluid metric, and cause an output device to provide the output.

[0139] In another implementation, a patient monitoring system may include an ECG sensor coupled to the patient and configured to detect one or more ECG signals of the patient, a vibration sensor coupled to the patient and configured to detect one or more cardiac vibration signals of the patient, an ultra-wideband radio frequency transceiver circuit including one or more radio frequency antennas and coupled to the patient, and one or more processors. The ultra-wideband radio frequency transceiver may be configured to cause the one or more radio frequency antennas to direct radio frequency electromagnetic energy toward and receive via the one or more radio frequency antennas and produce radio frequency information corresponding to reflected radio frequency electromagnetic energy reflected from within the patient's thoracic cavity. The one or more processors may be configured to process the one or more ECG signals, the one or more cardiac vibration signals, and the radio frequency information to generate a plurality of physiological parameters of the patient including one or more combined physiological parameters, perform a trend analysis of the plurality of physiological parameters including the combined physiological parameters to generate a trend result, update a monitoring schedule for the patient based on the trend result, determine a current clinical condition of the patient based on the trend result, compare the patient's current clinical condition with predetermined clinically actionable criteria to determine one or more clinically actionable events, and provide an output regarding the one or more clinically actionable events to an output device.

[0140] In another implementation, a patient monitoring system may include an ECG sensor coupled to the patient and configured to detect one or more ECG signals of the patient, a vibration sensor coupled to the patient and configured to detect one or more cardiac vibration signals of the patient, an ultra-wideband radio frequency transceiver circuit including one or more radio frequency antennas and positioned on at least one aorta of the patient, and one or more processors. The ultra-wideband radio frequency transceiver circuit may be configured to cause the one or more radio frequency antennas to direct radio frequency electromagnetic energy toward and receive via the one or more radio frequency antennas radio frequency electromagnetic energy reflected from within the at least portion of the patient and produce radio frequency information corresponding to the reflected radio frequency electromagnetic energy. The one or more processors may be configured to process the one or more ECG signals, the one or more cardiac vibration signals, and the radio frequency information to generate a plurality of physiological parameters of the patient including one or more combined physiological parameters, perform at least one of a predictive analysis and a correlation analysis of the plurality of physiological parameters including the combined physiological parameters to determine a current clinical condition of the patient, compare the current clinical condition of the patient with predetermined clinically actionable criteria to determine one or more clinically actionable events, and provide an output regarding the one or more clinically actionable events to an output device.

[0141] As described herein, one or more processes may be utilized to perform predictive and trend analysis of a patient. For example, a predictive analysis process utilizing data from a patient's medical devices, including information regarding recorded ECG data, recorded vibration data, and radio frequency information collected for the patient, may be generated for a variety of different cardiac diseases. The predictive analysis may be utilized to determine the patient's current symptoms, any clinically actionable events, and the patient's likely outcomes. FIG. 9 shows a sample overview of a predictive analysis process. A metric set 900, including Metric 1, Metric 2, and Metric 3, may be provided to the predictive process 905. Depending on the design of the processing and predicted output, various types of programming may be used for the predictive process 905. For example, the predictive process 905 may utilize machine learning and / or artificial neural networks, both of which are described in more detail below. By processing the input metrics 900, the predictive process 905 may produce one or more outputs 910. For example, the output 910 may include the patient's current symptoms, any likely adverse events that may occur to the patient in the near future (e.g., cardiac events that may occur within the next 30 days), and any clinically actionable actions for or regarding the patient. In some implementations, the output 910 may provide feedback to the prediction process so that the process learns from results and makes adjustments to the prediction score based on changes in the patient's symptoms.

[0142] FIG. 10 shows a sample overview of a trend analysis process. Process 1000 may include monitoring a first metric 1005. If a change is detected in the first metric 1005, or if the first metric exceeds a certain threshold, process 1000 may trigger monitoring of a second metric 1010. Similarly, if a change is detected in the second metric 1010, or if the second metric exceeds a certain threshold, process 1000 may trigger monitoring of a third metric 1015. If process 1000 detects a change in the third metric 1015, the process may output 1020 an indication, for example, to the patient or the patient's physician. For example, the output may include a recommended change in the treatment regimen, such as changing the patient's medication. As shown, output 1020 may be fed back into the trend analysis process to adjust the process's response to changes in the underlying metrics.

[0143] It should be noted that the overviews shown in Figures 9 and 10 are provided by way of example only. For example, three metrics are shown in both overviews as an example of the number of metrics that may be used. In practice, various other numbers of input metrics may be used for both the predictive analysis process and the trend analysis process.

[0144] 9 and 10, Metric 1 is an ECG metric, Metric 2 is a bio-vibration metric (including cardiac vibrations, pulmonary vibrations, and low-frequency patient motion data), and Metric 3 is a radio-frequency-based physiological metric. Alternatively or additionally, one or more of Metrics 1, 2, and 3 include a combined metric derived from one or more of the ECG metric, bio-vibration metric, and radio-frequency-based physiological metric.

[0145] 11 and 12 illustrate a patient monitoring system that indicates clinically actionable events based on changes in a patient's physiological information, and that may determine, analyze, and / or classify, and notify about clinically actionable events according to predetermined clinically actionable criteria.

[0146] 11 , the system 1100 receives various signals from the patient 1105. For example, the system 1100 may receive the patient's ECG signal 1110, bio-oscillation signal 1115, and radio frequency signal 1120, for example, from a wearable medical device coupled to the patient 1105. As shown, the output 1155 of the predictive analysis classifier or trend analysis classifier 1150 may be fed back to the predictive analysis classifier or trend analysis classifier 1150 in a closed-loop manner. For example, monitoring may begin with only one set of physiological parameters, such as the ECG signal 1110. Over time, or in response to a particular symptom described herein, monitoring may begin with an additional set of physiological parameters other than the ECG parameters, such as the bio-oscillation signal 1115 or the radio frequency-based signal 1120 (or a combination of these parameters).

[0147] In particular implementations, the ECG signal 1110 may be processed to provide ECG-based metrics 1125, such as, but not limited to, heart rate, heart rate variability, ST elevation, premature ventricular contractions (PVCs), cardiac rhythm morphology, and other similar ECG metrics. Specific details of the extraction of the ECG-based metrics 1125 are described in more detail below. Similarly, the bio-oscillation signal 1115 may be processed to provide bio-oscillation metrics 1130, such as cardiac vibration intensity values, pulmonary vibration intensity values, LVST values, and other similar bio-oscillation values. Additionally, the radio frequency signal 1120 may be processed to provide radio frequency metrics 1135, such as pleural fluid volume values, cardiac wall motion metrics, and blood pressure. The metrics 1125, 1130, and 1135 may be combined into a single set of physiological metrics 1145.

[0148] Furthermore, the above physiological metrics may be expanded to include combined physiological metrics. For example, a combined physiological metric may include a metric derived from two or more of metrics 1125, 1130, and / or 1135. One example of a combined physiological metric is EMAT. EMAT quantifies the time from the onset of the Q wave in an ECG signal determined from the S1 cardiac oscillation to mitral valve closure. Generally, long-term EMAT values ​​are associated with reduced left ventricular contractility. Another example of a combined physiological metric is percentage EMAT (%EMAT). %EMAT is calculated as EMAT divided by the dominant RR interval in the ECG signal and indicates the efficiency of the heart's pumping function. A %EMAT of >15% may be sufficient to predict the risk of readmission due to heart failure at and after patient discharge. In an example, heart sound vibration parameters derived from cardiac or chest vibration signals include LVST. LVST may be calculated as the interval from S1 to S2 (S1-S2). A combined metric based on LVST and RR interval information from a patient's ECG signal may include %LVST. For example, %LVST may be calculated as the ratio of the S1 to S2 interval to the RR interval (e.g., S1-S2 / RR). Another example of a combined physiological metric is the ratio of EMAT / LVST, which is useful in determining the degree of left ventricular dysfunction. Another example of a combined physiological metric is the systolic dysfunction index (SDI). The SDI is a multiplicative combination of ECG and oscillation parameters. The SDI has been shown to predict left ventricular systolic dysfunction with high specificity. The multiplicative score, SDI, is derived from QRS duration, QR interval, %EMAT, and S3 oscillation magnitude. The SDI is reported as a value between 0 and 10. Other combined metrics may include %LVST, left ventricular end-diastolic pressure (LVEDP), and other related combined metrics derived from two or more physiological metrics described herein.

[0149] 11, in some examples, physiological metrics 1145 may be populated with patient demographic information and patient history information 1140. For example, patient demographic information may include age, gender, race, etc. For example, patient history information may include previous medical conditions, diagnoses, length of previous hospitalizations, etc. A sample patient history record may include the information shown below, with sample input into a patient monitoring system further shown below in Table 3. Table 3 [Table 3]

[0150] In some implementations, an additional set of physiological metrics may be derived based on patient activity test data 1142 from one or more physical activity tests performed by a patient wearing a wearable medical device. For example, as shown below, the device may monitor one or more activity tests performed by the patient. In one example, the physical activity may be a physician-ordered physical assessment, such as a WalkTest® activity assessment administered by a LifeVest® WCD. The purpose of the assessment is to monitor the patient's process during an assessment that monitors parameters such as the patient's number of steps taken while walking for approximately six minutes, the distance traveled during the assessment, and the speed of movement.

[0151] Before and after walking, the patient may be asked to complete a health survey answering one or more questions regarding the patient's overall health. For example, the patient may be asked to indicate their level of shortness of breath (e.g., an exemplary scale could be 0 = normal sensation, 0.5 = able to maintain a conversation while walking, 1 = need to stop immediately). Another query may concern the patient's fatigue level (e.g., how tired the patient feels). After the assessment, the patient may be prompted to answer the same set of questions again or additional questions. Responses to these questions may be normalized to a set of physiological parameters, e.g., stored as values ​​ranging from 0 to 1, 0 to 10, 0 to 100, or any other suitable range, and input into predictive and / or trend analyses. While numerical ranges are described herein, other scales, ranges, and / or methods of quantifying the patient's response to the health survey may be used. The patient's physician prescriber may have the patient repeat the assessment daily or weekly. The device may remind the patient via the user interface when the patient is next due for assessment. Although the health surveys described herein are in the context of conducting a physical assessment, in some implementations, a health survey may be administered to a patient wearing the device separately from a physical assessment. For example, the device may periodically (e.g., daily, weekly, or on some other schedule) prompt the patient to complete the health survey. In some examples, the patient's physician may configure the device to administer the health survey as needed. In such cases, the patient's physician may instruct a technician to send one or more operational commands to a remote server to cause the wearable medical device to administer the health survey.

[0152] The distance walked over the assessment duration (e.g., 6 minutes) may indicate the patient's current clinical status. For example, if a patient walks less than a preset duration (e.g., 350 meters), the patient may be scored as being at a relatively higher risk of dying compared to a patient who travels more than 350 meters. In some examples, the input to the predictive and / or trend analysis may be the raw distance traveled by the patient. In other examples, the distance may be normalized according to a predetermined scale. For example, a normalized scale may be generated that indicates the distance traveled by the patient during an activity. For example, a scale from 0.0 to 1.0 indicating the distance traveled is shown in Table 4 below. Table 4 [Table 4]

[0153] In some examples, variations in measured distance may occur when a clinician inputs stride length into the device via a user interface, and the clinician-provided stride length is utilized. In some implementations, normalized stride length (e.g., based on similar patient population data) may be implemented when calculating distance traveled.

[0154] Additionally or alternatively, the device may track the number of steps taken by the patient over an evaluation period (e.g., 6 minutes). Accelerometer data may be analyzed, for example, to count heel strikes (e.g., this may be detected when the sum of the accelerometer's three measurement axes exceeds a preset threshold). For example, a scale from 0.0 to 1.0 indicating the number of steps taken is shown in Table 5 below. Table 5 [Table 5]

[0155] In some examples, the input to the predictive and / or trend analysis may be the raw number of steps taken by the patient, while in other examples, the number of steps may be normalized according to a predetermined scale.

[0156] Another parameter that may be tracked includes pace during assessment, e.g., steps per minute and / or distance traveled per minute. As before, in some examples, the input to the predictive and / or trend analysis may be the patient's raw pace information. In other examples, pace may be normalized according to a predetermined scale. For example, during the normalization process, these metrics may be converted and stored as values ​​ranging from 0 to 1, 0 to 10, 0 to 100, or any other suitable range, and input to the predictive and / or trend analysis. Although numerical ranges are described herein, other scales, ranges, and / or methods of quantifying patient metrics may be used. Additional patient metrics include: Heart rate during evaluation (average, maximum, expected, mode) Respiratory rate during assessment Whether the patient completed the test (completed / not completed) or what percentage of the test they completed The patient performed a walking test, which could be an indicator that the patient has problematic / worsening symptoms may include:

[0157] As noted above, all of the physical activity assessment metrics may be normalized according to a predetermined scale and input into predictive and / or trend analysis.

[0158] In some implementations, additionally or alternatively, a patient's overall performance on a physical assessment, such as the WalkTest®, may be scored on a normalized scale. The overall performance score may be based, for example, on individual scores for each component of the physical assessment, as well as information regarding the health assessment. For example, the overall performance may be rated on a scale ranging from 0 to 1, as shown in the table below: Table 6 [Table 6]

[0159] 11 , the entire data set of physiological metrics 1145 is sent to a predictive or trending classifier 1150, which generates an output 1155. Referring now to FIG. 12 , as shown in system 1200, the output 1155 of the predictive or trending classifier 1150 may include a score indicative of the patient's current overall clinical condition 1205. If the patient's condition worsens, e.g., the clinical condition score has a worsening trend and / or exceeds a threshold, the system may prompt a specific action. In one example, the classifier may output a score from 0 to 1 indicative of the patient's overall condition, with a score at the low end of the scale (e.g., 0.0 to 0.25) indicating a stable patient that does not raise any concerns, and a score at the high end of the scale (e.g., 0.50 or above) indicating an unstable patient that requires closer monitoring (e.g., indicating the possibility of an adverse event). For example, such adverse events may include one or more of an arrhythmia event (VT / VF), a seizure event, a syncope event, and a hospitalization event. Based on the patient's current health status 1205, the system 1200 may generate recommended clinically actionable events, such as adjusting the patient's treatment plan 1210, and send the generated recommendations to a caregiver, such as a physician 1220.

[0160] In certain implementations, information regarding changes in the treatment plan may be fed back to a predictive or trend analysis classifier 1150, as shown in Figure 11. Such a feedback loop may provide improvements to the predictive or trend analysis classifier 1150 with information regarding changes in the treatment plan and the patient's response to those changes.

[0161] Additionally or alternatively, to monitor trends in the output score, the score may be subject to one or more threshold conditions. For example, the threshold conditions may trigger one or more clinically actionable events. Table 7 below provides an example set of patient health score categories, threshold ranges, and the patient's current health status information. Table 7 [Table 7]

[0162] In some implementations, the score categories may be part of a specific scoring scheme for implementation in a specific device according to embodiments described herein. In other specific implementations, the process may classify patients according to a grade of heart failure adopted by a physician or hospital association, society, or other regulatory authority. For example, such authorities may be the American College of Cardiology (ACC), the American Heart Association (AHA), and the New York Heart Association (NYHA). In this regard, the device may automatically classify patients into an appropriate grade and / or stage of heart failure based on the similar scoring schemes described above. In some implementations, a user may configure in advance via a user parameter set which grading scheme the device will implement. For example, during setup, the user may be prompted to indicate the grading scheme according to one or more of the following options: The user may select these options via a user interface (e.g., either locally on the device or via remote configuration parameters on a server that are then sent to the device): Unique scoring system (described above) ·ACC / AHA method ·NYHA method You may indicate.

[0163] The ACC / AHA method can be implemented as follows.

[0164] Stage A: Patients at risk for heart failure who have not yet developed structural cardiac changes (ie, diabetic patients without prior infarction, patients with coronary heart disease).

[0165] Stage B: Patients with structural heart disease (i.e., low ejection fraction, left ventricular hypertrophy, left ventricular enlargement) who have not yet developed symptoms of heart failure.

[0166] Stage C: Patients who developed clinical heart failure.

[0167] Stage D: Patients with refractory heart failure who require advanced interventional therapy (i.e., biventricular pacemaker, left ventricular assist device, transplant surgery).

[0168] The techniques described above can be used to map scores from the processes taught herein to the stages defined in the ACC / AHA system. For example, Table 8 below shows the correlation between the scoring techniques described above and an ACC / AHA system implementation. Table 8 [Table 8]

[0169] The NYHA system can be implemented as follows: NYHA classified each diagnosis into four grades. Grades I and II are considered mild. Grade III is considered moderate, and Grade IV is severe. In Grade I, there are no limitations in physical activity. Patients generally do not complain of fatigue or feel short of breath. Patients can still control the disease. Regular exercise, limiting alcohol consumption, and a healthy diet (moderate sodium intake) are all actions that can be taken early on. High blood pressure will need to be treated. Smoking cessation is essential.

[0170] In grade II heart failure, patients will experience some limitations in everyday physical activities such as bending forward or walking. Patients may become fatigued and short of breath. Non-invasive treatments such as angiotensin-converting enzyme inhibitors or beta-blockers (depending on the patient) may be considered.

[0171] Patients with grade III heart failure experience significant limitations during physical activity. Patients may remain pain-free at rest, but most physical activity will cause excessive fatigue. Under medical supervision, the patient's diet and exercise may be monitored. Diuretics may be prescribed to reduce fluid retention.

[0172] Patients with grade IV heart failure engage in virtually any physical activity with discomfort. Significant signs of cardiac damage may be present even while at rest. Surgical options may be considered with the same considerations as those given to the treatment of grades I through III.

[0173] The techniques described above can be used to map scores from the processes taught herein to the stages defined in the NYHA system. For example, Table 9 below shows the correlation between the scoring techniques described above and NYHA system implementations. Table 9 [Table 9]

[0174] At any point, one or more events may be triggered depending on the patient's grade and / or score. These events may be either automatic events triggered without user input. For example, the system may automatically instruct an external medical device (e.g., a wearable defibrillator) connected to the patient to increase the sensitivity of its detection process during periods when the patient may be at high risk for a sudden cardiac arrhythmia event. Alternatively, the system may output information regarding the patient's current symptoms and instruct caregivers and / or others to take action. Such actions may include recommendations to modify the patient's treatment plan, such as changing medication dosages, or to place the patient under advanced monitoring or admit the patient to a hospital.

[0175] For example, a clinically actionable event may relate to an external medical device (e.g., a wearable defibrillator), such as, without limitation, triggering a change in the controller to shorten the time from detection of sudden cardiac arrest to treatment. Additionally, detecting a high risk score category may prompt the system to notify the patient, physician, responsible third party, medical team, and / or technical support of the change in risk. In this manner, patients in a high risk score category may be alerted for more frequent future monitoring and / or changes in treatment plan.

[0176] In some implementations, if a patient's clinical condition worsens, a caregiver and / or technical support personnel may cause the patient's external medical device to initiate real-time or substantially real-time streaming of the patient's ECG data (and / or other physiological data) to a remote server for additional monitoring and analysis. For example, as shown in FIG. 12, if the patient's current health status 1205 indicates a worsening condition, the system may send an indication to a remote server and take various notification and / or analysis actions 1215.

[0177] In some instances, the reporting and / or metric measurement and analysis period for such patients may be increased (e.g., from once every few hours or more to once every hour). In some instances, additional monitoring and / or reporting of the patient's symptoms and various physiological parameters may begin during a period of concern about the patient's condition (e.g., until the patient's condition stabilizes or improves).

[0178] In some implementations, if the patient's condition worsens but has not yet reached a predetermined critical threshold, i.e., if immediate action cannot be deemed necessary (e.g., if the classifier scores the patient's current health status as 0.5 or above but 0.65 or below in the above example), the doctor or technical support personnel may begin reporting the patient's physiological condition more frequently (e.g., medically reporting the patient's symptoms) or begin streaming ECG data. Furthermore, if the patient falls into score category 3 (e.g., 0.5 or above), the doctor or technical support personnel may take additional action, including, for example, instructing the external medical device to be checked more frequently and / or ensuring that continuous monitoring downtime is kept to a minimum.

[0179] If a patient in score category 4 has high levels of asystole, actions that may be taken include, but are not limited to, notifying the patient and a responsible third party to seek medical advice and notifying the responsible medical team of the patient's critical condition.

[0180] The treatment plan adjustment 1210 may be based on modifying medications administered to the patient to control and / or address the patient's underlying heart failure disease or symptoms. As previously discussed, heart failure is generally a long-term, chronic disease that can worsen over time. Under the ACC / AHA system, there are four stages of heart failure (stages A, B, C, and D). As a patient's symptoms worsen, the patient's heart muscle tends to pump less blood to the organs, and the patient progresses to the next stage of heart failure. In standard treatment plans, the goal of HF treatment is to prevent or slow the patient's progression through the stages. Treatment at each stage of heart failure may involve changes in medications, lifestyle behaviors, and cardiac devices. In various implementations, the patient monitoring system can base its recommendations on ACC / AHA or NYHA treatment guidelines. A basic treatment plan that may be applied to a patient according to stage / grade is described below.

[0181] Depending on the implementation, the patient monitoring device and / or patient monitoring system may recommend one or more of the following actions, which may be recommended to the patient directly via the device's user interface: In some cases, proposed treatment plans and / or changes to the treatment plans may be recommended to the patient's primary caregiver and / or physician:

[0182] Typical ACC / AHA Stage A is considered early heart failure. At this stage, patients are monitored for signs of hypertension, diabetes, coronary artery disease, metabolic syndrome, a history of alcoholism, a history of rheumatic fever, a family history of cardiomyopathy, and a history of taking medications that can damage the heart muscle, such as some anti-cancer drugs. Exemplary Stage A treatment plans that the patient monitoring system may recommend include the following: Perform ECG monitoring only The monitoring device must not initiate bio-vibration and / or radio frequency-based TFC monitoring of the patient. Regular exercise, being active, walking every day ·no smoking Hypertension treatment (medication, low-salt diet, active lifestyle) High cholesterol treatment - Avoid alcohol or energy boosters Medication: Angiotensin-converting enzyme inhibitors (ACE-I) or antihypertensive drugs (ARBs) if the patient has coronary artery disease, diabetes, hypertension, or other vascular or cardiac disease Beta-blockers if the patient has hypertension

[0183] ACC / AHA Stage B is considered early heart failure. This means that the patient has been diagnosed with at least left ventricular systolic dysfunction but has never experienced symptoms of heart failure. Most people with Stage B heart failure may have an echocardiogram (echo) that shows an ejection fraction (EF) of less than 40%. This category may include people with heart failure of any cause and low EF (HF-rEF). Exemplary Stage B treatment plans that a patient monitoring system may recommend include the following: Continuing or modifying treatments listed in Phase A For example, the monitoring device may initiate bio-vibration and / or radio frequency based TFC monitoring of the patient. For example, you may want to increase the frequency of your TFC measurements. If you were measuring your TFC weekly, you may want to measure it every two or three days. For example, the monitoring device may initiate administering a physical assessment and / or physical test to the patient. Adding an angiotensin-converting enzyme inhibitor (ACE-I) or an antihypertensive drug (ARB) (if the patient is not already on a treatment plan as part of their Phase A regimen) Add a beta-blocker if the patient has had a heart attack and if the patient's EF is 40% or less (if the patient is not already on a treatment plan as part of Phase A). Adding an aldosterone antagonist (to reduce the risk of the patient's heart muscle becoming enlarged and pumping insufficient) if the patient has had a heart attack or if the patient has diabetes and an EF less than 35% Enabling wearable defibrillator therapy Consider possible surgical or interventional treatments for coronary artery blockages, heart attacks, valve disease (where the patient may need valve repair or replacement surgery), or congenital heart disease.

[0184] Patients in ACC / AHA Stage C are generally diagnosed with heart failure and have current or previous signs and symptoms of the disease. There are many possible symptoms of heart failure. The most common are: ·shortness of breath Feeling tired (fatigue) -Inability to exercise Weak legs Waking up due to the need to urinate Swelling of the feet, ankles, lower legs, and abdomen (edema) is.

[0185] An exemplary Phase C treatment plan that the patient monitoring system may recommend includes: ·Continuing or modifying treatments listed in Phase A and Phase B. For example, the monitoring device may initiate bio-vibration and / or radio frequency based TFC monitoring of the patient (if not already monitoring). For example, you may increase the frequency of your TFC measurements. If you previously measured your TFC on a 2-3 day basis, you may change to daily measurements. For example, the monitoring device may initiate frequent physical assessments and / or physical tests on the patient. Adding a beta-blocker (if the patient is not already taking that medication) Adding an aldosterone antagonist (if the patient is not already taking it) if vasodilators (e.g., ACE-I, ARB, or angiotensin receptor / neprilysin inhibitor combinations) and beta-blockers do not relieve the patient's symptoms Adding a hydralazine / nitrate combination if other treatments do not relieve the patient's symptoms For example, adding medications that can slow the patient's heart rate if the patient's heart rate is greater than 70 beats per minute and the patient is still symptomatic · If chest fluid content and / or S3 strength continue to worsen, add diuretics ("diuretics"). Recommendations for dietary sodium (salt) restriction Daily patient weight tracking recommendations Fluid restriction as much as possible if chest fluid content and / or S3 strength continues to worsen Cardiac resynchronization therapy (e.g., biventricular pacemaker) whenever possible Implantable cardioverter defibrillator (ICD) therapy whenever possible If treating improves or prevents the patient's symptoms, the patient may still need to continue treatment to delay progression to Stage D.

[0186] Stage D and low EF patients (HF-rEF patients) generally suffer from advanced symptoms that do not improve with treatment. This is generally considered the end stage of heart failure. Exemplary Stage D treatment plans that the patient monitoring system may recommend include: Continuation of treatment and / or modification of treatment listed in Phase A, Phase B, and Phase C Heart transplant Artificial heart assist device Heart surgery Continuous infusion of intravenous inotropes Palliative care or hospice care ·Treatment research Evaluation of more advanced treatment options, including

[0187] The one or more physiological measures extracted from the ECG signal may include heart rate, heart rate variability, PVC load or PVC count, activity, noise quantification, atrial fibrillation, transient Pose , heart rate irregularity, QRS height, QRS width, change in morphology size or shape, cosine RT, artificial pacing, corrected QT interval, QT variation, T wave width, T wave alternans, T wave variation, ST segment changes, early repolarization, late potentials, fractionated QRS / HF content, and fractionated T wave / HF content.

[0188] The control unit may detect fiducial points in the ECG signal, e.g., points corresponding to P, Q, R, S, and T waves, to extract individual measurements, e.g., QRS, PVC, etc., from the physiological parameter data. For example, the QT interval may provide a measurement of the patient's heart failure, and the distance between the Q point and the T point may be determined and extracted from the physiological parameter signal.

[0189] 13 shows two ECG lead signals, including a plot 1300 of a left-right (SS) lead and a plot 1305 of an anterior-posterior (FB) lead. In particular implementations, a processing device may be configured to identify various information, such as QRS measurements and PVC parameters, within each ECG signal for extraction. For example, the processing device may apply a QRS detector and a PVC detector to the ECG signals to extract QRS measurement information 1302 and PVC measurement information 1304 from the SS lead plot 1300. The processing device may further extract PVC measurement information 1306, for example, from the FB lead plot 1305.

[0190] In certain implementations, the processing device may use the extracted QRS and PVC measurements and / or fiducial points determined by the QRS and PVC detectors to identify and extract various other measurements or parameters within the ECG signal, such as heart rate variability (HRV), RR intervals, etc. For example, the output from the QRS and PVC detectors may be fed to an AFIB detector within the control unit, which extracts measurements or parameters from the ECG signal. Pose The QRS detector may be utilized by a pulse wave detector, a pace detector, a morphology detector, a T-wave detector, and / or any other detector. QRS detectors, PVC detectors, and other detectors are well known in the art, as described, for example, in "The Principles of Software QRS Detection" (Kohler, Hennig, et al.), IEEE ENGINEERING IN MEDICINE AND BIOLOGY (January / February 2002), the entire contents of which are incorporated by reference.

[0191] HRV measurements quantify the temporal variation of R-R intervals in a patient's electrocardiogram (ECG) signal. The R wave of a particular beat corresponds to a point in the cardiac cycle during early systole, providing a reliable reference time for measuring cardiac cycle intervals from signal processing points on the diagram. HRV is influenced by the autonomic nervous system, which consists of the sympathetic nervous system (SNS) and parasympathetic nervous system (PNS). The observed HRV is considered an indicator of the dynamic interaction and balance between the SNS and PNS, providing a measure of nervous system performance. HRV is influenced by the autonomic nervous system and serves as an indicator for diagnosing and assessing various conditions, ranging from congestive heart failure to sleep apnea. For example, decreased HRV has been found to be a precursor to increased mortality in elderly individuals with coronary heart disease. Decreased HRV is also observed after sudden cardiac arrest.

[0192] A variability measure for HRV is T-wave alternan, which is a measure of the variability during myocardial recovery during diastole (relaxation) and measures the variability of the amplitude of the T-wave on the ECG.

[0193] In one example, extracting HRV data may include filtering an ECG signal to remove noise and artifacts, locating QRS complexes in the filtered ECG signal, finding R-R intervals between successive R peaks, and processing the R-R intervals to obtain HRV. For example, a band-pass filter is used to filter the ECG signal and locate the QRS complexes. A band-pass filter with an operating frequency range wider than the frequency components of the QRS complexes should be used. The frequency components of the QRS complexes are between 10 Hz and 25 Hz. Therefore, in one embodiment of the present disclosure, the operating frequency range of the band-pass filter is approximately 5 Hz to approximately 28 Hz.

[0194] In one example, the R wave may be located as follows: The first occurring maximum peak data value in the filtered ECG signal is located; Upper and lower amplitude thresholds from the located maximum peak value are determined; The peak value and the minimum values ​​on both sides of the peak value are located; in this embodiment of the present disclosure, both sides refer to the left and right sides of the peak value; whether the condition that the peak value exceeds the upper amplitude threshold and the minimum value is below the lower amplitude threshold is verified; if the condition is met, the position of the peak value is displayed as the R position; the position of the minimum value occurring closest to the left of the R position is displayed as the Q position, and the position of the minimum value occurring closest to the right of the R position is indicated as the S position. With respect to the time scale displaying the filtered ECG signal, the Q position occurs where the first minimum value occurs before the R position, and the S position occurs where the first minimum value occurs after the R position. Thus, the position of the QRS peak in the filtered ECG signal is determined.

[0195] In one embodiment of the present disclosure, a 1D array of ECG sample points x(n) is provided and an upper and lower amplitude threshold (T upper and T lower ) is set after finding the maximum (ref peak) within the first few seconds of data. The threshold is T upper =ref_peak+0.4*ref_peak T lower =ref_peak-0.35*ref_peak Then, the following condition is defined: x(i) is T upper and T lower Between x(i+1)-x(i)<0, x(i)-x(i-1)>0, The R wave is said to occur at point i if it is satisfied that the R peak is the point of maximum value.

[0196] The locations of other R waves in the filtered ECG signal may be located by repeating the process of locating other peaks and other minima on either side of another peak. If another peak exceeds the upper amplitude threshold and both other minima are below the lower threshold, the location of the peak is labeled as the R location. The location of the minimum occurring closest to the left of the R location is designated as the Q location, and the location of the minimum occurring closest to the right of the R location is labeled as the S location. In this manner, the locations of other QRS peaks are determined.

[0197] In an example, the heart rate variability data may include time domain data, frequency domain data, and geometric domain data.

[0198] The time domain data may include information regarding any one or more of the following parameters: mean value of RR intervals (mean RR), standard deviation of RR intervals (STD), mean value of instantaneous heart rate (mean HR), standard deviation of instantaneous heart rate (STD_HR), root mean square of adjacent RR interval differences (RMSSD), number of consecutive RR intervals that differ by 50 ms or more (NN50), and percentage of consecutive RR intervals that differ by 50 ms or more (pNN50).

[0199] Frequency domain data were analyzed using the following parameters: power in the very low frequency range (<=0.04 Hz) (VLF), power in the low frequency range (0.04 Hz to 0.15 Hz) (LF), power in the high frequency range (0.15 Hz to 0.4 Hz) (HF), total power (TP) estimated from the variation of the NN interval within the segment and measured in ms², ratio of LF power to HF power (LF / HF), LF power in normalized units (LF / (TP-VLF) × 100(LF)). norm ) and HF power in normalized units (HF / (TP-VLF) × 100(HF norm ) may include information regarding any one or more of:

[0200] All of the above physiological parameters and metrics related to the patient's health status are collected as described above. Within the physiological parameter data, one data set may not contain the same number of parameters compared to another data set. Furthermore, the patient parameters may be stored as digital data in which each of the four parameters has been converted from the form in which it was originally obtained (such as an analog signal), i.e., the original form of the measurement obtained. Data on patient characteristics, such as demographics and medical history, may be transmitted to the system via a wireless network distributed throughout the hospital or via a physician's network. All physiological parameters may be used to train the predictive analysis process. Furthermore, patient demographics (e.g., age, gender) and medical history information may be used to train the process. At the end of the training phase, parameters found to be most important for achieving high accuracy will then be used as inputs to the real-time detection system.

[0201] The physiological parameters may be stored in an electronic database. In an example, this data includes multiple data sets, each set having at least one of a first parameter related to cardiac vibration data and a second parameter related to radiofrequency-based thoracic fluid content level. Each of the multiple data sets further has a third "output" parameter related to the patient's health status, e.g., a patient health status score.

[0202] The electronic database used to store patient physiological parameters may be a memory module such as a hard disk drive, optical disk, or solid-state device. During the training phase of the artificial neural network, the training patient physiological parameters may be obtained from hospital admission records or from conducting an on-site survey of a patient pool, where the pool includes at least one group of patients designated to serve as a control group. Thus, the patient physiological parameters may include data on patients suffering from various cardiac-related conditions, patients who may be considered healthy, such as patients who show no signs of cardiac disease, and additional patients of various demographics.

[0203] A third output parameter is also trained within the process. This parameter is sometimes referred to as an outcome, such as death, other adverse event, recovery from an adverse event, or the patient's current health status. The electronic device may include a processor or memory module that stores instructions that implement the training process, such that the device can analyze the physiological parameters of the patient being examined. The output of the electronic device can then be used to predict the patient's outcome and thereby assist the surgeon or medical professional in making appropriate clinical decisions about how to treat the patient.

[0204] For example, the process utilized may depend on the type of medical event to be predicted, and depending on the event, the patient may be classified or categorized into an appropriate heart failure score category (i.e., NYHA grade or ACC / AHA stage). For example, the processing device may be configured to calculate an event rating for a cardiac arrest risk score using a first process or processes and calculate an event rating for a ventricular fibrillation risk score using a second, different process or processes. Thus, the imminent and serious risk of a patient's condition degenerating into cardiac arrest or other severe cardiopulmonary disease may be calculated in various ways. Various methods and processes may be used to calculate the event rating for a risk score for various time periods. For example, the processing device may be configured to calculate an event rating for a cardiac arrest risk score for a first time period using a first process or processes and calculate an event rating for a cardiac arrest risk score for a second, different time period using a second, different process or processes.

[0205] In some implementations, machine learning classifiers, described in more detail below, can be trained on large populations, such as populations that may contain thousands to tens of thousands of patient records containing electrophysiological, demographic, and medical history information. Machine learning tools can include, but are not limited to, classification and regression tree decision models, such as random forests and gradient boosting (e.g., implemented using R or any other statistical / mathematical programming language). Any other class-based machine learning tool can be used, including artificial neural networks (described in more detail below) and support vector machines. Because machine learning tools can be computationally intensive, some or all of the processing for the machine learning tools can be performed on a server separate from the medical device.

[0206] An overview of how a random forest tool may be applied to a given dataset may illustrate how the classification tool may function in interpreting a given parameter or metric. A random forest is a collection of decision trees. A decision tree is a flowchart-like structure in which each node represents a test against a metric and each branch represents the result of the test. The tree ends with a classification label, such as the decision made at the end after calculating each of the metrics. Each tree in a random forest tool receives one "vote" in classifying a given set of metrics. There are two components of randomness involved in building a random forest. First, in the generation of each tree, a random subsample of the total dataset is selected to grow the tree. Second, at each node of the tree, a "splitter variable" is selected to split the original patients into two classes. For example, patients in one class (e.g., response or occurrence of sudden cardiac arrest) can be separated from patients in another class (e.g., non-response). The tree grows with additional splitter variables until all terminal nodes (leaves) of the tree are entirely of one class or the other. The tree is "tested" against previously set-aside patient records. Each patient test record traverses the tree and descends one branch or another depending on the metric contained in the record for each splitter variable. The patient test record is assigned a predicted outcome (a vote) based on where in the tree it lands. The entire process may be repeated with new random distributions of the original dataset to generate additional trees and ultimately a "forest." In each case, different subsets of patients may be used to build the tree and test its performance.

[0207] In developing the results described in the exemplary implementation below, a predetermined number of model variants are trained. For example, each model variant is ranked sequentially (e.g., from 1 to 100 for 100 runs). In each run of the model, the software randomly sampled a predetermined percentage of the population (e.g., 80%) as a training set and reserved the remaining portion (e.g., 20%) as a validation set.

[0208] As described above, machine learning tools can train classifiers on a first percentage of the original dataset and validate them on a second percentage of the dataset or on another separate dataset. When evaluating the performance of each classifier, the performance of basic decisions within the random forest decision tree can be evaluated based on specificity and sensitivity parameters. For example, the sensitivity parameter can be based on a measure of the classifier's ability to correctly predict whether a patient is at risk of needing treatment according to the appropriate score category. For example, the sensitivity parameter can be based on the proportion of appropriately treated patients and at-risk patients who are correctly predicted by the model. The specificity parameter can be based on the proportion of untreated patients and patients predicted by the associated classifier as not at-risk of needing treatment. It may be advantageous to optimally balance individual performance variables, such as high levels of sensitivity and specificity. For example, by setting the specificity at a relatively high value, e.g., 95%, the basic threshold within the classifier model can be adjusted to minimize false positives. After the specificity is defined, the sensitivity measurement can be treated as a type of performance measure, e.g., generally in the range of 15% to 35% for a given model. However, smaller or larger sensitivity values ​​are possible.

[0209] For example, as described below, a validation protocol may be utilized to validate the predictive performance of the training model. In implementations, the validation phase may be used to ascertain appropriate threshold scores for classifying future patients (whose outcomes are currently unknown and for which outcome prediction is desired) and to determine the predictive performance of each classifier model generated by the machine learning tool. A second group of individuals, such as a validation population (or cohort), may be used to validate various classifiers and associated threshold scores. For example, the validation population used may be a new validation population. The outcomes of patients in the validation cohort are eventually learned as these patients progress to end-of-life use of the device. In one embodiment, the patients in the validation population may differ from the training and test groups of patients described above for training the classifier. For example, the validation population of patients and associated metrics (validation metrics) may be independent of the training population of patients and associated metrics (training metrics). In some implementations, there may be overlap between the validation metrics and the training metrics.

[0210] In some implementations, the validation population may be updated by at least one of 1) adjusting one or more metrics in the validation metric and 2) expanding the validation metric based on adding one or more additional patients to the population of patients comprising the validation population. Thresholds for classifying future patients may be refined based on the updated validation metric. For example, metrics of patients currently being treated or monitored, or who have not otherwise progressed to device end-of-life, may be used to adjust one or more metrics in the validation metric, or the patient's metrics may be added to the validation population as metrics from a new patient. As new metrics for a patient are determined during patient monitoring or treatment, the validation metric may be adjusted. In some examples, as a monitored patient progresses to device end-of-life, the patient's metrics may be added to the validation population and / or used to adjust metrics in the validation metric after the patient progresses to device end-of-life.

[0211] In some implementations, the training population may be updated by at least one of 1) adjusting one or more metrics in the training metrics and 2) expanding the training metrics based on adding one or more additional patients to the first plurality of patients. The machine learning classifier model may be retrained based on the updated training metrics. For example, as additional patient metrics are determined from current patients and / or metrics from new patients are determined, the machine learning classifier may be retrained, for example, on an increased number of metrics or new, different metrics, to provide an updated classifier model. The training population may be updated when new metrics for current patients and / or metrics for new patients are determined, or after a patient has progressed to end-of-life use of the device.

[0212] FIG. 14 shows a sample flow 1400 for training and validating one or more classifier models of the machine learning process described above. A set or population of known patient records may be provided as the dataset to use for training and validating the classifier model. For example, the dataset of known patient records may include 1,000 patients suffering from a particular type of cardiac disease, such as VT, their treatment plans, and their associated outcomes. A percentage of the data in the known patient records may be used as the training dataset 1405. For example, 80%, or 800 patient records, may be used as the training dataset 1405. As described above, the training data 1405 may include various available information for multiple patients, including patient ECG metrics 1125 (e.g., see FIG. 11 ), bio-vibration metrics 1130, radio frequency metrics (1135), combined physiological metrics 1145, demographic and medical history information 1140, and patient activity test data 1142. Training data 1405 may include, for each training data set, historical information regarding how each training data set was classified or scored into the appropriate score category (ie, NYHA grade or ACC / AHA stage).

[0213] The training dataset 1405 may be provided to a training module 1410. The training module 1410 may include one or more non-training data structures, such as a series of data trees (e.g., organized using the random forest tool described above). Using known input variables and known outcomes from the training dataset 1405, the training module 1410 may train its data structure to more accurately generate likely (and known) outcomes by iteratively processing each data point in the training set.

[0214] Once the training module 1410 empties the training dataset 1405, the training module may output one or more trained classifier models 1415. The one or more trained classifier models 1415 may represent a model set that provides the most accurate rating and produces results for a known set of input variables that can be generated from the training data 1405. The validation module 1420 may further be configured to refine the trained classifier models 1415 with additional patient records. For example, a validation dataset 1425 may be input to the validation module 1420 to validate the one or more trained classifier models 1415. To continue with the above example, the validation dataset 1425 may include 200 patient records. Generally, there is no overlap between the training dataset and the validation dataset, as there is no advantage to running the same dataset twice.

[0215] 14, a validation module 1420 may process a validation dataset 1425 to generate one or more validation classifier models 1430. Depending on the intended purpose of the validation classifier model 1430, the model may have a particular specificity or sensitivity as described above.

[0216] The generated results can be used as a validation classifier model to be used to classify new patients (e.g., to generate new outputs for the patient metric sets described herein) to further validate this process using a closed-loop feedback system. For example, once a patient is classified and treated, the results of that treatment can be included in the patient record and verified, for example, by the patient's physician. The patient record, now updated to include known results, can then be provided to the validation module 1420 as feedback 1435. The validation module can process the feedback 1435 and compare the generated outputs with the patient's known results. Based on this comparison, the validation module 1420 can further refine the validation classifier model 1430, thereby providing a closed-loop system in which the model is periodically updated and upgraded.

[0217] Alternatively, the system processes may be implemented as a network of interconnected nodes to form an artificial neural network. For example, Figure 15 shows the topography of a sample artificial neural network 1500. The artificial neural network 1500 may include, for example, an input layer 1502, one or more nodes configured in a series of hidden layers 1504, and one or more nodes configured in an output layer 1506.

[0218] In the artificial neural network, the nodes include a plurality of artificial neurons, each having at least one input with an associated weight. The artificial neural network can be trained using a patient's physiological parameters such that the associated weight of the at least one input of each artificial neuron of the plurality of artificial neurons is adjusted in response to first, second, and third parameters, respectively, of different data sets from the physiological parameters. This results in the artificial neural network being trained to generate a health status score for the patient.

[0219] An artificial neural network may be a mathematical or computational model that simulates the structure and / or functional aspects of a biological neural network. In embodiments, a node of an artificial neural network includes at least one input, at least one artificial neuron, and at least one output. A neuron may reside in a single hidden layer of the artificial neural network and may have two or more inputs. In instances where the artificial neural network has multiple neurons, the multiple neurons may be distributed across one or more hidden layers. When multiple layers are present, each layer may be interconnected with previous and subsequent layers.

[0220] An artificial neural network may be an adaptive system that changes based on external or internal information flowing through the artificial neural network during a training or learning phase. Specifically, the weights (or strengths) of connections within the artificial neural network (such as between adjacent artificial neurons or between input neurons and artificial neurons) are adapted to change to match known outputs.

[0221] In an example, a first parameter (e.g., S3 oscillation intensity) and a second parameter (e.g., tissue fluid content), or a combination of the first and second parameters, may be classified as a feature vector of the patient's physiological parameters. An artificial neural network may be trained with the feature vector. The artificial neural network may be implemented as instructions stored in a memory that, when executed by a processor, causes the processor to perform the functions of the artificial neural network. In an embodiment of the present disclosure, the artificial neural network may be based on a support vector machine architecture, and an associated weight of at least one input of each artificial neuron of the plurality of artificial neurons is initialized from a library used by the support vector machine.

[0222] The support vector machine may have an ensemble output with a decision function, which is given by:

number

[0223] For example, the artificial neural network may be based on an extreme learning machine architecture, where the associated weights of at least one input of each artificial neuron of the plurality of artificial neurons are initialized through random selection by an extreme learning machine. The artificial neural network may be implemented as a single-layer feedforward network, whereby the prediction of patient survivability is derived from the function:

number

[0224] The training of the artificial neural network may be based on backpropagation learning. For example, the backpropagation learning may use a Levenberg-Marquardt process. Each of the neurons of the artificial neural network may have an activation function selected from the group of functions comprising hard limit, sigmoid, sine, radial basis, and linear.

[0225] In an example, the outcome of the artificial neural network may be only two possible values: either the patient's health is stable and good (no adverse events expected), or the patient is hospitalized (adverse events expected).

[0226] In another example, the physiological parameter data may be divided into multiple portions. A first parameter set may relate to the patient's ECG metrics, a second parameter set may relate to the patient's cardiac vibration metrics, and a third parameter set may relate to the patient's demographics and medical history. The first, second, and third parameter sets are then provided as sets of normalized data values ​​that are input into a scoring scheme. For example, the scoring scheme may assign a category to each parameter in the first, second, and third parameter sets. Each category may have multiple predetermined value ranges, each of the multiple value ranges having a predetermined score. The score for each parameter in the first, second, and third parameter sets is determined by assigning a normalized data set to each predetermined value range and including a set of normalized data values ​​within the multiple value ranges of the category associated with each parameter in the first, second, and third parameter sets. A total score may be obtained based on the sum of the scores of each parameter in the first parameter set, the second parameter set, and the third parameter set as an indication of the patient's condition. The scoring scheme may be any suitable process or processes and may be implemented in an electronic database that can assign a score to each value range within each category associated with each parameter in the first parameter set, the second parameter set, and the third parameter set. For example, the scoring scheme may be based on a mathematical model using logistic regression, such as univariate analysis.

[0227] In an example, the output of the patient health score may be a numerical value that may be determined according to statistical information or standard of care information.

[0228] In addition to using a learning process, such as an artificial neural network or machine learning process, as described above, trend analysis may also be utilized to monitor a patient for changing symptoms and to provide an output, such as modifying the patient's treatment plan. In certain implementations, trend analysis may include determining the existence of a substantial relationship between changes in two or more physiological parameters. For example, as described above in connection with FIG. 7, detecting a change in one particular metric may trigger monitoring of another metric. This additional monitoring may be performed to confirm the trend of the monitored metric as well as to provide additional information regarding the patient's overall condition.

[0229] An example of trend analysis is correlation analysis. Correlation analysis quantifies the linear relationship between two variables, and correlation analysis can be expressed as a value between negative 1.0 and 1.0, collectively known as the Pearson product-moment correlation coefficient, denoted as r. When the correlation coefficient approaches 1.0, the correlation coefficient indicates a strong positive correlation between the two variables. When the correlation coefficient approaches negative 1.0, the correlation coefficient indicates a weak negative correlation between the two variables. A correlation value close to zero does not indicate a linear relationship between the two variables.

[0230] In a particular example, there may be two sets of data samples, X and Y. For example, the X data set may be data sampled from curve X, such as an S3 intensity curve of cardiac oscillations, and the Y data set may be data samples from curve Y, such as a curve of thoracic fluid content measurement. In a particular implementation, the correlation analysis may be performed using Pearson's product-moment correlation coefficient, Fisher transformation, or other similar correlation functions. For example, the Pearson's product-moment correlation coefficient may be used to calculate the correlation between n data pairs (X, Y), where X and Y are samples over a period of time. To continue the above example, X may be measured from the S3 intensity curve and Y may be measured from the TFC measurement curve. The above Pearson's product-moment correlation coefficient r may be obtained by exchanging the variance-covariance based samples of the X and Y data sets. The resulting value of r is between -1.0 and 1.0. In this example, a value of 1.0 would indicate a perfect correlation between X and Y, i.e., as S3 oscillation intensity increases, the TFC measurement increases accordingly. A value of -1.0 would indicate a negative perfect correlation. For example, as the S3 oscillation strength increases, the TFC measurement decreases. Conversely, as the S3 oscillation strength decreases, the TFC measurement increases. A value of 0.0 indicates that there is no linear relationship between the variables.

[0231] As described herein, multiple metrics can be correlated to determine a patient's symptoms. For example, if a patient is likely to experience an adverse cardiac event, such as arrhythmia, there may be a high correlation between thoracic fluid content (TFC) and S3 oscillation intensity. Using the process described in FIG. 10 , TFC is considered metric 1 and may be displayed on a normalized scale from 0 to 1000. Other normalized linear or nonlinear scales may be used. For example, the scale may be from 0 to 1, or from 0 to 10, or from 0 to 100. For example, the nonlinear scale may be a logarithmic scale. The processing device may monitor the TFC to detect any changes or trends in the patient's TFC value.

[0232] For example, as shown in FIG. 16A , a processing device may record a patient's TFC measurements over the course of several weeks. As shown in FIG. 16A , the TFC measurements are indicated by line 1600. For example, an initial measurement of a patient's TFC may include a TFC measurement of 120 TFC units. In some examples, the TFC units may be measured on a normalized scale of 0 to 200 TFC units. However, as shown in FIG. 16A , the plot is adjusted to focus on a range of 100 to 160 TFC units. At week 1, the TFC measurement may rise to 130 TFC units. However, at week 2, the TFC measurement may fall to 125 TFC units. At week 3, the TFC measurement may rise to 150 TFC units, and at week 4, the TFC measurement may rise to 155 TFC units. Depending on the programming of the processing device, the processing device may trigger a second metric measurement based on the change in the TFC measurement. For example, the TFC measurement may exceed a predetermined threshold (e.g., 150 TFC units) or threshold for a predetermined period of time (e.g., the second consecutive week of elevated TFC units, or the third to fourth week of elevated TFC units). These changes may trigger the processing device to monitor at least one additional metric. For example, as shown in FIG. 16A, in week 4, the processing device may also begin measuring S3 oscillation strength, as indicated by line 1602.

[0233] As described above, S3 vibration intensity may be measured on a normalized scale of 0 to 10 S3 intensity units, with 5.0 S3 intensity units indicating a threshold at which the patient may experience heart disease and / or require additional treatment. Other normalized linear or nonlinear scales may be used. For example, the scale may be 0 to 1 S3 intensity units, or 0 to 10 S3 intensity units, or 0 to 100 S3 intensity units. For example, the nonlinear scale may be a logarithmic scale. As shown in FIG. 16A , the initial readings shown may be adjusted to focus on a range of 3 to 9 S3 intensity units. At week 4, the patient's S3 vibration intensity is measured as 4.5 S3 intensity units. However, at week 5, the patient's S3 vibration intensity increases to 5.0 S3 intensity units, and by week 6, the patient's S3 vibration intensity increases to 6.0 S3 intensity units. During this same period, the patient's TFC measurements continued to rise as well, reaching 155 TFC units at week 5 and 165 TFC units at week 6.

[0234] In certain implementations, the processing device may perform a trend analysis to determine a correlation coefficient between both the TFC measurements and the S3 oscillation intensity. As shown in FIG. 16A, an approximate linear relationship may exist between the two metrics, potentially resulting in a correlation coefficient indicating a strong positive correlation between the two variables. Such a correlation may be interpreted by the processing device as indicating a particular adverse cardiac event, characterized by a high correlation coefficient between the TFC measurements and the S3 oscillation intensity. In certain implementations, a high correlation coefficient may further indicate a change in the patient's condition over a period of time, e.g., six weeks as shown in FIG. 16A.

[0235] In certain implementations, trend analysis can be used to determine how well a treatment plan is working for a patient. For example, as shown in FIG. 16B, a patient may be monitored for a six-week treatment plan. First, the patient's TFC measurements, shown in FIG. 16B as line 1610, can be monitored. As shown in FIG. 16B, during the first two weeks of monitoring the TFC measurements, the measurements decline. Such a downward trend can trigger additional monitoring. For example, as shown in FIG. 16B, S3 oscillation magnitude and %EMAT, measured in %EMAT units on a normalized scale of, for example, 0 to 100 EMAT units (adjusted to focus on a range of 10 to 70 %EMAT units, as shown in FIG. 16B), can be triggered by a downward trend in the TFC measurements. As shown in FIG. 16B, S3 oscillation magnitude is shown by line 1612, and %EMAT is shown by line 1614. As further shown in FIG. 16B, over the next four weeks, the TFC measurements, S3 oscillation intensity, and %EMAT each remained relatively stable or gradually decreased. Such results may indicate that the treatment plan is progressing well. In implementation, other metrics may be further verified. For example, if LVST also decreased during the same period, the device may indicate that the treatment plan is progressing well.

[0236] FIG. 16C shows a six-week trend analysis including the patient's overall grade score. As shown in FIG. 16C, the patient's TFC measurements over the first two weeks, represented by line 1620, may trend upward, producing an overall increase in the patient's overall grade score, represented by line 1626. As noted above, after a consistent trend in week two, additional monitoring may be triggered. In this example, monitoring of both S3 oscillation intensity and %EMAT may begin in week two. As shown in FIG. 16C, S3 oscillation intensity is represented by line 1622, and %EMAT is represented by line 1624. As further shown in FIG. 16C, the TFC measurements, S3 oscillation intensity, and %EMAT each continued to rise between weeks two and six.

[0237] As a result of these monitored metrics increasing, the patient's overall grade score continued to increase. As described above, the patient's overall grade score may be measured on a scale of 0.0 to 1.0 grade units, with 0.0 being the best condition and 1.0 being the worst condition. As shown in FIG. 16C , the displayed overall grade score reading may be adjusted to focus on a range of 0.1 to 0.7 grade units, for example. As the patient's overall score continues to increase, the monitoring device may determine whether the score exceeds a certain threshold (e.g., 0.50 grade units, or from one grade unit to another) and provide feedback accordingly, for example, to the patient's physician. In implementations, other metrics may also be verified. For example, if the LVST increases further during the same period, the device may identify that interventional therapy is needed to modify or manage the patient's treatment plan.

[0238] The metrics shown in FIGS. 16A through 16C are provided as examples. Additional metrics, such as LVST, may be similarly monitored and graphed over time. For example, as shown in FIG. 16D, a processing device may record a patient's TFC measurements over the course of several weeks, as indicated by line 1630. For example, an initial reading may include a TFC measurement of 120 TFC units. In week 1, the TFC measurement may rise to 130 TFC units. However, in week 2, the TFC measurement may fall to 125 TFC units. In week 3, the TFC measurement may rise to 150 TFC units, and in week 4, the TFC measurement may rise to 155 TFC units. Depending on the programming of the processing device, the processing device may trigger a second metric measurement based on a change in the TFC measurement. For example, the TFC measurement may exceed a predetermined threshold (e.g., 150 TFC units), or the increase may exceed a threshold for a predetermined period of time (e.g., the second week of consecutively elevated TFC units, or the third to fourth week of elevated TFC units). These changes may trigger the processing device to monitor at least one additional metric. For example, as shown in FIG. 16D, at week 4, the processing device also begins measuring LVST, as shown by line 1632.

[0239] For example, at week 4, the patient's LVST may be measured at approximately 250 ms. However, at week 5, the patient's LVST increased to 290 ms, and by week 6, the patient's LVST increased to 325 ms. During this same time period, the patient's TFC measurements also continued to increase, e.g., to 160 TFC units at week 5 and 165 TFC units at week 6. This increase in LVST may be used to confirm the interpretation that the patient's pleural fluid accumulation is worsening, and in such cases, interventional therapy may be recommended. Conversely, if the TFC measurements decrease substantially from week 1 to the next, e.g., a downward trend from high TFC units to low TFC units, the patient's LVST would be expected to decrease over the same period. The decrease in LVST may be used to confirm the interpretation that one or more interventional therapies to manage the pleural fluid appear to improve the patient's condition.

[0240] In certain implementations, the processing device may perform a trend analysis to determine a correlation coefficient for both the TFC measurement and the LVST. As shown in FIG. 16D, an approximate linear relationship may exist between the two metrics, potentially resulting in a correlation coefficient that indicates a positive correlation between the two variables. For example, a positive correlation coefficient may provide an indication that as the TFC measurement increases, the LVST increases. Such a correlation showing an upward trend may be interpreted by the processing device as indicating a particular adverse cardiac event characterized by a positive correlation coefficient between the TFC measurement and the LVST. In certain implementations, a positive correlation coefficient may also indicate a change in a patient's condition over a period of time, such as six weeks as shown in FIG. 16D.

[0241] 17 shows a sample process flow for a patient monitoring system described herein. For example, a series of sensors may be configured to detect 1705 raw signals from a patient. In a particular implementation, an ECG sensor may be configured to detect one or more ECG signals from the patient, a vibration sensor may be configured to detect one or more bio-vibration signals from the patient, and an ultra-wideband radio frequency transceiver may be configured to direct radio frequency electromagnetic energy toward the patient and receive, via the one or more antennas, radio frequency information corresponding to reflected radio frequency electromagnetic energy reflected from a portion of the patient, such as the aorta or thoracic cavity.

[0242] In particular implementations, the one or more processors may be configured to process the raw signals to determine 1710 a set of physiological parameters for the patient. For example, as described herein, the physiological parameters may include ECG metrics, bio-vibration metrics, radio frequency metrics, and / or combined physiological metrics. In some examples, the physiological parameters may be collected over a particular prior period of time. For example, the period may include at least 24 hours, at least 48 hours, at least one week, at least one month, at least six weeks, at least two months, at least four months, at least six months, at least one year, and at least two years. In some examples, the prior period corresponds to a period for which the patient's medical history is available to the patient.

[0243] The one or more processors may be further configured to use the physiological parameters to perform at least one of predictive analysis 1715 or trend analysis 1720. In particular implementations, the one or more processors may be configured to perform both predictive analysis 1715 and trend analysis 1720.

[0244] As described above, predictive analysis 1715 may include inputting the physiological parameters into a trained process. For example, the trained process may include an artificial neural network or machine learning process as described above. However, in certain implementations, the trained process may further include a deep learning process configured to analyze the physiological parameters and output current health status information for the patient.

[0245] Based on the output of the predictive analysis and / or trend analysis, the one or more processors may determine 1725 a current health status of the patient. Based on the patient's current health status, the one or more processors may determine 1730 one or more clinically actionable events for the patient. For example, clinically actionable events may include readmission, prescription of a wearable medical device such as a wearable defibrillator, an order to hospitalize the patient, a change in treatment plan, and other similar events. In particular implementations, clinically actionable events may include automatic events triggered without user input, such as increasing monitoring and / or analysis of one or more physiological signals. Similarly, clinically actionable events may include manual events triggered based on a user response to the generated output and may include one or more instructions to perform one or more actions, such as an order to hospitalize the patient.

[0246] The one or more processors may generate 1735 an output based on the clinically actionable event. In particular implementations, the clinically actionable event may include an automatic event triggered without user input, such as increasing the monitoring and / or analysis of one or more physiological signals. Similarly, the clinically actionable event may include a manual event triggered based on a user response to the generated output, and may include one or more instructions to take one or more actions, such as an instruction to hospitalize a patient.

[0247] Various validation studies were conducted to analyze and validate the concepts and techniques described herein. For example, in a heart failure analysis, validation studies were conducted to evaluate the value of the cardiology-based bio-vibration information described herein and the combination of the cardiology-based bio-vibration information with one or more ECG metrics. In the validation studies, a patient population including patients with a cardiac ejection fraction of <35% who were hospitalized with decompensated heart failure was enrolled in a prospective observational protocol. All patients were fitted with a WCD equipped with an accelerometer. The WCD was capable of recording bio-vibration and measuring ECG signals. Each patient was to wear the WCD in an outpatient setting. The patient population was monitored for heart failure events and emergency department visits for heart failure symptoms, and such information was collected for the patient population. Univariate and multivariate analyses were performed on the collected information to identify a model that best predicted the occurrence of early-stage heart failure rehospitalization or emergency department visits.

[0248] During validation, the patient population was divided into two groups based on the occurrence of heart failure events, resulting in a heart failure group and a non-heart failure group (based on the patient population's heart failure hospitalizations and emergency department visits). The heart failure group had a lower BMI (26±5 vs. 29±6) compared with the non-heart failure group, but all other demographic information was similar between the two groups. A multiparameter model utilizing heart rate, EMAT (QRS onset to S1), and S3 intensity (e.g., based on S3 timing, intensity, duration, and frequency) statistically produced the best predictive model for decompensated heart failure events. All-cause readmissions before the start of WCD use were similar between the two groups (heart failure 62% vs. non-heart failure 63%), as were NYHA class I and II grades (heart failure 68% vs. non-heart failure 69%). However, after WCD use and model validation, the ability of the NYHA grade to predict heart failure events was lower compared to using biofrequency information. For example, the biofrequency information predictor had a sensitivity of 68% and a positive predictive value of 28%, compared to the NYHA grade system's sensitivity of 30% and positive predictive value of 18%. Therefore, based on this validation, we can conclude that the use of biofrequency information alone, or in combination with ECG information and radiofrequency-determined lung fluid volume (discussed below), may be more useful than the traditional NYHA grade system in the early detection of patients at risk for decompensated heart failure. Such heart failure may result in patient readmission after an acute episode of decompensated heart failure.

[0249] A second validation study was conducted to provide insight into the use of radiofrequency-determined lung fluid volume to track disease status in patients with heart failure. The patient population included two groups. The first group included patients with acute heart failure with the following demographics: age 78 years + / - 8 years, 37% female, and BMI = 32 + / - 7. The second group included patients without acute heart failure (approximately 60% healthy and 40% stable heart failure patients), with the following demographics: age 56 years + / - 16 years, 27% female, and BMI = 27 + / - 5.

[0250] Each patient in the population underwent a supine chest CT scan to determine initial fluid volume. RF readings were then taken from a wearable patch device placed in the left mid-axillary line while the patient was lying supine. Using the reflected RF signal measured by the patch device and anthropometric data, an RF-based lung fluid model was constructed for each patient. Lung fluid was reported as a percentage of lung volume. A grading analysis technique was then utilized to compare the results and performance of the RF patch device with those of the chest CT scan.

[0251] The results of the above experiment showed that all patients with acute heart failure were measured as having higher lung fluid volume than patients with non-acute heart failure. Based on the statistics, RF patch designs and chest CT scans were performed to identify patients with acute heart failure from patients with non-acute heart failure. Sample data showing various statistical results determined from this experiment can be seen in Table 10 below. Table 10 [Table 10]

[0252] The data obtained from the two validations described above indicate that radio frequency data can be used to evaluate whether a particular patient should be considered a heart failure patient or an acute heart failure patient. Additionally, using the techniques described above, a patient's biovibration data can be incorporated with the patient's radio frequency data to indicate trends or changes in the patient's condition. For example, using the combined physiological parameter analysis described herein, patient information such as biovibration information and radio frequency data can be combined with measured ECG information into one or more combined metrics. Using the correlation analysis described herein (or, in other examples, another similar machine learning technique described above), the patient information can be transformed or otherwise processed to quantify one or more trends or changes in the patient's condition. The trend can be further analyzed to determine what additional information can be determined from the trend. For example, the trend can be further analyzed to determine information such as whether the patient's overall health is improving / worsening, whether the patient's cardiac health is improving / worsening, and / or whether the patient is likely to suffer another heart failure event in the near future (e.g., in the next 30 days) that may require an emergency department visit or readmission. For example, the combined physiological parameters based on the patient's radio frequency data and bio-vibration data may be further used to determine whether a patient being discharged from the hospital is likely to experience a heart failure event in the near future and / or whether the patient should be prescribed a portable medical device such as a WCD as described in the example below.

[0253] In an example, a patient may complain of chest pain when performing certain activities, such as sitting in a reclined position or lying on their back. The patient's physician may prescribe the patient a wearable radio frequency sensing device configured to measure the patient's TFC. The patient may wear the device over the course of several weeks. Over that period, the device may measure a steady increase in the patient's intrathoracic fluid content. Upon return to the physician, the physician may download the data recorded by the radio frequency sensing device into a trend analysis process. The process may detect trends indicating a steady increase in the patient's TFC. The physician may wish to obtain additional physiological information for the patient and may further prescribe a mobile cardiac monitoring device that includes a vibration sensor configured to measure bio-vibrations. The patient wears both the radio frequency sensing device and the mobile cardiac monitoring device for a period of time. Upon return to the physician, the recorded information from both monitoring devices may be input into a trend analysis process. The process may identify trends indicating an increase in the patient's TFC by analyzing information regarding the patient's S3 vibration intensity. The process may determine a correlation between increased TFC and the patient's S3 oscillation strength and provide an indication that the patient may be suffering from a cardiac condition such as congestive heart failure. The process may provide such an indication to a physician along with recommended actions, such as changes to the patient's prescribed medication.

[0254] In another example, a hospitalized patient may be fitted with at least one ECG sensor, at least one bio-vibration sensor, and at least one radio frequency sensor. Prior to discharge, data recorded by these sensors may be combined into one or more combined physiological parameters and input into a trained process, such as an artificial neural network (or, in other examples, another similar machine learning technique described above). The artificial neural network may convert the input information to provide clinical symptom information for the patient. This clinical symptom information may be further analyzed to determine the patient's current health status, whether the patient should be approved for discharge, and what the patient's risk of being readmitted to the hospital within a certain period of time, e.g., within 30 days. In some examples, even if a patient meets all the conditions necessary for discharge, the process may indicate that the patient has a high risk of being readmitted to the hospital in the near future. In such examples, a physician or other similar caregiver may decide to keep the patient hospitalized for additional monitoring, modify the patient's treatment plan, or otherwise alter the treatment being provided to the patient.

[0255] Although the subject matter contained herein has been described in detail for purposes of illustration, it should be understood that such detail is for that purpose only, and that the disclosure is not limited to the disclosed embodiments, but is instead intended to cover modifications and equivalent arrangements within the spirit and scope of the appended claims. For example, it should be understood that the disclosure contemplates that, to the extent possible, one or more features of any embodiment may be combined with one or more features of any other embodiment.

[0256] Other examples are within the scope and spirit of the claims. Furthermore, the particular functionality described above may be implemented using software, hardware, firmware, hardwiring, or any combination thereof. Features that implement the functionality may also be physically located in different locations, including distributing portions of the functionality to be implemented at different physical locations.

Claims

1. 1. A patient monitoring system comprising: an ECG sensor coupled to a patient and configured to detect one or more ECG signals of the patient; a vibration sensor coupled to the patient and configured to detect one or more cardiac vibration signals of the patient; an ultra-wideband radio frequency transceiver circuit coupled to the patient, the ultra-wideband radio frequency transceiver circuit having one or more radio frequency antennas; directing the one or more radio frequency antennas to direct radio frequency electromagnetic energy toward the patient's thoracic cavity; generating radio frequency information corresponding to reflected radio frequency electromagnetic energy received via the one or more radio frequency antennas and reflected from within the thoracic cavity; an ultra-wideband radio frequency transceiver circuit configured to one or more processors, a) processing the one or more ECG signals and b) the one or more cardiac vibration signals to generate a plurality of physiological parameters for the patient, the combined physiological parameter being based on one or more fiducial points in the one or more ECG signals and one or more occurrences of the one or more cardiac vibration signals; processing the radio frequency information to generate one or more radio frequency-based thoracic fluid parameters; performing a trend analysis of the plurality of physiological parameters comprising the combined physiological parameter to determine a trend of a current clinical condition of the patient, the trend analysis comprising: correlating the combined physiological parameter to at least one of the one or more radio frequency-based thoracic fluid parameters to generate a correlation coefficient when the combined physiological parameter exceeds a threshold; and determining the trend of the current clinical condition of the patient based on the correlation coefficient; determining one or more clinically actionable events based on the trend of the current clinical condition of the patient; and providing an output regarding said one or more clinically actionable events to an output device. one or more processors configured to: A patient monitoring system including:

2. The patient monitoring system of claim 1 , wherein the combined physiological parameter is electromechanical activation time (EMAT).

3. 10. The patient monitoring system of claim 1, wherein the one or more radio frequency based thoracic fluid parameters are thoracic fluid content (TFC).

4. 10. The patient monitoring system of claim 1, wherein the trend analysis is performed on physiological parameters collected over a previous period having one or more of at least 24 hours, at least 48 hours, at least one week, at least two weeks, at least one month, at least six weeks, at least two months, at least four months, at least six months, at least one year, and at least two years.

5. 10. The patient monitoring system of claim 1, wherein said trend analysis is performed on physiological parameters collected over a previous period corresponding to at least available medical history of said patient.

6. the one or more clinically actionable events: Automatic events that are triggered without user input; a manual event triggered based on a user response to the output and having one or more instructions to perform one or more actions; 10. The patient monitoring system of claim 1, comprising at least one of:

7. 7. A patient monitoring system according to any one of the preceding claims, wherein the correlation coefficient indicates a change in the patient's condition.

8. 10. The patient monitoring system of claim 1, wherein the vibration sensor is further configured to sense one or more pulmonary vibrations of the patient, the one or more pulmonary vibrations comprising at least one of bronchial vibrations, wheezing, crackles, wheezing, bullae, pleural rubs, squawks, glottal sounds, or pharyngeal sounds.

9. 2. The patient monitoring system of claim 1, wherein the one or more ECG signals include at least one of heart rate, heart rate variability, PVC load or PVC count, atrial fibrillation load, pause, heart rate turbulence, QRS height, QRS width, changes in size or shape of the morphology of the one or more ECG signals, cosine R-T, artificial pacing, QT interval, QT variation, T wave width, T wave alternans, T wave variation, and ST segment changes.

10. The patient monitoring system of claim 1 , wherein the one or more cardiac vibration signals include at least one of an S1 vibration, an S2 vibration, an S3 vibration, an S4 vibration, and a heart murmur vibration.

11. The patient monitoring system of claim 1 , wherein the radio frequency information includes a measurement of fluid content within the thoracic cavity of the patient.

12. The patient monitoring system of claim 1 , wherein the one or more processors are incorporated within a cardiac monitoring device adapted to be worn by the patient.

13. The patient monitoring system of claim 1 , wherein the one or more processors are incorporated in a remote processing device.

14. 10. The patient monitoring system of claim 1, wherein the one or more processors are incorporated into a wearable defibrillation device adapted to be worn by the patient.

15. 1. A patient monitoring system comprising: an ECG sensor coupled to a patient and configured to detect one or more ECG signals of the patient; a vibration sensor coupled to the patient and configured to detect one or more cardiac vibration signals of the patient; an ultra-wideband radio frequency transceiver circuit coupled to the patient, the ultra-wideband radio frequency transceiver circuit including one or more radio frequency antennas; directing the one or more radio frequency antennas to direct radio frequency electromagnetic energy toward the patient's thoracic cavity; generating radio frequency information corresponding to reflected radio frequency electromagnetic energy received via the one or more radio frequency antennas and reflected from within the patient's thoracic cavity; an ultra-wideband radio frequency transceiver circuit configured to one or more processors, processing the one or more ECG signals and the one or more cardiac vibration signals to generate a plurality of physiological parameters for the patient, including a combined physiological parameter based on one or more fiducial points in the one or more ECG signals and one or more occurrences of the one or more cardiac vibration signals; processing the radio frequency information to generate one or more radio frequency-based thoracic fluid parameters; performing a trend analysis of the plurality of physiological parameters, including the combined physiological parameter, to generate a clinical trend, the trend analysis including: correlating the combined physiological parameter to at least one of the one or more radiofrequency-based thoracic fluid parameters to generate a correlation score when the combined physiological parameter exceeds a threshold; and determining a current clinical trend of the patient based on the correlation score; updating a monitoring schedule for the patient based on the trend in the clinical condition; determining one or more clinically actionable events based on the trend of the current clinical condition of the patient; causing an output device to provide an output regarding said one or more clinically actionable events. one or more processors configured to: A patient monitoring system including:

16. 16. The patient monitoring system of claim 15, wherein the combined physiological parameter is electromechanical activation time (EMAT).

17. 16. The patient monitoring system of claim 15, wherein the one or more radio frequency based thoracic fluid parameters are thoracic fluid content (TFC).

18. the one or more processors further process the one or more ECG signals, the one or more cardiac vibration signals, and the radio frequency information in accordance with the updated monitoring schedule to generate updated physiological parameters of the patient; 16. The patient monitoring system of claim 15, configured to perform a trend analysis of the updated plurality of physiological parameters including the one or more combined physiological parameters to generate an updated clinical symptom trend.

19. the one or more clinically actionable events: Automatic events that are triggered without user input; a manual event triggered based on a user response to the output, the manual event comprising one or more instructions to perform one or more actions; 20. A patient monitoring system according to any preceding claim 15 or 18, comprising at least one of:

20. 16. The patient monitoring system of claim 15, wherein the one or more ECG signals include at least one of heart rate, heart rate variability, PVC burden or PVC count, atrial fibrillation burden, pause, heart rate turbulence, QRS height, QRS width, changes in size or shape of the morphology of the one or more ECG signals, cosine R-T, artificial pacing, QT interval, QT variation, T wave width, T wave alternans, T wave variation, and ST segment changes.

21. 16. The patient monitoring system of claim 15, wherein the one or more cardiac vibration signals include at least one of an S1 vibration, an S2 vibration, an S3 vibration, an S4 vibration, and a heart murmur vibration.

22. 16. The patient monitoring system of claim 15, wherein the radio frequency information includes a measurement of fluid content within the thoracic cavity of the patient.

Images