Phonocardiogram (PCG) signal processing systems and methods for determining heart tissue and valve blood flow parameters

Abstract

A non-invasive, passive, fully automated heart sound-based system and method is provided that provides estimation of blood velocity, tissue movement, and lumen size parameters for heart assessment. The system uses a computer processor and software to receive PCG acoustic signals from one or more sensors and simultaneously receive electrocardiogram (ECG) signals from one or more sensors attached to the patient. A phonocardiogram (PCG) processing system and method computes metrics of indices of echocardiogram parameters of heart tissue movement and valve blood flow for evaluation.

Description

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application claims priority to and the benefit of U.S. Provisional Patent Application Serial No. 63 / 332,074, filed on April 18, 2022, which is incorporated herein by reference in its entirety.

[0003] STATEMENT REGARDING FEDERALLY SPONSORED RESEARCH OR DEVELOPMENT

[0004] not applicable

[0005] Notice of Copyrighted Material

[0006] Portions of the material in this patent document are copyrighted under the copyright laws of the United States and other countries. The copyright owner has no objection to the facsimile reproduction by anyone of the patent document or the patent disclosure as it appears in the United States Patent and Trademark Office publicly available file or records, but otherwise reserves all copyright rights. The copyright owner hereby does not waive any rights to have this patent document maintained in confidence, including but not limited to the right pursuant to 37 CFR §1.14. Technical Field

[0007] The present technology relates generally to systems and methods for providing cardiac assessments, and more particularly to non-invasive devices and methods for generating cardiac assessments by calculating values ​​for blood velocity and tissue motion based on heart sounds. Background Art

[0008] The heart is a muscle that pumps blood through the body and is divided into four chambers and has four valves, Figure 1A and Figure 1B During one cardiac cycle, deoxygenated blood flows from the body into the right side of the heart, through the lungs into the left side of the heart, and is then pumped from the heart back to the rest of the body, as shown in Figure 1. Figure 2 shown.

[0009] The pumping efficiency of the heart is usually measured by calculating the left ventricular (LV) ejection fraction, which is defined as the fraction of blood in the left ventricle that is pumped out of the heart to the body during each cardiac cycle. This measurement can be represented by the following equation:

[0010]

[0011] In some cases, the heart's ability to pump is compromised because the muscle of the left ventricle cannot fully relax or is too stiff to allow normal filling of blood before pumping, which increases LV filling pressures. This condition is called LV diastolic dysfunction and is one of the biggest causes of the development of heart failure in humans.

[0012] Patients who exhibit symptoms of heart failure (such as shortness of breath, fatigue, or decreased exercise tolerance) are recommended to undergo an ultrasound interrogation of the heart, also known as an echocardiogram. During this imaging technique, technicians visualize the blood flow and muscle movement of the heart, and then calculate some parameter values ​​to assess LV size, wall motion, and valve blood flow patterns. Such echocardiography-based parameters include peak E velocity (peak velocity of blood flow through the mitral valve during early diastole), E / A ratio (ratio of early peak flow velocity to late peak flow velocity through the mitral valve during diastole), e' velocity (average flow velocity through the mitral valve during early diastole), peak TR velocity (peak velocity of blood regurgitation through the tricuspid valve during systole), and LAVi (maximum volume of the left atrium associated with body surface area). The cutoff values ​​for each of these parameters are then analyzed in order to grade the degree of diastolic dysfunction and estimate LV filling pressures. Although physicians can analyze these echocardiographic parameters independently, the accuracy of calculating these parameters depends on the quality of the echocardiographic images and the technician's level of training and experience. Therefore, echocardiography is a resource-intensive tool.

[0013] Although there are non-invasive systems that are capable of cardiac assessment, existing systems like ultrasound are not passive and do not use modalities other than heart sounds, such as ballistocardiography, pressocardiography, or bioimpedance. Existing non-invasive and passive systems based on heart sounds generally claim the ability to perform cardiac assessment and abnormal sound detection, and may also identify cardiac pain by name, but they do not explicitly claim or provide any evidence for the ability to assess cardiac tissue and blood flow parameter calculations. In other words, there are existing systems based on heart sounds that calculate blood volume (such as stroke volume or ejection fraction) and blood pressure (such as systolic or diastolic blood pressure or pulmonary pressure), but do not calculate blood flow velocity (such as blood flow velocity through the heart valves), cardiac tissue motion (such as valve or heart tissue deflection), or ventricular size (such as atrial volume).

[0014] Therefore, new non-invasive, low-cost and easy-to-operate systems, devices and protocols are needed to allow reliable and accurate measurement of blood flow velocity, cardiac tissue motion and ventricular dimensions. Summary of the invention

[0015] A non-invasive and automated phonocardiogram (PCG) processing system and method is provided that calculates proxy metrics for echocardiographic parameters of cardiac tissue motion and valvular blood flow. The system uses a computer processor and software to receive PCG acoustic signals from one or more sensors and simultaneously receive electrocardiogram (ECG) signals from one or more sensors attached to a patient.

[0016] Preferably, the collected PCG signal is denoised and processed into one or more of the time characteristics, amplitude characteristics, frequency characteristics or spectral entropy characteristics of each heartbeat of the same patient.

[0017] The processed features are then converted into one of several cardiac tissue and valve blood flow parameter analogs (index metrics) based on a set of predetermined conversion equations. In one embodiment, the amplitude features of all heartbeats of the same patient are processed into an index metric for the peak velocity of blood flow through the patient's mitral valve during early diastole (peak E velocity).

[0018] In another embodiment, the extracted frequency features of all heartbeats of the same patient are processed into index metrics for the mean flow velocity through the patient's mitral valve during early diastole (e' velocity) and the maximum volume of the patient's left atrium correlated to the patient's body surface area (LAVi).

[0019] In another embodiment, the spectral entropy features of all heartbeats of the same subject are processed into index metrics for the ratio of early peak flow velocity to late peak flow velocity through the patient's mitral valve during diastole (E / A ratio) and the peak velocity of blood regurgitation through the patient's tricuspid valve during systole (peak TR velocity).

[0020] In contrast to existing methods and systems, the present technology provides automated methods and systems for cardiac assessment by calculating values ​​of blood velocity and tissue motion based on collected heart sound (phonocardiogram PCG) data. PCG signal acquisition is passive, that is, it does not involve any application of energy to the body (unlike echocardiography) and is non-invasive. Instead, the method only involves recording the sounds generated by the heart over a period of time. The methods and systems proposed in the present disclosure require signal processing that goes beyond simple mental processes and cannot be accomplished using simple calculations or observations. These methods and systems are automated and do not require expert supervision. These signal processing methods and systems have been developed using insights from real-world clinical PCG data. The technology is based on the concept of calculating these blood velocity and tissue motion parameters based on heart sounds, and on their utility in helping healthcare practitioners diagnose and assess heart disease.

[0021] Other aspects of the technology described herein will be presented in the following portions of the specification, wherein the detailed description is for the purpose of fully disclosing preferred embodiments of the technology without limiting the same. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] A more complete understanding of the present invention will be obtained by referring to the following drawings which are provided for illustrative purposes only.

[0023] Technology described:

[0024] Figure 1A is a schematic cross-sectional view of the human heart anatomy showing the two upper chambers (atria) and the two lower chambers (ventricles).

[0025] Figure 1B is a schematic cross-sectional view of the human heart anatomy showing the four heart valves.

[0026] Figure 2 It is a conceptual description of the circulation of blood through the heart, lungs, and organs.

[0027] Figure 3 is a schematic diagram illustrating methods and systems according to a technique for using heart sound data to calculate corresponding indices for echocardiographic parameters for assessing left ventricular function in heart failure.

[0028] Figure 4 is a schematic diagram of a signal processor / diagnostic processor architecture according to one embodiment of the present technology.

[0029] Figure 5 is a conceptual diagram showing the general placement of four acoustic sensors for collecting PCG signals and three electrodes for collecting ECG signals using a signal collection sensor according to one embodiment of the present technology.

[0030] Figure 6 is a conceptual diagram of modules and associated module functions of a PCG signal processor according to an embodiment of the present technology. The input of the PCG signal processor includes a PCG signal acquired from a subject (upper left), and the output includes an indicator of an echocardiographic parameter provided to a care provider (lower right).

[0031] Figure 7 is a functional block diagram of a non-invasive and automated phonocardiogram (PCG) processing method for calculating index metrics for echocardiographic parameters such as cardiac tissue motion and valvular blood flow.

[0032] Figure 8 It is from Figure 5 Schematic diagram of the PCG signal waveform (top) and ECG signal waveform (bottom) of three consecutive heart beats obtained by the signal sensor shown.

[0033] Fig. 9 is an example of a PCG signal of a heartbeat that has been filtered and segmented by a PCG signal processor. The schematic diagram shows the diastolic interval, the first heart sound (S1), the systolic interval and the second heart sound (S2).

[0034] Fig. 10A is a functional block diagram of the signal processing steps performed by the PCG signal processor for amplitude feature extraction.

[0035] Fig. 10B is a functional block diagram of the signal processing steps performed by the PCG signal processor for frequency feature extraction.

[0036] Fig. 10C is a functional block diagram of the signal processing steps performed by the PCG signal processor for spectral entropy feature extraction.

[0037] Fig.11A is a functional block diagram of the signal processing steps performed by the PCG signal processor for the calculation of the index metric for Peak E Velocity.

[0038] Fig. 11B is a functional block diagram of the signal processing steps performed by the PCG signal processor for the calculation of the E / A ratio indicator metric.

[0039] Fig. 11C is a functional block diagram of the signal processing steps performed by the PCG signal processor for the calculation of the index metric of e' velocity.

[0040] Fig.11D is a functional block diagram of the signal processing steps performed by the PCG signal processor for the calculation of the index metric for peak TR velocity.

[0041] Fig.11E Figure 1 is a functional block diagram of the signal processing steps performed by the PCG signal processor for the calculation of the LAVi indicator metric. The processing steps show the signal type used (raw or denoised), source (aortic, pulmonary or mitral), heartbeat (total or mass), signal interval (systolic or diastolic), and characteristics (amplitude, frequency or spectral entropy).

[0042] Fig. 12A Concept illustration of the diagnostic criteria for assessing LV diastolic function as described in the 2016 joint recommendations of the American Society of Echocardiography (ASE) and the European Association of Cardiovascular Imaging (EACVI).

[0043] Fig. 12B is a conceptual diagram of the diagnostic criteria used to estimate mean left atrial pressure as a reliable approximation of LV filling pressure from the same source. The cutoff value for "mean" e' velocity was chosen as the average of the cutoff values ​​for "septal" and "lateral" e' velocities. DETAILED DESCRIPTION

[0044] Referring more particularly to the drawings, for purposes of illustration, a system and method for phonocardiogram (PCG) signal processing that calculates a proxy metric for an echocardiogram parameter is generally shown. Figures 1A to 12B Several embodiments of the present technology are generally described in order to illustrate the features and functions of devices, systems and methods. It will be appreciated that the methods may vary with respect to specific steps and sequences, and the systems and devices may vary with respect to structural details, without departing from the basic concepts as disclosed herein. The method steps are merely examples of the order in which these steps may occur. These steps may occur in any desired order so that they still perform the objectives of the claimed technology.

[0045] Now go to Figure 4 , schematically shows the overall structure of an apparatus and system 10 for calculating an index metric for an echocardiographic parameter based on a phonocardiogram (PCG) signal. The apparatus of the system 10 has a control and computing device 12, which has one or more ECG sensors 14 that generate electrocardiogram (ECG) signals and one or more PCG sensors 16 that generate phonocardiogram (PCG) signals, and the ECG signals and PCG signals are inputs 18 to the apparatus 12. The control / processing device 12 has at least one processor 20, a non-transitory memory 22, and application software 24 with programming, which processes the inputs 18 of the ECG signals and PCG signals and extracts features, creates or applies models, and calculates index metrics as described herein. The apparatus may also have a display 26. The display 26 can be used to output the calculated analysis or diagnosis results to a healthcare provider.

[0046] In another embodiment, the diagnostic processor may be a separate device or part of the PCG signal processor, in which case the application software will also contain instructions for performing the diagnostic processing functions described herein.

[0047] In one embodiment, Figure 3As generally shown in FIG. 1 , a wearable system 10 is provided that uses heart sound (phonocardiogram PCG) signals to calculate PCG-based indices as equivalents to echocardiographic parameters. The calculation utilizes PCG signal analysis techniques that determine physiological properties of blood flow and muscle movement, and the fully automatic system operates without any expert supervision. The physiological properties used to calculate the PCG-based index metrics are substantially the same as those that would otherwise be measured by echocardiographic parameters. The calculation involves complex signal processing that cannot be accomplished simply as a mental process, mere observation, or with pen and paper. The resulting index metric values ​​have a heart failure screening accuracy comparable to that of echocardiographic parameters, and the system can be used in a clinical care setting to assess heart failure patients.

[0048] exist Figure 5 In the illustrated embodiment, three ECG sensors 14 are placed at specific locations 30 and four PCG acoustic sensors 16 are placed at other locations 32 on the left and right sides of the patient's body 28 for signal acquisition by the control / processing device 12. Figure 5 The number and location of sensors generally described in are preferred, but the number and location of each sensor type may vary.

[0049] exist Figure 6 An overview of the sensor signal acquisition and processing steps is generally set forth in the context of use. The processing modules and associated module functions of the PCG signal processor 34 are shown from patient acquisition to evaluation of the results by a healthcare practitioner 42. In this embodiment, the sensor signal from the patient 28 is received by the PCG signal processor 34 for processing. At the first module 36, filtering and noise reduction are optionally utilized to improve the signal. In one embodiment, a spectral noise reduction algorithm commonly used in speech processing is utilized to remove noise artifacts from speech, motion, and other interferences from the original PCG signal using bandpass filtering and noise reduction to obtain a PCG signal that is qualitatively higher in audio fidelity than the original signal.

[0050] Then, in Figure 6 In the illustrated embodiment, the denoised signal is processed using a heartbeat segmentation and quality assurance module 38. In this module, the start and end times of each heartbeat in the PCG signal are identified using the ECG signal as a reference. The start of the R wave in each cardiac cycle of the ECG signal can be considered as a transition point between the end of one heartbeat and the beginning of the next heartbeat.

[0051] The quality of individual heartbeats may also be considered for identification and further processing. A heartbeat may be identified as "good quality" if: (1) both S1 and S2 are successfully identified, (2) there is no signal drift in the systolic and diastolic intervals, and (3) the heartbeat duration is within ±20% of the median duration for the subject.

[0052] The third processing module 40 is used for feature extraction and index metric calculation. In one embodiment, the PCG signal processor processes the PCG signal to convert the signal into three types of features. It is important to note that the feature is not a segment of the PCG signal, but represents a distinguishing property obtained from a segment of the PCG signal. Therefore, feature calculation is not a mental process and cannot be completed using simple observation or calculation, but is completed using the complex signal processing techniques described here. Once the features are extracted, the PCG signal processor processes those features to convert the features into PCG-based index metrics for each echocardiographic parameter. These PCG-based index metrics are analogs of cardiac tissue and valve blood flow parameters. Index metric calculation involves extracting features that characterize physiological phenomena (such as cardiac pressure gradients, tissue motion, and blood flow that are otherwise measured by echocardiographic parameters). In one embodiment, the features include amplitude features, frequency features, and spectral entropy features. In a preferred embodiment, the index metrics include peak E velocity, e' velocity, LAVi, E / A ratio, and peak TR velocity.

[0053] Figure 7 An alternative method 44 of calculating an index metric for an echocardiographic parameter from a phonocardiogram (PCG) signal is shown. At block 46, PCG acoustic signals from one or more acoustic sensors are received or acquired simultaneously with electrocardiogram (ECG) signals from one or more sensors attached to a subject.

[0054] Optionally, at box 48, the simultaneously received signals may be denoised using filters or other signal quality improvement methods. In one embodiment, the PCG signal is denoised by applying a bandpass filter with cutoff frequencies of 25 Hz and 140 Hz, and then applying a spectral noise reduction algorithm that involves estimating the noise spectrum during brief pauses in heart sound activity. Thereafter, at box 48, the estimate is subtracted from the spectrum of the entire signal to obtain a clean heart sound signal.

[0055] In another embodiment, a fourth order Butterworth low pass filter is applied to the phonocardiogram signal and the resulting signal envelope is divided into multiple frames. A total noise spectrum is calculated and subtracted from each discrete Fourier transform of the signal envelope frame, and the noise free frames are added together to reconstruct the phonocardiogram signal from the entire PCG signal.

[0056] Other denoising methods may be employed at block 48 to produce a clean heart sound signal. Fig. 9 A schematic diagram of a PCG signal for a heartbeat that has been filtered and segmented by a PCG signal processor is shown in FIG. , which shows the diastolic interval, the first heart sound ( S1 ), the systolic interval, and the second heart sound ( S2 ).

[0057] exist Figure 7 At block 50, the original or denoised PCG signal is processed into one or more of time features, amplitude features, frequency features and / or spectral entropy features for each heartbeat of the same subject. The feature extraction process can be selected using a specific case of the original or denoised phonocardiogram signal, where the original or denoised phonocardiogram signal belongs to all or exclusively good quality heartbeats depending on the underlying physiological function being characterized.

[0058] In a preferred embodiment, Fig. 10A Following the steps shown in Figure 7 The PCG signal is processed into a time feature at box 50 of the present invention. The process 60 begins by acquiring a PCG signal corresponding to a heartbeat at box 62. The acquired signal is processed into an amplitude feature by first applying a Hilbert transform to the PCG signal that requires signal processing in the time domain and the frequency domain at box 64. At box 66, a low pass filter with a cutoff frequency of 51 Hz is applied to generate a signal envelope. The 60th percentile value of the resulting signal envelope for the heartbeat is then calculated at box 68 to generate an amplitude feature 70.

[0059] Also refer to Fig. 10B , showing that in Figure 7 The present invention is an embodiment of a process 72 for generating a frequency signature at block 50 of the present invention. The frequency signature may be processed from the PCG signal by applying a Hamming window at block 76 to a segment of the PCG signal acquired at block 74 corresponding to a heartbeat. A 64-point discrete Fourier transform is applied at block 78, and the centroid of the frequency distribution of the heartbeat between 16 Hz and 160 Hz is calculated at block 80 to generate a frequency signature 82.

[0060] In a preferred embodiment, the Fig. 10C The spectral features are calculated by the process 84 shown in FIG. A PCG signal corresponding to one heartbeat is collected at box 86, and the PCG signal is processed into a spectral entropy feature by the following steps: First, a signal distribution probability estimate is obtained from the PCG signal at box 88, and a negative product of the signal probability distribution estimate and its logarithm is calculated for the heartbeat at box 90 to generate a spectral entropy feature 92.

[0061] Then, if Figure 7As shown in the box 52 of , one or more of these calculated features can be used to formulate at least one of several index metrics. Although a specific process for generating these features is shown, other feature extraction processes can be used at the box 50 to obtain features for formulating index metrics at the box 52.

[0062] exist Figure 7 The indicator metric selected and formulated at block 52 of is a measure of a physiological phenomenon that can be compared to a standard. One or more indicator metrics may be formulated at block 52, and Figure 7 At block 54 , various aspects of the condition of the subject patient's heart may be assessed.

[0063] exist FIG. 11A to FIG. 11E Describes the use of FIG. 10A to FIG. 10C The PCG-based index metrics for several transformations of the extracted features of are presented to illustrate the method. The formulated PCG-based index metrics for peak E velocity, e' velocity, LAVi, E / A ratio, and peak TR velocity can be output to a diagnostic processor configured to be used in Figure 7 At box 54, diastolic function and left atrial pressure are assessed based on PCG-based index metrics.

[0064] In one embodiment, Fig.11A 1 , a process 100 for determining a PCG-based index metric 110 for deriving an echocardiogram-based peak E velocity parameter is shown. The index metric is obtained by first acquiring a denoised PCG signal for the diastolic interval of good quality heartbeats at box 102. The ratio of the amplitude characteristic values ​​of the pulmonary signal and the aortic signal is calculated for each heartbeat at box 104. Then, the average of the available ratios for all good quality heartbeats of each subject is determined at box 106. Then, the average of the available ratios for all good quality heartbeats of each subject is determined at box 107. Fig.11A The average is fit to a predefined linear model using a predetermined transformation equation at box 108 to produce an indicator 110 of peak E velocity.

[0065] Can be used Fig. 11B The process 112 shown in FIG. 1 is used to determine a PCG-based index metric 122 for an echocardiogram-based E / A ratio parameter. In this embodiment, the index metric for the E / A ratio parameter is determined by obtaining a raw lung PCG signal for a good-quality heartbeat at box 114. The ratio of the spectral entropy eigenvalues ​​of the early diastolic interval and the late diastolic interval of each heartbeat is determined at box 116. At box 118, the average of the available ratios for all good-quality heartbeats of each subject is calculated. Then, at box 120, the calculated average is fit to a predefined linear model using a predetermined transformation equation to produce the E / A ratio index metric 122.

[0066] Fig. 11C An embodiment 124 for processing an index metric 134 for an e' velocity parameter is shown. Here, a PCG-based index metric for an echocardiography-based e' velocity parameter is determined by obtaining a de-noised aortic PCG signal for all heartbeats at box 126, calculating frequency characteristic values ​​of the late systolic interval for each heartbeat at box 128, calculating the average of the available ratios for all heartbeats for each subject at box 130, and then fitting the calculated average to a predefined linear model using a predetermined transformation equation at box 132 to produce an index 134 for e' velocity.

[0067] Can be used Fig.11D The process 136 shown in FIG. 1 is used to determine a PCG-based index metric 146 for an echocardiography-based peak TR velocity parameter. In this embodiment, a PCG-based index metric for an echocardiography-based peak TR velocity parameter is determined by obtaining a denoised PCG signal for the diastolic interval of good quality heartbeats at box 138. The ratio of the spectral entropy eigenvalues ​​of the pulmonary signal and the aortic signal is then calculated for each heartbeat at box 140. The average of the available ratios for all good quality heartbeats for each subject is calculated at box 142. The calculated average is then fitted to a predefined linear model using a predetermined transformation equation at box 144 to produce an index 146 for peak TR velocity.

[0068] Can be used Fig.11E The process 148 shown in FIG. 1 is used to determine a PCG-based index metric 158 for an echocardiogram-based LAVi parameter. In this embodiment, at box 150, the LAVi parameter is determined by obtaining a raw mitral valve PCG signal for all heartbeats. Then, at box 152, the frequency characteristic value of the early diastolic interval of each heartbeat is calculated. At box 154, the average of the available ratios for all heartbeats of each subject is calculated. At box 156, the calculated average is fit to a predefined linear model using a predetermined transformation equation to produce an index 158 for LAVi.

[0069] Preferably use training data to generate FIG. 11A to FIG. 11E The process shown is a predefined linear model to determine a transformation equation that can be used to calculate the index metric for any object that has never been seen before, as is the intent of the invention presented herein. In other embodiments, the signal processing is accompanied by advanced machine learning methods, rather than linear regression, to provide accurate calculation of the index metric in these processing steps.

[0070] In embodiments of such a system where the user can select a set of sensors with characteristics different from those applied here, it is expected that the user will have to train the system again. This is because the characteristics of the known acoustic sensor signals are determined by the specific characteristics of the selected sensors. Therefore, the linear model and the resulting transformation equations must be trained using linear regression using a training data set obtained using signal acquisition of the selected sensors applied to an object for which corresponding object condition realities have also been obtained.

[0071] Therefore, these PCG-based systems and methods can be used as part of a routine assessment of patients presenting with symptoms of dyspnea or heart failure, and can help them engage in an accelerated care pathway. The device and method are the first to use heart sounds for non-invasive and passive calculations of cardiac tissue and valve blood flow parameters. The end-to-end PCG signal processing, feature extraction, and index metric calculation algorithms for LV diastolic function assessment can be operated in a fully automatic manner without expert supervision. Although the clinical value of the index metric is determined using the 2016ASE / EACVI algorithm, the index metric calculation itself is independent of the algorithm and is therefore not affected by any guideline revisions that may be introduced in the future. This demonstrates the utility and potential of the proposed PCG-based system in providing echocardiogram-like parameters to interpreting physicians within a few minutes of signal acquisition in a real-world environment in a hospital or clinic, thereby allowing individuals with heart failure and other heart diseases to engage in an accelerated care pathway.

[0072] The technology described herein may be better understood with reference to the accompanying examples, which are provided for illustrative purposes only and should not be construed in any sense to limit the scope of the technology described herein, as defined in the appended claims.

[0073] Example 1

[0074] In order to prove the operating principle of the technology, the system was designed and verified using a research population of 34 adult inpatients who were scheduled for right heart catheterization at Oregon Health and Science University Hospital (Portland, OR). An echocardiogram report was obtained for each subject, which consisted of 2-dimensional and Doppler parameters from a transthoracic examination performed in close proximity to the right heart catheterization. Each report included one or more of the five parameters based on the quality of the echocardiogram study. Referring to (Table 1). The echocardiogram reports of these 34 subjects in the subject population constituted the subject status actual data.

[0075] The signal acquisition, filtering, heartbeat segmentation, and quality assurance features of individuals from this group are also presented and evaluated.During signal acquisition, the subjects lay supine on a catheterization laboratory bed.

[0076] Use four commonly Figure 5 Acoustic sensors placed as shown collected phonocardiogram (PCG) signals from each subject at a sampling rate of 512 Hz. Each sensor consisted of an electret microphone housed in an acrylonitrile-butadiene-styrene copolymer plastic body with a 0.4 mm thick black nitrile rubber membrane at one end.

[0077] The sensor is placed membrane-side down by the care provider at four traditional auscultatory points on the chest wall: aorta (2nd intercostal space, right sternal border), lung (2nd intercostal space, left sternal border), tricuspid valve (4th intercostal space, left sternal border), and mitral valve (5th intercostal space, left midclavicular line). Sensor placement is determined relative to the suprasternal notch and these locations do not require provider intervention. Use Figure 5 The three ECG electrodes shown in the figure simultaneously collect ECG signals at a sampling rate of 300 Hz. The electrodes are placed adjacently on both upper limbs and the (lower left) abdomen. The PCG signal acquisition and ECG signal acquisition duration for each subject ranged from 4 minutes to 80 minutes according to the catheterization laboratory schedule, and then these signals were stored in Matlab (MathWorks, MA) for offline analysis. Example PCG signal waveforms and ECG signal waveforms from continuous heartbeats are shown in Figure 8 As described above, the PCG data and ECG data of each of the 34 subjects in the subject group obtained by such signal acquisition together with the above-mentioned subject condition actual situation data constitute a training data set.

[0078] The PCG signal processor is designed to perform Figure 6 and Figure 7 The general PCG signal processing steps shown in , namely, filtering and noise reduction, heartbeat segmentation and quality assurance, and feature extraction and indicator metric calculation.

[0079] To demonstrate the signal filtering and noise reduction features, the signals collected from each subject were denoised. Noise artifacts from speech, motion, and other interference were removed from the original PCG signal using filtering and noise reduction to obtain a PCG signal that is qualitatively higher in audio fidelity than the original signal. In this illustration, the first step involves applying a bandpass filter with cutoff frequencies of 25 Hz and 140 Hz to retain the maximum amount of heart sound information while removing most low-frequency and high-frequency noise artifacts.

[0080] The remaining noise artifacts in the mid-frequency range overlap with the frequency range of heart sounds and are removed by applying a spectral noise reduction algorithm commonly used in speech processing. This technique involves estimating the noise spectrum during brief pauses in heart sound activity and then subtracting this estimate from the spectrum of the entire signal to obtain a clean heart sound signal. Regions of the PCG signal corresponding to pauses in heart sound activity are identified in the amplitude distribution of the signal envelope on a frame-by-frame basis. For this purpose, a fourth-order Butterworth low-pass filter with a cutoff frequency of 38 Hz is applied to the phonocardiogram signal. The signal envelope obtained as a result is then divided into 93 milliseconds long frames with an overlap of 31 milliseconds (33%) between adjacent frames. The frame and overlap lengths are determined empirically to provide optimal noise reduction.

[0081] Next, the amplitude distribution of the signal envelope is obtained by arranging the root mean square amplitude values ​​of each frame in increasing order. The amplitude distribution has a roughly bimodal shape, with one peak being a lower amplitude value corresponding to the pause of heart sound activity and the other peak being a higher amplitude value corresponding to physiological heart sounds and pathological heart sounds.

[0082] The individual spectra of the selected frames in the first peak are calculated using a discrete Fourier transform, and then these spectra are averaged to approximate an overall noise spectrum of the PCG signal. This average noise spectrum is then subtracted from the individual discrete Fourier transforms of all available signal envelope frames (including those belonging to heart sound activity), and their corresponding time domain signals are recovered by performing an inverse Fourier transform.

[0083] The noise-free signals in each frame are then added together while taking into account the original 33% overlap to reconstruct the phonocardiogram signal of the entire PCG signal. Finally, the same bandpass filter with cutoff frequencies of 25Hz and 140Hz is then applied again and the resulting signal is now noise-free.

[0084] Heartbeat segmentation and quality assurance features applied to the denoised signal or the original signal are also presented. For segmentation, the ECG signal is used as a reference to identify the start and end times of each heartbeat in the processed PCG signal. The start of the R wave in each cardiac cycle of the ECG signal is considered as the transition point between the end of one heartbeat and the beginning of the next heartbeat. The ECG signal between two consecutive starting points is then identified as a cardiac cycle, and therefore the corresponding PCG signal is identified as a heartbeat. The short-time periodicity of the cardiac cycle is then exploited to identify the first heart sound (S1) and the second heart sound (S2) within each heartbeat, even in severely affected cases. To this end, the PCG signal is divided into overlapping frames containing two heartbeats, each heartbeat having one overlapping beat between consecutive frames, and the signal envelope of each frame is calculated using a low-pass filter with a cutoff frequency of 10 Hz.

[0085] Then, for each frame, a cross-correlation is performed using a comb function that has a value of zero at all points except t=0 and t=T (where T is the time period of the first of two heartbeats). The position of the pulse in this function is expected to be close to the start of the first S1 peak in the PCG signal frame, and the position of the second S2 peak is T seconds after the first S1 peak. Once the two S1 peak positions are known, the start time and end time of the S1 heart sound are determined by searching back and forward from the peak for the time point corresponding to 60% of the peak height. A similar method is used to identify the second heart sound, however this time a cutoff frequency of 15 Hz is used to calculate the signal envelope. Previous signal analysis experiments revealed that the S2 peak is found within 0.2T and 0.55T seconds of the S1 peak, so the peak in the cross-correlation time series that occurs in this time interval after the first most significant peak informs the position of the S2 peak of the first heartbeat. The start time and end time of the S2 heart sound are determined in a similar manner to that described above.

[0086] Identifying the endpoints of the S1 and S2 heart sounds allows identification of the systolic and diastolic intervals between these heart sounds, such as Fig. 9 As shown. The segmented denoised signal is used to mark the corresponding endpoints in the original PCG signal. Not all heartbeats in the resulting dataset are perfectly segmented, and therefore a subset of high-quality heartbeats is created for applications that do not require the entire dataset of heartbeats. A heartbeat is considered to be a high-quality heartbeat if its signal: (1) both S1 and S2 are successfully identified, (2) there is no signal offset between the systolic and diastolic intervals, and (3) the heartbeat duration is within ±20% of the median duration of the subject.

[0087] Example 2

[0088] To illustrate the feature extraction process of the method, the acquired and processed signals of the initial subject group are used as sources. The feature extraction process utilizes a specific case selection of raw or denoised phonocardiogram signals, which belong to all or exclusively good quality heartbeats depending on the underlying physiology being characterized.

[0089] In this example, the PCG signal processor processes the PCG signal to transform the signal into three types of features including amplitude features, frequency features, and spectral entropy features. These features can then be used to formulate index metrics including peak E velocity, e' velocity, LAVi, E / A ratio, and peak TR velocity.

[0090] like Fig. 10AAs shown in the process of , the amplitude feature in this example is generated by applying a Hilbert transform to a selected PCG signal segment and then applying a low pass filter with a cutoff frequency of 51 Hz, thereby producing a signal envelope. The 60th percentile value of the resulting signal envelope is then calculated and designated as the amplitude feature. The amplitude feature is then used to calculate an index metric for the peak E velocity parameter. The PCG-based index metric calculation is performed on a per heartbeat basis.

[0091] By following Fig. 10B The process of applying a Hamming window to the PCG signal segment and then applying a 64-point discrete Fourier transform to generate a frequency signature. The centroid of the frequency distribution between 16 Hz and 160 Hz is then calculated and designated as the frequency signature. The frequency signature is then used to calculate index metrics for e' velocity and LAVi parameters. The PCG-based index metric calculation is performed on a per heartbeat basis.

[0092] Finally, if Fig. 10C As described in the process shown in , a spectral entropy feature is generated by obtaining a signal distribution probability estimate from a PCG signal segment and then calculating the negative product of the signal probability distribution estimate and its logarithm. The spectral entropy feature is then used to calculate index metrics for E / A ratio and peak TR velocity parameters. The PCG-based index metric calculation is performed on a per heartbeat basis.

[0093] Example 3

[0094] To demonstrate the formulation of the index metric, the final feature value for each subject was calculated by taking the average feature value of the selected heartbeats of that subject. Noise reduction, heartbeat segmentation, feature extraction, and index metric calculation were performed in a fully automatic manner. The extracted features directly characterize physiological phenomena that are otherwise measured by echocardiographic parameters. A summary of the features used for each echocardiographic parameter is shown in Table 2, and the procedure used is generally described in FIG. 11A to FIG. 11E Described in.

[0095] Each subject's feature values ​​are plotted against their echocardiographic parameters, and an index metric is estimated for each subject using a linear fit. The linear fit involves first modeling the relationship between the echocardiographic parameters and the PCG features by calculating the constants (slope and intercept) of the linear regression between the two, and then using these calculated constants to estimate the index echocardiographic metric for any given PCG feature value. The index metric is adjusted by subtracting the intercept of the linear model and dividing by its slope, and any index values ​​outside the physiologically feasible range are truncated accordingly. It can be seen that these predefined linear models can be used to calculate the index metric for any never-seen subject, and new model derivation may not be required for each analysis.

[0096] Through Fig.11A As shown, a denoised PCG signal of the diastolic interval of a high-quality heartbeat is obtained, the ratio of the amplitude eigenvalues ​​of the pulmonary signal and the aortic signal is calculated, the average of the available ratios for each object is calculated, and the calculated average is fitted to a linear model to determine an indicator for peak E velocity.

[0097] The peak E velocity parameter is a measure of the peak early diastolic flow velocity at the tip of the mitral valve leaflet during passive emptying of the left atrium into the left ventricle. The parameter value reflects the pressure gradient between the left atrium and the left ventricle and is affected by any changes in the left ventricular diastolic rate or left atrial pressure. Subjects in the group with high flow velocities exhibit correspondingly high signal amplitudes. Therefore, the ratio of the feature based on the pulmonary amplitude to the aortic amplitude calculated for the diastolic denoised phonocardiogram signal of the good quality beat was selected to characterize this trend.

[0098] Because the direction of diastolic blood flow is away from the aortic auscultation point location and the lung auscultation point location, lower amplitude-based eigenvalues ​​are seen for high Peak E velocity values. Here, the ratio of the lung eigenvalue to the aortic eigenvalue allows this trend to be compared on the left and right sides of the heart. As a result, a larger diastolic amplitude ratio is seen for subjects with larger Peak E velocity values, and thus this ratio is selected to calculate the index metric for the Peak E velocity parameter.

[0099] Through Fig. 11B As shown, the original pulmonary PCG signal of the high-quality heartbeat is obtained, the ratio of the spectral entropy eigenvalues ​​of the early diastolic interval and the late diastolic interval is calculated, the average of the available ratios for each subject is calculated, and the calculated average is fitted to a linear model to determine an indicator for the E / A ratio.

[0100] The E / A ratio parameter is a measure of the ratio of the early to late peak diastolic flow velocity at the tip of the mitral valve leaflet during passive and subsequent active emptying of the left atrium into the left ventricle. The value of this parameter is used to identify the state of left ventricular function: normal, impaired relaxation, moderate diastolic dysfunction (pseudoformal filling), or restrictive left ventricular filling (impaired left ventricular compliance).

[0101] A spectral entropy-based feature was used to identify diastolic phonocardiographic signal segments associated with muscle contraction related to left ventricular filling. Lower spectral entropy values ​​were observed in late diastolic signal segments corresponding to active left atrial contraction when compared to early diastolic signal segments corresponding to passive left atrial emptying. This trend was strongest for raw phonocardiographic signals in good quality heartbeats acquired at the lung auscultation point. Therefore, the ratio of the features based on the spectral entropy of early to late lung diastolic signals was selected to calculate the index metric for the E / A ratio parameter.

[0102] Through Fig. 11CAs shown in , the de-noised aortic PCG signals of all heartbeats are obtained, the frequency characteristic values ​​of the late systolic intervals are calculated, the average of the available ratios for each subject is calculated, and the calculated average is fitted to a linear model to determine an index for e' velocity.

[0103] The e' velocity parameter is a measure of the average early diastolic flow velocity at the mitral annulus during passive emptying of the left atrium into the left ventricle. It can be seen that the value of this parameter is associated with the time constant of left ventricular relaxation. The left ventricular blood flow dynamics responsible for these early diastolic mitral annular deflections are indirectly estimated during systole.

[0104] High frequency vibrations associated with high blood flow exhibit elevated levels of corresponding high frequency signal components, while low frequency vibrations associated with myocardial motion exhibit elevated levels of corresponding low frequency signal components. Subjects with high e' velocity values ​​due to larger mitral annular deflection also exhibit larger muscle motion related low frequency components during systole. This phenomenon is characterized by calculating a frequency-based feature of the de-noised end-systolic phonocardiogram signal in all heartbeats collected at the aortic auscultation point. Therefore, this feature is selected to calculate the index metric for the e' velocity parameter.

[0105] Through Fig.11D As shown, a denoised PCG signal of the diastolic interval of a high-quality heartbeat is obtained, the ratio of the spectral entropy eigenvalues ​​of the pulmonary signal and the aortic signal is calculated, the average of the available ratios for each subject is calculated, and the calculated average is fitted to a linear model to determine an indicator for the peak TR velocity.

[0106] The peak TR velocity parameter is a measure of the peak regurgitant systolic jet velocity at the tricuspid valve during right ventricular systole. The parameter value provides an indirect measure of the pulmonary artery systolic pressure, which can be seen to be directly related to the left atrial pressure. It has been observed that subjects with larger peak TR velocity values, and therefore higher pulmonary artery pressures, exhibit organized cardiac sound patterns in the phonocardiographic signals acquired at the lung auscultation point. These patterns are characterized by calculating the ratio of spectral entropy-based features of the diastolic phonocardiographic signals acquired at the lung auscultation point and the aortic auscultation point.

[0107] For subjects with larger peak TR velocity values, lower spectral entropy values ​​were observed at the lung auscultation point. This trend was strongest for the denoised phonocardiogram signal in good quality heartbeats, so this ratio was chosen to calculate the index metric for the peak TR velocity parameter.

[0108] Through Fig.11E As shown, raw mitral valve PCG signals for all heartbeats are obtained, frequency characteristic values ​​of the early diastolic interval are calculated, the average of the available ratios for each subject is calculated, and the calculated average is fitted to a linear model to determine an index for LAVi.

[0109] The LAVi parameter is a metric of the maximum left atrial volume associated with body surface area. The parameter value reflects the cumulative effect of the increase in left atrial pressure over time. Objects with larger LAVi values ​​and therefore larger left atria exhibit larger low-frequency signal components related to muscle movement during left ventricular filling in early diastole. This trend is characterized by calculating the frequency-based features of the early diastolic interval phonocardiogram signal collected at the mitral valve auscultation point, and this trend is strongest for the original phonocardiogram signal in all heartbeats. Therefore, this feature is selected to calculate the index metric for the LAVi parameter.

[0110] Due to occasional signal quality deficiencies associated with measurements in a noisy cath lab environment, index metrics could not be calculated for all subjects. Peak E velocity index and peak TR velocity index were not available for each of the six subjects, e' velocity was not available for three subjects, and LAVi was not available for one subject.

[0111] Example 4

[0112] The index metric results were compared with existing echocardiographic and standard methods to validate the diagnostic accuracy of the device and method. The clinical value of the index metric was assessed using a diagnostic processor employing a custom diagnostic algorithm based on the algorithm described in the 2016 joint recommendations of the American Society of Echocardiography (ASE) and the European Association of Cardiovascular Imaging (EACVI).

[0113] Fig. 12A The schematic diagram shown represents elements of an algorithm for assessing left ventricular diastolic dysfunction in patients with normal left ventricular ejection fraction (LVEF). The algorithm is customized to use peak E velocity, e' velocity, peak TR velocity, and LAVi to identify subjects with LV diastolic dysfunction in the presence of normal LV ejection fraction values. Fig. 12A The standard parameters used are shown. Fig. 12B The second part of the algorithm was tailored to use the above 4 parameters and the E / A ratio to estimate the mean left atrial pressure (as an indirect measure of LV filling pressure) in subjects with underlying myocardial disease, with reduced ejection fraction values ​​or with normal ejection fraction values. The parameters used were Fig. 12B is shown in the schematic diagram.

[0114] The goal of this example is to compare PCG-based index metrics with echocardiographic parameters for LV diastolic function assessment using this 2016ASE / EACVI algorithm. Use their echocardiographic parameters to obtain actual diastolic dysfunction and left atrial pressure assessment for each subject regardless of their ejection fraction value.

[0115] The index metric identified LV diastolic dysfunction in 29 subjects with 87.5% accuracy and elevated LV filling pressures in 17 subjects with 75% accuracy. These figures are very consistent with those reported in reference studies comparing the diagnostic accuracy of echocardiographic parameters with gold standard invasive catheter pressure measurements. Potential sources of error in the calculation of the index metric are PCG signals that are not recorded simultaneously with the echocardiographic parameters or due to occasional signal quality defects during measurements in the noisy catheterization laboratory environment.

[0116] From the description herein, it will be understood that the present disclosure encompasses a number of embodiments, which include but are not limited to the following:

[0117] A system for calculating an index metric of an echocardiographic parameter, comprising: (a) one or more phonocardiogram (PCG) sensors; (b) one or more electrocardiogram (ECG) sensors; (c) a computer processor; and (d) a non-volatile memory storing instructions executable by the computer processor; (e) wherein the instructions, when executed by the processor, cause the processor to perform steps comprising: (i) receiving PCG acoustic signals from one or more PCG sensors attached to a subject; (ii) simultaneously receiving ECG signals from one or more electrocardiogram (ECG) sensors attached to the subject; (iii) optionally, denoising the received PCG acoustic signals; (iv) processing the PCG signals into one or more of a time characteristic, an amplitude characteristic, a frequency characteristic, or a spectral entropy characteristic for each heartbeat of the subject; and (v) converting the processed characteristics into one or more index metrics of cardiac tissue and valve blood flow parameters using a set of predetermined conversion equations.

[0118] A system according to any preceding or following embodiment, wherein denoising a received PCG acoustic signal comprises: applying a bandpass filter with cutoff frequencies of 25 Hz and 140 Hz to the PCG acoustic signal; estimating the spectral noise spectrum during brief pauses in heart sound activity; and subtracting the estimate from the spectrum of the entire signal to obtain a clean heart sound signal.

[0119] A system according to any of the foregoing or following embodiments, wherein the instructions, when executed by the processor, further perform steps including: using the ECG signal as a reference to identify the start time and end time of each heartbeat in the PCG signal; identifying the first heart sound (S1) and the second heart sound (S2) as well as the diastolic interval and the systolic interval within each identified heartbeat; and evaluating whether the heartbeat meets the conditions of a high-quality heartbeat by: determining whether both S1 and S2 have been successfully identified, determining whether there is no signal offset in the systolic interval and the diastolic interval, and determining whether the heartbeat duration is within ±20% of the median duration of the subject.

[0120] A system according to any of the preceding or following embodiments, wherein the amplitude feature processing comprises: applying a Hilbert transform to the PCG signal using signal processing in the time domain and the frequency domain; applying a low-pass filter with a cutoff frequency of 51 Hz to generate a signal envelope, and calculating the 60th percentile value of the resulting signal envelope of the heartbeat.

[0121] A system according to any preceding or following embodiment, wherein the frequency feature processing includes: isolating a PCG signal corresponding to a heartbeat; applying a Hamming window to a segment of the PCG signal; applying a 64-point discrete Fourier transform, and calculating the centroid of the frequency distribution of the heartbeat between 16 Hz and 160 Hz.

[0122] A system according to any of the preceding or following embodiments, wherein the spectral entropy frequency feature processing includes: isolating a PCG signal corresponding to a heartbeat; obtaining a signal distribution probability estimate from the PCG signal; and calculating the negative product of the signal probability distribution estimate of the heartbeat and its logarithm.

[0123] A system as in any preceding or following embodiment, wherein each transformation equation is generated using linear regression applied to a training data set of subject condition ground truth data and sensor signal data obtained for a population of subjects.

[0124] A system according to any of the preceding or following embodiments, wherein calculating a PCG-based index metric for an echocardiogram-based peak E velocity parameter comprises: obtaining a denoised PCG signal for the diastolic interval; identifying good quality heartbeats; calculating a ratio of amplitude characteristic values ​​of the pulmonary signal and the aortic signal for each heartbeat; calculating an average of the available ratios for all good quality heartbeats of each subject; and fitting the calculated average to a predefined linear model using a predetermined transformation equation.

[0125] A system according to any of the preceding or following embodiments, wherein calculating a PCG-based indicator metric for an echocardiogram-based E / A ratio parameter comprises: obtaining a raw pulse PCG signal for quality heartbeats; calculating the ratio of spectral entropy eigenvalues ​​of the early diastolic interval and the late diastolic interval of each heartbeat; calculating the average of the available ratios for all quality heartbeats of each subject; and fitting the calculated average to a predefined linear model using a predetermined transformation equation.

[0126] A system according to any of the preceding or following embodiments, wherein calculating a PCG-based indicator metric for an echocardiogram-based e' velocity parameter comprises: obtaining a denoised aortic PCG signal for all heartbeats; calculating frequency characteristic values ​​of the late systolic interval for each heartbeat; calculating an average of the available ratios for all heartbeats of each subject; and fitting the calculated average to a predefined linear model using a predetermined transformation equation.

[0127] A system according to any of the preceding or following embodiments, wherein calculating a PCG-based indicator metric for an echocardiogram-based peak TR velocity parameter comprises: obtaining a denoised PCG signal for the diastolic interval of good quality heartbeats; calculating the ratio of spectral entropy eigenvalues ​​of the pulmonary signal and the aortic signal for each heartbeat; calculating the average of the available ratios for all good quality heartbeats of each subject; and fitting the calculated average to a predefined linear model using a predetermined transformation equation.

[0128] A system according to any of the preceding or following embodiments, wherein calculating a PCG-based indicator metric for an echocardiography-based LAVi parameter comprises: obtaining raw mitral valve PCG signals for all heartbeats; calculating frequency characteristic values ​​of the early diastolic interval for each heartbeat; calculating an average of the available ratios for all heartbeats of each subject; and fitting the calculated average to a predefined linear model using a predetermined transformation equation.

[0129] A system according to any preceding or following embodiment, wherein PCG-based indicator metrics for peak E velocity, e' velocity, LAVi, E / A ratio, and peak TR velocity are output to a diagnostic processor configured to assess diastolic function and left atrial pressure based on the PCG-based indicator metrics.

[0130] A method for calculating an index metric for an echocardiographic parameter, the method comprising: (a) receiving a PCG acoustic signal from one or more sensors attached to a subject; (b) simultaneously receiving an electrocardiogram (ECG) signal from one or more sensors attached to the subject; (c) processing the denoised PCG signal into one or more of a time feature, an amplitude feature, a frequency feature, or a spectral entropy feature for each heartbeat of the same subject; (d) based on a set of predetermined conversion equations, converting the extracted features into a plurality of cardiac tissue and valve blood flow parameter analogs (index metrics), the conversion comprising: (i) processing the amplitude features of all heartbeats of the same subject into a time feature for early diastole; (ii) converting the amplitude features of all heartbeats of the same subject into a frequency feature for early diastole; and (iii) converting the amplitude features of all heartbeats of the same subject into a frequency feature for early diastole. (i) processing the frequency features of all heartbeats of the same subject into an index measure for the peak velocity of blood flow through the subject's mitral valve during diastole (peak E velocity); (ii) processing the frequency features of all heartbeats of the same subject into an index measure for the average flow velocity through the subject's mitral valve during early diastole (e' velocity) and the maximum volume of the subject's left atrium (LAVi) associated with the subject's body surface area; and (iii) processing the spectral entropy features of all heartbeats of the same subject into an index measure for the ratio of early peak flow velocity to late peak flow velocity through the subject's mitral valve during diastole (E / A ratio) and the peak velocity of blood regurgitation through the subject's tricuspid valve during systole (peak TR velocity).

[0131] The method according to any of the preceding or following embodiments also includes denoising the received PCG acoustic signal, the denoising comprising: applying a bandpass filter with cutoff frequencies of 25 Hz and 140 Hz to the PCG acoustic signal; estimating the spectral noise spectrum during brief pauses in cardiac sound activity; and subtracting the estimate from the spectrum of the entire signal to obtain a clean cardiac sound signal.

[0132] The method according to any of the foregoing or following embodiments also includes: using the ECG signal as a reference to identify the start time and end time of each heartbeat in the PCG signal; identifying the first heart sound (S1) and the second heart sound (S2) as well as the diastolic interval and the systolic interval within each identified heartbeat; and evaluating whether the heartbeat meets the conditions of a high-quality heartbeat by: determining whether both S1 and S2 have been successfully identified, determining whether there is no signal offset in the systolic interval and the diastolic interval, and determining whether the heartbeat duration is within ±20% of the median duration of the subject.

[0133] A method according to any of the preceding or following embodiments, wherein amplitude feature processing comprises: applying a Hilbert transform to the PCG signal using signal processing in the time domain and the frequency domain; applying a low-pass filter with a cutoff frequency of 51 Hz to generate a signal envelope, and calculating the 60th percentile value of the resulting signal envelope of the heartbeat.

[0134] A method according to any of the preceding or following embodiments, wherein frequency feature processing includes: isolating a PCG signal corresponding to a heartbeat; applying a Hamming window to a segment of the PCG signal; applying a 64-point discrete Fourier transform, and calculating the centroid of the frequency distribution of the heartbeat between 16 Hz and 160 Hz.

[0135] A method according to any of the preceding or following embodiments, wherein spectral entropy frequency feature processing includes: isolating a PCG signal corresponding to a heartbeat; obtaining a signal distribution probability estimate from the PCG signal; and calculating the negative product of the signal probability distribution estimate of the heartbeat and its logarithm.

[0136] A method according to any of the preceding or following embodiments, wherein calculating a PCG-based index metric for an echocardiogram-based peak E velocity parameter comprises: obtaining a denoised PCG signal for the diastolic interval; identifying high-quality heartbeats; calculating the ratio of amplitude characteristic values ​​of the pulmonary signal and the aortic signal for each heartbeat; calculating the average of the available ratios for all high-quality heartbeats of each subject; and fitting the calculated average to a predefined linear model using a predetermined transformation equation.

[0137] A method according to any of the preceding or following embodiments, wherein calculating a PCG-based indicator metric for an echocardiogram-based E / A ratio parameter comprises: obtaining a raw pulse PCG signal for high-quality heartbeats; calculating the ratio of spectral entropy eigenvalues ​​of the early diastolic interval and the late diastolic interval of each heartbeat; calculating the average of the available ratios for all high-quality heartbeats of each subject; and fitting the calculated average to a predefined linear model using a predetermined transformation equation.

[0138] A method according to any of the preceding or following embodiments, wherein calculating a PCG-based index metric for an echocardiography-based e' velocity parameter comprises: obtaining a denoised aortic PCG signal for all heartbeats; calculating frequency characteristic values ​​of the late systolic interval for each heartbeat; calculating an average of the available ratios for all heartbeats of each subject; and fitting the calculated average to a predefined linear model using a predetermined transformation equation.

[0139] A method according to any of the foregoing or following embodiments, wherein calculating a PCG-based indicator metric for an echocardiography-based peak TR velocity parameter comprises: obtaining a denoised PCG signal for the diastolic interval of quality heartbeats; calculating the ratio of spectral entropy eigenvalues ​​of the pulmonary signal and the aortic signal for each heartbeat; calculating the average of the available ratios for all quality heartbeats of each subject; and fitting the calculated average to a predefined linear model using a predetermined transformation equation.

[0140] A method according to any of the preceding or following embodiments, wherein calculating a PCG-based indicator metric for an echocardiography-based LAVi parameter comprises: obtaining raw mitral valve PCG signals for all heartbeats; calculating frequency characteristic values ​​of the early diastolic interval for each heartbeat; calculating an average of the available ratios for all heartbeats of each subject; and fitting the calculated average to a predefined linear model using a predetermined transformation equation.

[0141] As used herein, the term "embodiment" is intended to include, but not be limited to, examples, or other forms of practicing the techniques described herein.

[0142] As used herein, the singular terms "a" and "an" and "the" may include plural referents unless the context clearly indicates otherwise. Unless explicitly stated, reference to an object in the singular is not intended to mean "one and only one" but rather "one or more."

[0143] Word constructions such as "A, B, and / or C" in the present disclosure describe where A, B, or C may be present, or any combination of the items A, B, and C. Word constructions indicating such as "at least one of" which precedes a group of elements listed indicate that at least one of the group of elements is present, including any possible combination of the listed elements, as applicable.

[0144] References in this disclosure to "an embodiment," "at least one embodiment," or similar embodiment terms indicate that a particular feature, structure, or characteristic described in connection with the described embodiment is included in at least one embodiment of the present disclosure. Thus, these various embodiment phrases are not necessarily all referring to the same embodiment, or to a particular embodiment that is different from all other embodiments described. The embodiment phrase should be interpreted to mean that the particular features, structures, or characteristics of a given embodiment may be combined in any suitable manner in one or more embodiments of the disclosed apparatus, system, or method.

[0145] As used herein, the term "group" refers to a collection of one or more objects. Thus, for example, a group of objects may include a single object or multiple objects.

[0146] Relational terms such as first and second, top and bottom, and the like may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions.

[0147] The terms “comprises, comprising, includes, including,” “has, having,” “contains, containing,” or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises, has, or contains a list of elements includes not only those elements but may also include other elements not expressly listed or inherent to such process, method, article, or apparatus. An element preceded by “comprises,” “comprising,” or “including” does not, without more constraints, preclude the presence of additional identical elements in the process, method, article, or apparatus that comprises, has, or contains that element.

[0148] As used herein, the terms "approximately," "approximately," "substantially," "essentially," and "about," or any other form thereof, are used to describe and illustrate small variations. When used in conjunction with an event or circumstance, the term may refer to instances in which the event or circumstance occurs precisely as well as instances in which the event or circumstance occurs close to approximately. When used in conjunction with a numerical value, the term may refer to a range of variation of less than or equal to ±10% of the numerical value, such as less than or equal to ±5%, less than or equal to ±4%, less than or equal to ±3%, less than or equal to ±2%, less than or equal to ±1%, less than or equal to ±0.5%, less than or equal to ±0.1%, or less than or equal to ±0.05%. For example, "substantially" alignment may refer to an angular variation range of less than or equal to ±10°, such as less than or equal to ±5°, less than or equal to ±4°, less than or equal to ±3°, less than or equal to ±2°, less than or equal to ±1°, less than or equal to ±0.5°, less than or equal to ±0.1°, or less than or equal to ±0.05°.

[0149] In addition, amounts, ratios and other numerical values ​​may sometimes be presented in the form of ranges herein. It should be understood that the use of such ranges is for convenience and brevity, and should be flexibly understood to include numerical values ​​that are clearly specified as the limits of the range, and also include all individual numerical values ​​or sub-ranges encompassed within the range, as if each numerical value and sub-range were clearly specified. For example, a ratio within the range of about 1 to about 200 should be understood to include clearly listed limits of about 1 and about 200, but also includes separate ratios, such as about 2, about 3, and about 4, and sub-ranges, such as about 10 to about 50, about 20 to about 100, etc.

[0150] The term "coupled" as used herein is defined as connected, but not necessarily directly, and not necessarily mechanically. A device or structure "configured" in a certain way is configured in at least that way, but may also be configured in ways that are not listed.

[0151] Benefits, advantages, solutions to problems, and any element(s) that may make any benefit, advantage, or solution present or more significant should not be construed as critical, required, or essential features or elements of the techniques described herein or in any or all the claims.

[0152] In addition, in the foregoing disclosure, various features may be combined together in various embodiments in order to make the disclosure smooth. This disclosure method should not be interpreted as reflecting the intention that the claimed embodiments require more features than the features explicitly recited in each claim. The subject matter of the invention may lie in less than all the features of a single disclosed embodiment.

[0153] The Abstract of the Disclosure is provided to allow the reader to quickly ascertain the nature of the technical disclosure. It is understood that the Abstract is not intended to interpret or limit the scope or meaning of the claims.

[0154] It should be understood that after submitting an application, the practice of some jurisdictions may require deleting one or more parts of the disclosure. Therefore, the reader should consult the application submitted for the original content of the disclosure. Any deletion of the disclosure should not be interpreted as giving up, losing or contributing to the public any theme of the application as originally submitted.

[0155] The following claims are hereby incorporated into this disclosure, with each claim standing on its own as a separately claimed subject matter.

[0156] Although the description herein contains many details, these details should not be interpreted as limiting the scope of the present disclosure, but should be interpreted as only providing illustrations of some currently preferred embodiments. Therefore, it will be understood that the scope of the present disclosure fully encompasses other embodiments that may become obvious to those skilled in the art.

[0157] All structural and functional equivalents of the elements of the disclosed embodiments known to those of ordinary skill in the art are expressly incorporated herein by reference and are intended to be covered by the claims. In addition, none of the elements, components, or method steps in the present disclosure are dedicated to the public, regardless of whether the elements, components, or method steps are clearly stated in the claims. Unless the element is clearly stated using the phrase "means for...", the claim elements herein should not be interpreted as "means plus function" elements. Unless the element is clearly stated using the phrase "step for...", no claim element herein is interpreted as "step plus function" element.

[0158] Table 1

[0159] Summary of parameters available in echocardiogram reports

[0160]

[0161]

[0162] Table 2 Summary of characteristics used to calculate indices for echocardiographic parameters

[0163]

Claims

1. A system for calculating an index metric of an echocardiographic parameter, comprising: (a) one or more phonocardiogram (PCG) sensors; (b) one or more electrocardiogram (ECG) sensors; (c) computer processors; and (d) a non-transitory memory storing instructions executable by the computer processor; (e) wherein the instructions, when executed by the processor, cause the processor to perform steps comprising: (i) receiving PCG acoustic signals from one or more PCG sensors attached to the subject; (ii) simultaneously receiving electrocardiogram (ECG) signals from one or more ECG sensors attached to the subject; (iii) optionally, denoising the received PCG acoustic signal; (iv) processing the PCG signal into one or more of a time feature, an amplitude feature, a frequency feature or a spectral entropy feature of each heartbeat of the subject; and (v) converting the processed features into one or more index measures of cardiac tissue and valve blood flow parameters using a set of predetermined conversion equations.

2. The system according to claim 1, wherein: De-noise the received PCG acoustic signal, including: applying a bandpass filter with cutoff frequencies of 25 Hz and 140 Hz to the PCG acoustic signal; estimating the spectral noise spectrum during brief pauses in heart sound activity; and The estimate is subtracted from the spectrum of the entire signal to obtain a clean heart sound signal.

3. The system according to claim 1, wherein: The instructions, when executed by the processor, further perform steps including: Using the ECG signal as a reference to identify the start and end times of each heartbeat in the PCG signal; identifying the first heart sound (S1) and the second heart sound (S2) as well as the diastolic interval and the systolic interval within each identified heartbeat; as well as A heartbeat is assessed for quality by determining whether both S1 and S2 have been successfully identified, whether there is no signal offset between systolic and diastolic intervals, and whether the heartbeat duration is within ±20% of the subject's median duration.

4. The system according to claim 1, wherein: The amplitude feature processing includes: applying a Hilbert transform to the PCG signal using signal processing in the time domain and the frequency domain; Apply a low pass filter with a cutoff frequency of 51 Hz, thus generating the signal envelope, and The 60th percentile value of the resulting signal envelope of the heartbeat is calculated.

5. The system according to claim 1, wherein: The frequency characteristic processing includes: Isolate the PCG signal corresponding to one heartbeat; applying a Hamming window to a segment of the PCG signal; Apply a 64-point discrete Fourier transform, and The centroid of the frequency distribution of the heartbeat between 16 Hz and 160 Hz is calculated.

6. The system according to claim 1, wherein: The spectral entropy frequency feature processing includes: Isolate the PCG signal corresponding to one heartbeat; Obtaining a signal distribution probability estimate from the PCG signal; and The negative product of the signal probability distribution estimate of the heartbeat and its logarithm is calculated.

7. The system according to claim 1, wherein: Each transformation equation is generated using linear regression applied to a training data set of object condition ground truth data and sensor signal data obtained for a population of objects.

8. The system according to claim 1, wherein: Calculate PCG-based metrics for echocardiographic peak E velocity parameters including: Obtain a denoised PCG signal for the diastolic interval; Identify quality heartbeats; Calculate the ratio of the amplitude characteristic values ​​of the pulmonary signal and the aortic signal for each heartbeat; Calculate the average of the available ratios of all good quality heartbeats for each object; and The calculated means were fit to a predefined linear model using a predetermined transformation equation.

9. The system according to claim 1, wherein: Calculate PCG-based metrics for echocardiographic-based E / A ratio parameters including: Get the original pulse PCG signal for high-quality heartbeat; Calculate the ratio of the spectral entropy eigenvalues ​​of the early diastolic interval and the late diastolic interval of each heartbeat; Calculate the average of the available ratios of all good quality heartbeats for each object; and The calculated means were fit to a predefined linear model using a predetermined transformation equation.

10. The system according to claim 1, wherein: Calculate PCG-based metrics for echocardiographic e' velocity parameters including: Obtain denoised aortic PCG signals for all heartbeats; Calculate the frequency characteristic value of the late systolic interval of each heartbeat; Calculate the average of the available ratios for all heartbeats of each subject; and The calculated means were fit to a predefined linear model using a predetermined transformation equation.

11. The system according to claim 1, wherein: Calculate PCG-based metrics for echocardiographic peak TR velocity parameters including: Obtain a denoised PCG signal for the diastolic interval of a high-quality heartbeat; Calculate the ratio of the spectral entropy eigenvalues ​​of the pulmonary signal and the aortic signal for each heartbeat; Calculate the average of the available ratios of all good quality heartbeats for each object; and The calculated means were fit to a predefined linear model using a predetermined transformation equation.

12. The system of claim 1, wherein: PCG-based index metrics for echocardiographic LAVi parameters were calculated, including: Obtain raw mitral valve PCG signals for all heartbeats; Calculate the frequency characteristic value of the early diastolic interval of each heartbeat; Calculate the average of the available ratios for all heartbeats of each subject; and The calculated means were fit to a predefined linear model using a predetermined transformation equation.

13. The system of claim 1, wherein: The PCG-based index metrics for peak E velocity, e' velocity, LAVi, E / A ratio, and peak TR velocity are output to a diagnostic processor configured to assess diastolic function and left atrial pressure based on the PCG-based index metrics.

14. A method for calculating an index metric for an echocardiographic parameter, the method comprising: (a) receiving PCG acoustic signals from one or more sensors attached to a subject; (b) simultaneously receiving electrocardiogram (ECG) signals from one or more sensors attached to the subject; (c) processing the denoised PCG signal into one or more of a time feature, an amplitude feature, a frequency feature or a spectral entropy feature of each heartbeat of the same subject; (d) converting the extracted features into a plurality of cardiac tissue and valve blood flow parameter analogs (index metrics) based on a set of predetermined conversion equations, the conversion comprising; (i) processing the amplitude feature of all heartbeats of the same subject into an index measure for the peak velocity of blood flow through the subject's mitral valve during early diastole (peak E velocity); (ii) processing the frequency features of all heartbeats of the same subject into index measures for the mean flow velocity through the subject's mitral valve during early diastole (e' velocity) and the maximum volume of the subject's left atrium (LAVi) correlated to the subject's body surface area; as well as (iii) processing the spectral entropy features of all heartbeats of the same subject into index metrics for the ratio of early peak flow velocity to late peak flow velocity through the subject's mitral valve during diastole (E / A ratio) and the peak velocity of blood regurgitation through the subject's tricuspid valve during systole (peak TR velocity).

15. The method according to claim 14, further comprising denoising the received PCG acoustic signal, wherein the denoising comprises: applying a bandpass filter with cutoff frequencies of 25 Hz and 140 Hz to the PCG acoustic signal; estimating the spectral noise spectrum during brief pauses in heart sound activity; as well as The estimate is subtracted from the spectrum of the entire signal to obtain a clean heart sound signal.

16. The method according to claim 14, further comprising: Using the ECG signal as a reference to identify the start and end times of each heartbeat in the PCG signal; identifying the first heart sound (S1) and the second heart sound (S2) as well as the diastolic interval and the systolic interval within each identified heartbeat; as well as A heartbeat is assessed for quality by determining whether both S1 and S2 have been successfully identified, whether there is no signal offset between systolic and diastolic intervals, and whether the heartbeat duration is within ±20% of the subject's median duration.

17. The method according to claim 14, wherein: The amplitude feature processing includes: applying a Hilbert transform to the PCG signal using signal processing in the time domain and the frequency domain; Apply a low pass filter with a cutoff frequency of 51 Hz, thus generating the signal envelope, and The 60th percentile value of the resulting signal envelope of the heartbeat is calculated.

18. The method according to claim 14, wherein: The frequency characteristic processing includes: Isolate the PCG signal corresponding to one heartbeat; applying a Hamming window to a segment of the PCG signal; Apply a 64-point discrete Fourier transform, and The centroid of the frequency distribution of the heartbeat between 16 Hz and 160 Hz is calculated.

19. The method according to claim 14, wherein: The spectral entropy frequency feature processing includes: Isolate the PCG signal corresponding to one heartbeat; Obtaining a signal distribution probability estimate from the PCG signal; and The negative product of the signal probability distribution estimate of the heartbeat and its logarithm is calculated.

20. The method according to claim 14, wherein: Calculate PCG-based metrics for echocardiographic peak E velocity parameters including: Obtain a denoised PCG signal for the diastolic interval; Identify quality heartbeats; Calculate the ratio of the amplitude characteristic values ​​of the pulmonary signal and the aortic signal for each heartbeat; Calculate the average of the available ratios of all good quality heartbeats for each object; and The calculated means were fit to a predefined linear model using a predetermined transformation equation.

21. The method according to claim 14, wherein: Calculate PCG-based metrics for echocardiographic-based E / A ratio parameters including: Get the original pulse PCG signal for high-quality heartbeat; Calculate the ratio of the spectral entropy eigenvalues ​​of the early diastolic interval and the late diastolic interval of each heartbeat; Calculate the average of the available ratios of all good quality heartbeats for each object; and The calculated means were fit to a predefined linear model using a predetermined transformation equation.

22. The method according to claim 14, wherein: Calculate PCG-based metrics for echocardiographic e' velocity parameters including: Obtain denoised aortic PCG signals for all heartbeats; Calculate the frequency characteristic value of the late systolic interval of each heartbeat; Calculate the average of the available ratios for all heartbeats of each subject; and The calculated means were fit to a predefined linear model using a predetermined transformation equation.

23. The method according to claim 14, wherein: Calculate PCG-based metrics for echocardiographic peak TR velocity parameters including: Obtain a denoised PCG signal for the diastolic interval of a high-quality heartbeat; Calculate the ratio of the spectral entropy eigenvalues ​​of the pulmonary signal and the aortic signal for each heartbeat; Calculate the average of the available ratios of all good quality heartbeats for each object; and The calculated means were fit to a predefined linear model using a predetermined transformation equation.

24. The method according to claim 14, wherein: PCG-based index metrics for echocardiographic LAVi parameters were calculated, including: Obtain raw mitral valve PCG signals for all heartbeats; Calculate the frequency characteristic value of the early diastolic interval of each heartbeat; Calculate the average of the available ratios for all heartbeats of each subject; and The calculated means were fit to a predefined linear model using a predetermined transformation equation.

Images