Quantifying hemodynamics from non-invasive waveform analysis phase information

Abstract

A method includes generating, via a sensor, a signal representing vibrations originating from a blood vessel of a patient and processing the signal to yield amplitudes and phases of the vibrations that correspond to oscillation frequencies of the vibrations. The method also includes identifying a first subset of the amplitudes and a second subset of the phases that correspond to the oscillation frequencies that are integer multiples of a heart rate of the patient. The method also includes, in response to providing the first subset of the amplitudes and the second subset of the phases to a computational model, determining a health metric for the patient and generating, via a user interface, an output indicative of the health metric.

Description

Quantifying Hemodynamics from Non-Invasive Waveform Analysis Phase InformationCROSS REFERENCE TO RELATED APPLICATION

[0001] This application claims priority to US provisional application no. 63 / 636,213, filed on April 19, 2024, the entire contents of which are hereby incorporated herein.STATEMENT REGARDING FEDERALLY SPONSORED RESEARCH OR DEVELOPMENT

[0002] This invention was made with government support under Grant No.R44HL 140669 awarded by the National Institutes of Health. The government has certain rights in the invention.BACKGROUND

[0003] Some methods of determining blood volume status or related metrics of patient health involve measurement of pulmonary capillary wedge pressure (PCWP), central venous pressure (CVP), or central arterial pressure (CAP) via insertion of a catheter. Unfortunately, these measurements are cumbersome and invasive. Conventional vital sign monitoring fails to detect euvolemia or hypervolemia during resuscitation, often resulting in unguided and / or excessive fluid administration.SUMMARY

[0004] A first example includes a method comprising: generating, via a sensor, a signal representing vibrations originating from a blood vessel of a patient; processing the signal to yield amplitudes and phases of the vibrations that correspond to oscillation frequencies of the vibrations; identifying a first subset of the amplitudes and a second subset of the phases that correspond to the oscillation frequencies that are integer multiples of a heart rate of the patient; in response to providing the first subset of the amplitudes and the second subset of the phases to a computational model, determining a health metric for the patient; and generating, via a user interface, an output indicative of the health metric.

[0005] A second example includes a non-transitory computer readable medium storing instructions that, when executed by one or more processors of a computing device, cause the computing device to perform the method of the first example.

[0006] A third example includes a system comprising: a sensor; and a computing device comprising: one or more processors; a user interface; and a computer readable mediumstoring instructions that, when executed by the one or more processors, cause the computing device to perform the method of the first example.

[0007] A fourth example includes a method comprising: accessing multiple sets of training data, each set of the multiple sets indicating: first amplitudes that correspond to first oscillation frequencies of a first signal that represents vibrations originating from a first blood vessel of a reference patient, wherein the first oscillation frequencies are integer multiples of a first heart rate of the reference patient; first phases that correspond to the first oscillation frequencies; and a first health metric for the reference patient, the first health metric corresponding to the first amplitudes and the first phases; and using the multiple sets of training data to train a computational model to (a) access input indicating (i) second amplitudes that correspond to second oscillation frequencies of a second signal that represents vibrations originating from a second blood vessel of a test patient, wherein the second oscillation frequencies are integer multiples of a second heart rate of the test patient and (ii) second phases that correspond to the second oscillation frequencies, and (b) responsively generate an output indicating a second health metric for the test patient.

[0008] A fifth example is a non-transitory computer readable medium storing instructions that, when executed by one or more processors of a computing device, cause the computing device to perform the method of the fourth example.

[0009] A sixth example is a computing device comprising: one or more processors; and a computer readable medium storing instructions that, when executed by the one or more processors, cause the computing device to perform the method of the fourth example.

[0010] When the term “substantially” or “about” is used herein, it is meant that the recited characteristic, parameter, or value need not be achieved exactly, but that deviations or variations, including, for example, tolerances, measurement error, measurement accuracy limitations, and other factors known to those of skill in the art may occur in amounts that do not preclude the effect the characteristic was intended to provide. In some examples disclosed herein, “substantially” or “about” means within + / - 0-5% of the recited value.BRIEF DESCRIPTION OF THE FIGURES

[0011] The above, as well as additional, features will be better understood through the following illustrative and non-limiting detailed description of example embodiments, with reference to the appended drawings.

[0012] Figure l is a block diagram of a system, according to an example.

[0013] Figure 2 is a front view of a system, according to an example.

[0014] Figure 3 a graphical representation of a signal that represents vibrations originating from a blood vessel of a patient, according to an example.

[0015] Figure 4 shows many signal components that represent vibrations originating from the blood vessel of the patient, according to an example.

[0016] Figure 5 shows an amplitude spectrum of the signal shown in Figure 3, according to an example.

[0017] Figure 6 is a block diagram of a method, according to an example.

[0018] Figure 7 is a block diagram of a method, according to an example.

[0019] All the figures are schematic, not necessarily to scale, and generally only show parts which are necessary to elucidate example embodiments, wherein other parts may be omitted or merely suggested.DETAILED DESCRIPTION

[0020] As discussed above, determination of blood volume status via catheter insertion and measurement of PCWP, CVP, CAP, or other metrics has diagnostic value, but is inherently invasive and can be costly. Disclosed herein are methods and systems for using non-invasive venous waveform analysis (NIVA) to determine or detect blood volume status, PCWP, CVP, CAP, mechanical in vivo properties of a patient’s blood vessels, the presence of edema in the patient, and other metrics such as mean pulmonary arterial pressure, pulmonary artery diastolic pressure, left ventricular end diastolic pressure, left ventricular end diastolic volume, cardiac output, stroke volume, left atrial pressure, congestion, total blood volume, volume overload, dehydration, hemorrhage, and volume responsiveness. One or more of these metrics may be used to diagnose or treat various disorders that afflict a patient or be used for real time assessment and resuscitation of a patient.

[0021] Methods disclosed herein generally involve non-invasively measuring a peripheral arterial waveform (PAW) or a peripheral vein waveform (PVW) using a sensor positioned over a patient’s artery or vein. For example, a piezoelectric sensor can be positioned in contact with the patient’s skin to generate a signal. The signal represents vibrations originating from the blood vessel of the patient. The vibrations are generally caused by blood flowing though the vessel and / or the physiological reaction of the vessel or the surrounding tissue to the blood flow. The sensor generates a signal representing the vibrations as a function of time. A computing device receives the signal via a wired or wireless connection and processes the signal to yield amplitudes and phases of the vibrations that correspond to oscillation frequencies of the vibrations. For example, a Fourier transform is used to yield theamplitudes and phases corresponding to the oscillation frequencies of the signal. The computing device identifies a first subset of the amplitudes and a second subset of the phases that correspond to the oscillation frequencies that are integer multiples of a heart rate of the patient. Thereafter, the first subset of the amplitudes and the second subset of the phases that correspond to the oscillation frequencies that are integer multiples of the patient’s heart rate are provided as input to a computational model. Finally, the computational model responsively uses the input to determine a health metric for the patient. The computing device generates output indicative of the health metric via a user interface. Methods disclosed herein also involve the process of training the computational model.

[0022] The present disclosure generally relates to the quantification of volume status of a patient which may be experiencing hemorrhage, congestion / volume overload, or dehydration. The disclosure is related to real-time assessment of resuscitation in hospital or field settings. More specifically, the present disclosure relates to systems and methods of using peripherally sensed non-invasive venous waveform analysis to assess parameters such as pulmonary capillary wedge pressure, pulmonary artery diastolic pressure, mean pulmonary pressure, blood volume, and other metrics. In contrast to other methods of assessing a patient’s blood volume status, the methods of this disclosure utilize phase information within the signal generated by the sensor.

[0023] Fluid overload detection can be difficult. Fluid overload can be caused by administering excessive fluids or pathologic conditions. Unfortunately, many patients discharged after hospital admission for heart failure are re-admitted within 30 days for recurrent symptoms of congestion. A clinical trial showed a significant reduction in admissions related to heart failure with the use of a wireless implantable hemodynamic monitoring system. This was attributed to optimization of current heart failure treatments, specifically titration of diuretic regimens to reduce increased pulmonary pressures. Further analysis suggested diuresis guided by pulmonary pressure monitoring reduced likelihood of readmission within 30 days. While the results of these studies were promising, implantation of the device is invasive because it is performed through a right heart catheterization.

[0024] Heart failure (HF) related hospitalizations are commonplace, costly, and portend a poor prognosis. These hospitalizations are often precipitated by systemic congestion. Current measures of systemic congestion rely largely on clinical signs and symptoms including jugular venous pressure (JVP), subjective shortness of breath (SOB), weight changes, degree of peripheral edema, and laboratory values. Current guidelines recommend volume status should be monitored in heart failure patients using serial assessments of weight, edema, JVP,and orthopnea (SOB). Unfortunately, these signs and symptoms lack adequate sensitivity for detecting hemodynamic congestion. The primary focus of most acute HF admissions is the treatment of hemodynamic decongestion with diuretic therapy. Despite best efforts, some patients continue to be discharged with residual congestion and are at risk for hospital readmission.

[0025] Hemodynamic measures of congestion, though more sensitive and specific than clinical and laboratory measures, currently require invasive procedures thus limiting their use. Measurement of pulmonary capillary wedge pressure (PCWP) by right heart catheterization (RHC) is the primary clinical measurement used to estimate left-sided cardiac filling pressures in patients with heart failure. Despite this procedure being both costly and invasive, the utility of direct measurement of intracardiac pressures is supported by the observation that elevated cardiac filling pressures are associated with increased hospitalizations and mortality, and often precede the development of symptoms related to HF decompensation. Direct measurement of pulmonary pressures using an implantable monitor has been shown to reduce heart failure related hospitalization.

[0026] Additionally, hemorrhage and dehydration are critical issues that need to be addressed in a timely manner. Clinicians often rely on heart rate (HR) and blood pressure (BP) in the hospital setting for estimating fluid status and as a surrogate for quantification of blood loss and as a guide to fluid replacement in hypovolemic patients experiencing hemorrhage. However, HR and BP have routinely been shown to be unreliable at detecting reduced intravascular volume. In one clinical study, 100 blood donors were evaluated and HR and BP were found not sufficient for reliably detecting an acute 450 mL blood loss. Typically, a blood volume of 25-35% would be lost prior to a significant rise in HR. There are a host of pre-hospital or on-arrival scoring systems used as a screening tool to assess for blood loss. One such scoring system is the ratiometric relationship of pulse rate (PR) derived from pulse oximetry, to SBP, also known as the shock index. However, typically over 1000 mL of blood loss is needed before shock index alone is suggestive of hemorrhage. The ABC (assessment of blood consumption) and Trauma-Associated Severe Hemorrhage (TASH) scores are predictors for the need of massive transfusions (MT), but have not demonstrated the ability to detect small volumes of blood loss (< 1000 mL) prior to a need for MT. Therefore, there is a large unmet need for the advancement of non-invasive real-time assessment of volume status in patients.

[0027] Figure 1 is a block diagram of a system 10. The system 10 includes a computing device 100 and a sensor 114.

[0028] The computing device 100 includes one or more processors 102, a non-transitory computer readable medium 104 storing instructions 106 and a computational model 115, a communication interface 108, a display 110, and a user interface 112. Components of the computing device 100 are linked together by a system bus, network, or other connection mechanism 116. The computing device 100 can take the form of a desktop computer, a laptop computer, a tablet computer, or a smartphone, for example.

[0029] The one or more processors 102 can be any type of processor(s), such as a microprocessor, a digital signal processor, or a multicore processor, etc., coupled to the non- transitory computer readable medium 104.

[0030] The non-transitory computer readable medium 104 can be any type of memory, such as volatile memory like random access memory (RAM), dynamic random access memory (DRAM), static random access memory (SRAM), non-volatile memory like readonly memory (ROM), flash memory, magnetic or optical disks, or compact-disc read-only memory (CD-ROM), among other devices used to store data or programs on a temporary or permanent basis.

[0031] Additionally, the non-transitory computer readable medium 104 stores instructions 106. The instructions 106 are executable by the one or more processors 102 to cause the computing device 100 to perform any of the functions or methods described herein. The non- transitory computer readable medium 104 also stores the computational model 115 which can take the form of a machine learning model or a regression model.

[0032] The communication interface 108 includes hardware to enable communication within the computing device 100 and / or between the computing device 100 and one or more other devices, such as the sensor 114. The hardware can include transmitters, receivers, and antennas, for example. The communication interface 108 can be configured to facilitate communication with one or more other devices, in accordance with one or more wired or wireless communication protocols. For example, the communication interface 108 can be configured to facilitate wireless data communication for the computing device 100 according to one or more wireless communication standards, such as one or more Institute of Electrical and Electronics Engineers (IEEE) 801.11 standards, ZigBee standards, Bluetooth standards, etc. As another example, the communication interface 108 can be configured to facilitate wired data communication with one or more other devices.

[0033] The display 110 can be any type of display component configured to display data. As one example, the display 110 can include a touchscreen display. As another example, thedisplay 110 can include a flat-panel display, such as a liquid-crystal display (LCD) or a lightemitting diode (LED) display.

[0034] The user interface 112 can include one or more pieces of hardware used to provide data and control signals to the computing device 100. For instance, the user interface 112 can include a mouse or a pointing device, a keyboard or a keypad, a microphone, a touchpad, or a touchscreen, among other possible types of user input devices. Generally, the user interface 112 can enable an operator to interact with a graphical user interface (GUI) provided by the computing device 100 (e.g., displayed by the display 110).

[0035] The sensor 114 generally takes the form of a piezioelectric sensor, an optical sensor, a multiplexed optical array, a pressure sensor, a force resistive sensor, a tonometer, an ultrasound sensor, a capacitive sensor, or a pressure transducer. For example, the sensor 114 can be a piezoelectric sensor that communicates wirelessly or via a wired connection with the communication interface 108. Additionally or alternatively, the sensor 114 is securable to a skin of a patient via a strap and / or encapsulated within rubber, a polymer, polyurea, and / or silicone.

[0036] Figure 2 is a front view of the system 10. As shown, the system 10 includes the sensor 114 that communicates via a wired connection with the computing device 100. Also shown in Figure 2 are the user interface 112 and the display 110.

[0037] Figure 3 is a graphical representation of a signal 202 that represents vibrations originating from a blood vessel of a patient. The vertical axis represents vibration intensity in arbitrary units and the horizontal axis represents time in arbitrary units. In Figure 3, the signal 202 includes information about approximately one period of the patient’s heart beat. The sensor 114 generates the signal 202 while being worn by the patient (e.g., pressed against the patient with a pressure of 10 mmHg to 60 mmHg). The sensor 114 provides the signal 202 to the computing device 100. In various examples, the sensor 114 is pressed against a wrist, an ankle, an ear canal, an eye, or a neck of the patient while generating the signal 202. The signal 202 can represent vibrations originating from a peripheral vein or a peripheral artery of the patient. In another sense, the signal 202 indicates displacement of the sensor 114 as a function of time. In some examples, the signal 202 represents an average of many signal components.

[0038] Figure 4 shows many signal components 203 that represent vibrations originating from the blood vessel of the patient. Each of the signal components 203 represent a duration that is equal to a reciprocal of the heart rate (e.g., a period of the patient’s heart beat). For example, if the patient’s heart rate is 1.0 Hz, each signal component 203 represents onesecond of vibrations originating from the blood vessel of the patient. That is, each signal component 203 represents one full period of the patient's heart beat. The computing device 100 can average the signal components 203 to yield the signal 202.

[0039] Figure 5 shows an amplitude spectrum 206 of the signal 202. The vertical axis represents amplitude in arbitrary units and the horizontal axis represents frequency in arbitrary units. The computing device 100 processes the signal 202 to yield amplitudes and phases of the vibrations that correspond to oscillation frequencies of the vibrations. For example, the computing device 100 can perform a Fourier transform, a fast Fourier transform, a Hilbert-Huang transform, an empirical mode deconvolution, or a Laplace transform to yield the amplitudes and phases.

[0040] As an example, processing the signal 202 using a Fourier transform inherently generates the amplitudes and phases of the vibrations that correspond to oscillation frequencies of the vibrations. Referring to equation (1) below, x(t) is the signal 202 in the time domain and X(jco) is the signal 202 in the frequency domain. |X(j co)| is the frequency dependent magnitude of the signal 202 and (X(jco)) is the frequency dependent phase of the signal 202.

[0042] Next, the computing device 100 identifies a first subset of the amplitudes and a second subset of the phases that correspond to the oscillation frequencies that are integer multiples of a heart rate of the patient. With respect to the signal 202 and the amplitude spectrum 206 shown in Figure 5, the first subset of the amplitudes and the second subset of the phases are shown in the table below:

[0043] The computing device 100 can identify the subset of the amplitudes and the subset of the phases that correspond to integer multiples of the heart rate in several ways. In someexamples, the computing device 100 first identifies the frequency (FO) that corresponds to the heart rate of the patient, and then identifies the integer multiples of the frequency (FO). The computing device 100 can identify FO as being the most intense amplitude within a range of 0.3 Hz to 3.5 Hz. In other examples, the computing device 100 receives input via the user interface 112 that identifies F0. In another example, a dedicated pulsometer is used to determine F0. Other examples are possible. Equation (1) defines a relationship between the subset of the amplitudes and the subset of the phases such that the phases that correspond to the integer multiples of the frequency F0 can also be identified.

[0044] The computing device 100 uses the amplitudes and phases that correspond respectively to F0, Fl, F2, F3, F4, F5, and F6 as input for the computational model 115. In various examples, the computational model 115 is a regression model generated by regression analysis of data collected from multiple patients and / or multiple cardiac states. In other examples, the computational model is a machine learning model trained using data collected from multiple patients and / or multiple cardiac states. In response to receiving the input, the computational model 115 responsively determines a health metric for the patient. The health metric can indicate or be related to one or more of a pulmonary capillary wedge pressure (PCWP), a pulmonary artery diastolic pressure, a central venous pressure, a central arterial pressure, a mean pulmonary pressure, a blood volume, an intravascular volume status, a renal or hepatic function, autonomic function, intravascular blood clotting, or vascular compliance. The user interface 112 and / or the display 110 generates visual, audio, and / or tactile output indicative of the health metric.

[0045] In some examples, a doctor will commence, adjust, or cease a treatment for the patient based on the output. Such treatment could be for a condition of the patient, such has one or more of abnormal heart valve function, a comorbidity, dyspnea, volume overload, dehydration, hemorrhage, pulmonary embolism, deep vein thrombosis, or vascular endothelial function.

[0046] The computational model 115 is trained and / or iteratively adjusted to be useful for determining health metrics for patients. For example, the computing device 100 accesses multiple sets of training data. Each set of the training data corresponds to a single patient and a single session of data collection of signals generated by the sensor 114. The sets of training data as a whole can correspond to many patients (e.g., hundreds of patients) and to many cardiac states (e.g., relaxation, exertion, dehydration, sleep etc.).

[0047] Each set of the training data indicates amplitudes that correspond to oscillation frequencies of a first signal that represents vibrations originating from a first blood vessel of areference patient. The first oscillation frequencies are integer multiples of a first heart rate of the reference patient. Each set of the training data also indicates first phases that that correspond to the first oscillation frequencies. Lastly, each set of the training data indicates a first health metric for the reference patient. The first health metric corresponds to the first amplitudes and the first phases. That is, the first health metric was determined using conventional techniques such as catheterization. Thus, each set of the training data includes amplitudes and phases corresponding to integer multiples of the patient’s heart rate. These amplitudes and phases are derived from a signal that was generated by the sensor 114 while the patient was determined to be characterized by the health metric included in the set of the training data.

[0048] Next, the computing device 100 uses the multiple sets of training data to train the computational model 115 to access input and responsively generate an output. The input indicates second amplitudes that correspond to second oscillation frequencies of a second signal that represents vibrations originating from a second blood vessel of a test patient. The second oscillation frequencies are integer multiples of a second heart rate of the test patient. The input also includes second phases that correspond to the second oscillation frequencies. The computing device 100 trains the computational model 115 to responsively generate an output indicating a second health metric for the test patient.

[0049] That is, the computing device 100 trains the computational model 115 to identify patterns that relate (i) the amplitudes and the phases of the integer multiples of the patient’s heart rate and (ii) the patient’s health metric that corresponds to the amplitudes and the phases. The patterns can be mathematically defined as a regression model or a machine learning model. In some examples, each hidden layer of the computational model 115 is a hyperbolic tangent function of a weighted sum of the amplitudes and phases that constitute the input to the computational model 115 at runtime. The health metric of the test patient can be calculated in the form of a weighted sum of these hidden layers (e.g., eight hidden layers).

[0050] Figure 6 and Figure 7 are a block diagrams of a method 300 and a method 400, which in some examples are performed by the system 10. As shown in Figure 6 and Figure 7, the method 300 and the method 400 include one or more operations, functions, or actions as illustrated by blocks 302, 304, 306, 308, 310, 402, and 404. Although the blocks are illustrated in a sequential order, these blocks may also be performed in parallel, and / or in a different order than those described herein. Also, the various blocks may be combined into fewer blocks, divided into additional blocks, and / or removed based upon the desired implementation.

[0051] At block 302, the method 300 includes the system 10 generating, via the sensor 114, the signal 202 representing vibrations originating from a blood vessel of a patient. Functionality related to block 302 is described above with reference to Figures 3-4.

[0052] At block 304, the method 300 includes the computing device 100 processing the signal 202 to yield amplitudes and phases of the vibrations that correspond to oscillation frequencies of the vibrations. Functionality related to block 304 is described above with reference to Figure 5.

[0053] At block 306, the method 300 includes the computing device 100 identifying a first subset of the amplitudes and a second subset of the phases that correspond to the oscillation frequencies that are integer multiples of a heart rate of the patient. Functionality related to block 306 is described above with reference to Figure 5.

[0054] At block 308, the method 300 includes, in response to providing the first subset of the amplitudes and the second subset of the phases to the computational model 115, determining a health metric for the patient. Functionality related to block 308 is described above with reference to Figures 3-5.

[0055] At block 310, the method 300 includes the computing device 100 generating, via the user interface 112, an output indicative of the health metric. Functionality related to block 310 is described above with reference to Figures 3-5.

[0056] At block 402, the method 400 includes the computing device 100 accessing multiple sets of training data, each set of the multiple sets indicating first amplitudes, first phases, and a first health metric for a reference patient. The first amplitudes correspond to first oscillation frequencies of a first signal that represents vibrations originating from a first blood vessel of a reference patient. The first oscillation frequencies are integer multiples of a first heart rate of the reference patient. The first phases correspond to the first oscillation frequencies. The first health metric corresponds to the first amplitudes and the first phases. Functionality related to block 402 is described above with reference to Figures 3-5.

[0057] At block 404, the method 400 includes using the multiple sets of training data to train the computational model 115 to access input indicating second amplitudes and second phases and responsively generate an output indicating a second health metric for a test patient. The second amplitudes correspond to second oscillation frequencies of a second signal that represents vibrations originating from a second blood vessel of the test patient. The second oscillation frequencies are integer multiples of a second heart rate of the test patient. The second phases correspond to the second oscillation frequencies. Functionality related to block 404 is described above with reference to Figures 3-5.

[0058] While some embodiments have been illustrated and described in detail in the appended drawings and the foregoing description, such illustration and description are to be considered illustrative and not restrictive. Other variations to the disclosed embodiments can be understood and effected in practicing the claims, from a study of the drawings, the disclosure, and the appended claims. The mere fact that certain measures or features are recited in mutually different dependent claims does not indicate that a combination of these measures or features cannot be used. Any reference signs in the claims should not be construed as limiting the scope.

Claims

CLAIMSWhat is claimed is:

1. A method compri sing : generating, via a sensor, a signal representing vibrations originating from a blood vessel of a patient; processing the signal to yield amplitudes and phases of the vibrations that correspond to oscillation frequencies of the vibrations; identifying a first subset of the amplitudes and a second subset of the phases that correspond to the oscillation frequencies that are integer multiples of a heart rate of the patient; in response to providing the first subset of the amplitudes and the second subset of the phases to a computational model, determining a health metric for the patient; and generating, via a user interface, an output indicative of the health metric.

2. The method of claim 1, wherein the sensor comprises a piezoelectric sensor.

3. The method of claim 1, wherein the sensor comprises an optical sensor, a multiplexed optical array, a pressure sensor, a force resistive sensor, a tonometer, an ultrasound sensor, a capacitive sensor, or a pressure transducer.

4. The method of any one of claims 1-3, further comprising pressing the sensor against the patient while generating the signal.

5. The method of claim 4, wherein pressing the sensor against the patient comprises pressing the sensor against the patient with a pressure of 10 mmHg to 60 mmHg.

6. The method of any one of claims 4-5, wherein pressing the sensor against the patient comprises pressing the sensor to a wrist, an ankle, an ear canal, an eye, or a neck of the patient.

7. The method of any one of claims 1-6, wherein the blood vessel comprises a peripheral vein.

8. The method of any one of claims 1-6, wherein the blood vessel comprises a peripheral artery.

9. The method of any one of claims 1-8, wherein the signal indicates displacement of the sensor as a function of time.

10. The method of any one of claims 1-9, wherein processing the signal comprises performing a Fourier transform upon the signal.

11. The method of any one of claims 1-10, wherein processing the signal comprises performing a Fast Fourier transform (FFT) upon the signal.

12. The method of any one of claims 1-10, wherein processing the signal comprises performing a Hilbert-Huang transform (HHT) upon the signal.

13. The method of any one of claims 1-10, wherein processing the signal comprises performing an empirical mode deconvolution upon the signal.

14. The method of any one of claims 1-10, wherein processing the signal comprises performing a Laplace transform upon the signal.

15. The method of any one of claims 1-14, wherein identifying the first subset of amplitudes comprises identifying a first amplitude corresponding to a first frequency that is the heart rate of the patient.

16. The method of claim 15, wherein identifying the first subset of amplitudes comprises identifying a second amplitude corresponding to a second frequency that is an integer multiple of the first frequency.

17. The method of any one of claims 15-16, wherein identifying the first amplitude comprises identifying the first amplitude as corresponding to a most intense amplitude within a range of 0.3 Hz to 3.5 Hz.

18. The method of any one of claims 15-16, further comprising receiving, via the user interface, an input identifying the first frequency, wherein identifying the first amplitude comprises identifying the first amplitude based on the input.

19. The method of any one of claims 15-16, further comprising receiving from an additional sensor an input identifying the first frequency, wherein identifying the first amplitude comprises identifying the first amplitude based on the input.

20. The method of any one of claims 1-19, further comprising generating signal components representing vibrations originating from the blood vessel of the patient, wherein the signal components each represent a duration that is equal to a reciprocal of the heart rate, wherein generating the signal comprises averaging the signal components.

21. The method of any one of claims 1-20, wherein the computational model is a regression model generated by regression analysis of data collected from multiple patients.

22. The method of any one of claims 1-20, wherein the computational model is a machine learning model trained using data collected from multiple patients.

23. The method of any one of claims 1-22, further comprising commencing, adjusting, or ceasing a treatment for the patient based on the output.

24. The method of claim 23, wherein the treatment is a treatment for a condition of the patient, the condition including one or more of abnormal heart valve function, a comorbidity, dyspnea, volume overload / congestion, dehydration, hemorrhage, pulmonary embolism, deep vein thrombosis, or vascular endothelial function.

25. The method of any one of claims 1-24, wherein the health metric is indicative of one or more of a pulmonary capillary wedge pressure (PCWP), a pulmonary artery diastolic pressure, a central venous pressure, a central arterial pressure, a mean pulmonary pressure, a blood volume, an intravascular volume status, renal or hepatic function, autonomic function, or vascular compliance.

26. A non-transitory computer readable medium storing instructions that, when executed by one or more processors of a computing device, cause the computing device to perform the method of any one of claims 1-25.

27. A system comprising: a sensor; and a computing device comprising: one or more processors; a user interface; and a computer readable medium storing instructions that, when executed by the one or more processors, cause the computing device to perform the method of any one of claims 1-25.

28. A method comprising: accessing multiple sets of training data, each set of the multiple sets indicating: first amplitudes that correspond to first oscillation frequencies of a first signal that represents vibrations originating from a first blood vessel of a reference patient, wherein the first oscillation frequencies are integer multiples of a first heart rate of the reference patient; first phases that correspond to the first oscillation frequencies; and a first health metric for the reference patient, the first health metric corresponding to the first amplitudes and the first phases; and using the multiple sets of training data to train a computational model to (a) access input indicating (i) second amplitudes that correspond to second oscillation frequencies of a second signal that represents vibrations originating from a second blood vessel of a test patient, wherein the second oscillation frequencies are integer multiples of a second heart rate of the test patient and (ii) second phases that correspond to the second oscillation frequencies, and (b) responsively generate an output indicating a second health metric for the test patient.

29. The method of claim 28, wherein the second blood vessel comprises a peripheral vein.

30. The method of claim 28, wherein the second blood vessel comprises a peripheral artery.

31. The method of any one of claims 28-30, wherein the computational model is a machine learning model trained using data collected from multiple patients.

32. A non-transitory computer readable medium storing instructions that, when executed by one or more processors of a computing device, cause the computing device to perform the method of any one of claims 28-31.

33. A computing device comprising: one or more processors; and a computer readable medium storing instructions that, when executed by the one or more processors, cause the computing device to perform the method of any one of claims 28- 31.

Images