Template-based analysis and classification of cardiovascular waveforms

Abstract

In various embodiments, a first classification assigned to a periodic component of an electrical waveform representing electrical activity in a heart of a patient can be identified (302). A corresponding periodic component of a hemodynamic waveform representing hemodynamic activity in a cardiovascular system of the patient can be analyzed (306, 318, 328). The corresponding periodic component is causally related to the periodic component of the electrical waveform. Based on the analysis, responsive to determining that a previously assigned classification also applies to the corresponding periodic component based on the analysis, the previously assigned classification can be assigned (312, 324) to the corresponding periodic component of the hemodynamic waveform. In a database (130) of hemodynamic templates, a hemodynamic template associated with the previously assigned classification can be updated (314) to include one or more features of the corresponding periodic component of the hemodynamic waveform.

Description

TECHNICAL FIELD

[0001] The present disclosure relates generally to healthcare. More particularly, but not exclusively, the various methods and apparatus disclosed herein relate to template-based analysis and classification of cardiovascular waveforms. BACKGROUND

[0002] Cardiovascular waveforms representing arterial blood pressure ("ABP"), pulmonary artery pressure ("PAP"), central venous pressure ("CVP"), and plethysmography can be affected by physiological changes such as ectopic beats and / or arrhythmias. In particular, electrical activity in a patient's heart (measured, for example, by electrocardiogram ("ECG")) can affect the shape of these waveforms. Physicians can then analyze the shape of these waveforms to identify abnormalities in the patient's heartbeat that require further investigation. However, the signals obtained by ECG and other devices are not perfect, and noise and artifacts can be introduced that also affect the waveform shape. These shapes are sometimes mistaken for abnormalities even though they were introduced by machinery or other factors other than the patient's physiology. Reference is made to WO 2011 / 080189 Al ("Monitoring a property of the cardiovascular system of a subject") herein. SUMMARY

[0003] Algorithms for filtering and assessing waveform quality sometimes incorrectly classify atypical electrical activity as noise or artifact (e.g., due to interference and / or patient movement) when, in fact, the activity can prove to be a cardiovascular abnormality. For example, some algorithms compare electrical waveforms and / or hemodynamic waveforms to templates of normal waveforms and abnormal waveforms. However, many abnormal waveforms can not match existing templates and, as a result, can lead to noisy classifications when, in fact, a true cardiovascular abnormality exists. Therefore, it would be beneficial to provide a method and system of analyzing and classifying cardiovascular waveforms in a manner that better identifies cardiovascular abnormalities and confirms / rejects classifications made using existing algorithms.

[0004] The present disclosure relates to inventive methods and apparatus for template-based analysis and classification of cardiovascular waveforms. For example, the present disclosure describes techniques for classifying (or annotating) periodic components (e.g., heartbeats) of hemodynamic waveforms based on templates such as normal waveforms and abnormal waveforms, and / or for confirming and / or reclassifying periodic components of electrical waveforms representing electrical activity in a patient's heart. Furthermore, templates associated with normal waveforms and various types of abnormal waveforms can be updated to include features of newly identified normal waveforms and abnormal waveforms.

[0005] Generally, in one aspect, a method can include identifying a previously assigned classification associated with a periodic component of an electrical waveform, wherein the electrical waveform is representative of electrical activity in a heart of a patient; analyzing a corresponding periodic component of a hemodynamic waveform representative of hemodynamic activity in a cardiovascular system of the patient, wherein the corresponding periodic component is causally related to the periodic component of the electrical waveform; assigning the previously assigned classification to the corresponding periodic component of the hemodynamic waveform in response to determining, based on the analysis, that the previously assigned classification is also applicable to the corresponding periodic component; and updating, in a database of hemodynamic templates, a hemodynamic template associated with the previously assigned classification to include one or more features of the corresponding periodic component of the hemodynamic waveform.

[0006] In various embodiments, the method can further include receiving electrophysiology data associated with the patient, wherein the electrophysiology data includes the electrical waveform and one or more previously assigned classifications associated with one or more periodic components of the electrical waveform; and receiving hemodynamic data associated with the patient, wherein the hemodynamic data includes the hemodynamic waveform.

[0007] In various embodiments, the electrophysiology data is received from one or more electrodes of an electrocardiogram. In various embodiments, the hemodynamic data can include a signal indicative of arterial blood pressure of the patient. In various embodiments, the hemodynamic data can include a signal indicative of pulmonary blood pressure of the patient. In various embodiments, the hemodynamic data can include a signal indicative of central venous pressure. In various embodiments, the hemodynamic data can include a signal from a plethysmograph.

[0008] In various embodiments, the method can further include identifying an unclassified periodic component of the same hemodynamic waveform or a different hemodynamic waveform associated with a different patient; matching the unclassified periodic component to a template of the database of hemodynamic templates; and assigning a classification associated with the matched template to the unclassified periodic component of the hemodynamic waveform. In various embodiments, the method can further include updating the matched template to include one or more features of the now classified periodic component of the hemodynamic waveform.

[0009] In various embodiments, the previously assigned classification can include an abnormality classification. The assigning can include assigning the abnormality classification to the corresponding periodic component of the hemodynamic waveform in response to determining, based on the analysis, that a difference between the corresponding periodic component of the hemodynamic waveform and a previous periodic component of the hemodynamic waveform satisfies a threshold.

[0010] In various embodiments, the method can further include identifying an artifact classification of another periodic component of the electrical waveform that is deemed to be an artifact; analyzing another corresponding periodic component of the hemodynamic waveform that is causally related to the other periodic component of the electrical waveform; responsive to determining, based on the analysis, that a difference between the other corresponding periodic component of the hemodynamic waveform and another prior periodic component of the hemodynamic waveform satisfies a threshold, assigning an abnormality classification to the other corresponding periodic component of the hemodynamic waveform; and reclassifying the other periodic component of the electrical waveform having the abnormality classification.

[0011] In various embodiments, the previously assigned classification can include a normal classification, and the assigning can include assigning the normal classification to the corresponding periodic component of the hemodynamic waveform responsive to determining, based on the analysis, that the corresponding periodic component satisfies a signal quality indicator (SQI).

[0012] In various embodiments, the analysis can include matching the corresponding periodic component to templates of the database of hemodynamic templates. In various versions, the updating can include merging the corresponding periodic component with a jointly matched periodic component of a hemodynamic template stored.

[0013] It will be recognized that all combinations of the foregoing concepts and additional concepts (if any) discussed in greater detail below (provided such concepts are not mutually inconsistent) are contemplated as being part of the subject matter disclosed herein. In particular, all combinations of claimable subject matter appearing at the end of this disclosure are contemplated as being part of the subject matter disclosed herein. It will be appreciated that specified BRIEF DESCRIPTION OF DRAWINGS

[0014] In the drawings, like reference numerals refer to like parts throughout the various views. Also, the drawings are not necessarily to scale, emphasis instead being placed upon illustrating the principles of the disclosure.

[0015] Figure 1 An example environment in which the disclosed technology can be implemented is shown in accordance with various implementations.

[0016] Figure 2 An example waveform that can be analyzed in accordance with various embodiments is depicted.

[0017] Figure 3 And Figure 4 An example method in accordance with various embodiments is depicted.

[0018] Figure 5 Example "normal" and "abnormal" templates are depicted in accordance with various embodiments.

[0019] Figure 6 Components of an example computer system are depicted. DETAILED DESCRIPTION

[0020] Cardiovascular waveforms such as those representing ABP, PAP, and / or CVP, as well as signals from a plethysmograph ("PLETH"), can be affected by physiological changes such as ectopic beats and / or arrhythmias. In particular, electrical activity in a patient's heart (e.g., as measured by an ECG) can affect the shape of various hemodynamic waveforms. Existing algorithms for filtering and assessing waveform quality sometimes incorrectly classify atypical electrical activity as noise or artifact, when in fact the activity can prove to be a cardiovascular abnormality. For example, some algorithms compare electrical waveforms and / or hemodynamic waveforms to templates of normal waveforms and abnormal waveforms. However, many abnormal waveforms can not match existing templates and thus can result in a noisy classification, when in fact there is a true cardiovascular abnormality. Accordingly, there is a need in the art for analyzing and classifying cardiovascular waveforms in a manner that better identifies cardiovascular abnormalities and validates / rejects classifications made using existing algorithms. More generally, it is recognized and appreciated that it would be beneficial to continuously learn new patterns associated with normal cardiovascular waveforms and abnormal cardiovascular waveforms. In view of the foregoing, various embodiments and implementations of the present disclosure are directed to classifying or annotating periodic components (e.g., heartbeats) of cardiovascular waveforms based on various signals such as templates of normal waveforms and abnormal waveforms, and / or for validating and / or reclassifying periodic components of electrical waveforms representing electrical activity in a patient's heart. Further, a database of templates of normal waveforms and abnormal waveforms can be updated to include newly identified features of normal waveforms and abnormal waveforms.

[0021] REFERENCE Figure 1 An example environment 100 in which the disclosed technology can be implemented is depicted. A patient 102 can be connected to various medical devices to monitor electrical activity in the patient's heart and / or hemodynamic activity in the patient's vascular system. For example, the patient 102 can be monitored by one or more medical devices such as an ECG 104, an ABP 106, a PAP 108, a CVP 110, and / or a PLETH 112. These devices are merely examples and are not meant to be limiting. Other electrical signals and / or hemodynamic signals can be obtained and monitored.

[0022] In the case of ECG 104, in some implementations, one or more electrodes (not shown) can be attached to patient 102 to detect electrical activity in the patient's heart to produce signal 114. In the case of ABP 106, an instrument such as a blood pressure cuff can be used to measure the ABP of patient 102, for example at regular intervals or continuously, to produce signal 116. PAP 108 can take various forms, for example transthoracic echocardiography ("TTE") and / or right heart catheterization, and can produce signal 118. In various embodiments, CVP 110 can be measured by connecting a central venous catheter (not depicted) to an infusion pump to produce signal 120. In the case of PLETH 112, signal 122 can be produced by a plethysmograph, which is an instrument used to measure volume changes within an organ or the entire body of patient 102.

[0023] In various embodiments, signals 114-122 can be provided to a cardiovascular analysis system 124. Operations performed by cardiovascular analysis system 124 can be distributed across multiple computer systems. For example, cardiovascular analysis system 124 can be implemented as a computer program running on one or more computers in one or more locations, which are coupled to each other by a network (not shown). Cardiovascular analysis system 124 can include various engines and / or modules implemented using any combination of hardware, software, or both. In various embodiments, these modules and / or engines can include an initial annotation engine 125, an analysis engine 126, and / or a template engine 128. In some implementations, one or more of engines 125, 126, and 128 can be omitted. In some implementations, all or aspects of one or more of engines 125, 126, and 128 can be combined. In some implementations, one or more of engines 125, 126, and 128 can be implemented in components separate from cardiovascular analysis system 124.

[0024] The initial annotation engine 125 can be configured to analyze the electrical signals 114 produced by the ECG 104 individually to obtain various characteristics. In some embodiments, the initial annotation engine 125 can annotate various “periodic components” of the electrical waveforms as “normal,” “abnormal,” or “artifact” (also referred to as “noise”). As used herein, a “periodic component” of a waveform can refer to any component that generally recurs in the waveform (although from occurrence to occurrence it need not be identical). In some embodiments, the periodic component can be a recurring “peak” of the waveform. For example, in an electrical waveform, each “peak” can represent a surge of electrical activity. In other embodiments, the periodic waveform can refer to a “trough” or another generally recurring visual feature of the waveform, such as a peak / trough pair (even though the height or phase of the peak and the depth of the trough can differ from occurrence to occurrence). The modifier “periodic” as used herein refers to the component being something that generally recurs periodically, or at least should recur periodically. However, it should be understood that in various scenarios, particularly for patients in critical condition, the waveforms produced by the patient can or can not exhibit periodic components, and / or for a particular patient, the periodic components can not actually recur periodically (which can or can not be a cause or result of the patient’s illness). In some embodiments, the annotations can be added by the initial annotation engine 125 using various algorithms, such as the segment and arrhythmia analysis ST / AR ECG and / or DXL algorithms, although other algorithms can be used to add annotations as well.

[0025] In various implementations, the analysis engine 126 can be configured to analyze the annotated waveforms representing electrical activity in the patient’s heart and / or hemodynamic activity in the patient’s vascular system to determine whether one or more of the annotations or “classifications” are correct, or incorrect, and so on. In a hemodynamic waveform, each “peak” (or a combination of peaks and troughs) can represent a heartbeat. In various implementations, the analysis engine 126 can be configured to identify classifications / annotations assigned to periodic components of the electrical waveforms representing electrical activity in the patient’s 102 heart, e.g., by the initial annotation engine 125. For example, the analysis engine 126 can identify a particular peak that was classified or annotated as “normal,” “abnormal,” or “artifact.” The analysis engine 126 can then analyze a corresponding periodic component of a hemodynamic waveform representing hemodynamic activity in the patient’s vascular system (e.g., representing one or more of the signals 116-122). The corresponding periodic component can be causally related to the periodic component of the electrical waveform.

[0026] Based on this analysis, the analysis engine 126 can classify (or reclassify, if already classified elsewhere) the corresponding periodic component of the hemodynamic waveform with the same classification or a different classification as assigned to the periodic component of the electrical waveform. For example, if the analysis of the hemodynamic waveform confirms the "normal" classification assigned to the periodic component of the electrical waveform by the initial annotation engine 125, then the analysis engine 126 can classify the corresponding periodic component of the hemodynamic waveform as "normal." In another aspect, assume that the analysis negates the "artifact" classification assigned to the periodic component of the electrical waveform by the initial annotation engine 125. For example, assume that the analysis engine 126 determines, using the techniques described herein, that the corresponding periodic component of the hemodynamic waveform actually indicates an abnormality. In this case, the analysis engine 126 can classify the corresponding periodic component of the hemodynamic waveform as "abnormal," and can reclassify the periodic component of the electrical waveform from "artifact" to "abnormal."

[0027] In various implementations, once the analysis engine 126 classifies and / or reclassifies the periodic components of the electrical waveform and / or the hemodynamic waveform, the analysis engine 126 can effectively "update" its knowledge and / or the knowledge of the cardiovascular analysis system 124. For example, in some embodiments, the template engine 128 can maintain one or more databases of hemodynamic templates. In various embodiments, each hemodynamic template can include one or more "features" of one or more periodic components of one or more hemodynamic waveforms having the same classification as the periodic component of the hemodynamic waveform. In Figure 1 In particular embodiments, the template engine 128 maintains a database 130 of hemodynamic templates that includes templates associated with a "normal" classification, and templates having various types of "abnormal" classifications associated with different diseases (e.g., ectopic beats, arrhythmias, etc.). However, this is not meant to be limiting. In other embodiments, templates classified as normal, abnormal, or even "artifact" can be stored in separate databases.

[0028] After the analysis engine 126 performs the above analysis, in various embodiments it can request the template engine 128 to update one or more of the template database 130 with one or more features of the corresponding periodic component of the hemodynamic waveform. Various features of the periodic component of the waveform or underlying hemodynamic activity can be used to update the templates, ranging from summary data representative of the entire waveform to specific features such as maximum / minimum amplitude, distance from previous peak, rate of increase / decrease, oscillations, blood density, average velocity, blood viscosity, etc. In some embodiments, features of multiple waveforms (e.g., electrical and hemodynamic) can be added to the templates. For example, a "pulse transit time" can be the time difference between a peak of an ECG waveform and a trough of an ABP or PLETH waveform. In some embodiments, the pulse transit time can be included in the template as an annotation to the periodic component.

[0029] When the analysis engine 126 is analyzing a new or otherwise unclassified hemodynamic waveform, it can compare the periodic components of the new hemodynamic waveform to the templates in the database 130. For example, a template can match a particular periodic component under analysis when one or more features of the periodic component under analysis are sufficiently similar to corresponding features of the template.

[0030] For example, in some embodiments, a feature vector can be extracted from the periodic component under analysis. Various machine learning techniques can be employed to compute a similarity measure between the template feature vector associated with a template and the feature vector of the periodic component under analysis. In some embodiments, a mathematical model such as a neural network and / or logistic regression can be trained. Various learning algorithms for training such models can be used, e.g., batch or stochastic gradient descent and / or application of normal equations. If the computed similarity satisfies one or more thresholds, then a match can exist. Thus, the periodic component under analysis can be classified as being the same as the template feature vector. In some implementations, one or more features from the feature vector extracted from the periodic component can be incorporated into the template of the template database 130, e.g., to aid in future comparisons.

[0031] In other embodiments, a hemodynamic template can include one or more periodic components that, for example, represent a consolidation of multiple previously classified periodic components. The periodic component under analysis can be correlated with one or more periodic components of the template to determine whether a match exists (e.g., whether the difference between the two satisfies or does not satisfy one or more thresholds). In some embodiments, the periodic component under analysis can be subtracted from the template periodic component(s), and the difference can be compared to one or more thresholds. In other embodiments, techniques such as a Fast Fourier Transform ("FFT") or a covariance shift can be employed to correlate the periodic component under analysis with the template periodic component.

[0032] Referring now to Figure 2 , two example waveforms are depicted. A first waveform 240, depicted in solid line, represents electrical activity of the patient, such as the signal 114 produced by the ECG 104. A second waveform 242, depicted in dashed line, represents hemodynamic activity in the patient's vasculature, such as the signal 116 produced by the ABP 106. In this example, five periodic components in the first waveform 240 (in the form of peaks, from left to right) are annotated / classified with the letter "N" to indicate normal electrical activity in the patient's heart, such as determined by the initial annotation engine 125. A sixth electrical peak is annotated with the letter "A" to indicate an "artifact" perceived by the initial annotation engine 125, such as due to a sudden increase in amplitude and / or a sudden decrease in phase. Following the artifact, there are four periodic components of the first waveform that are classified as normal.

[0033] The periodic components of the second waveform 242 (in the form of peaks in this example) are causally related to the periodic components of the first waveform 240. Specifically, an electrical pulse (represented by a peak) in the first waveform 240 is followed by a concomitant peak of the second waveform 242. This is because an electrical pulse in the patient's heart triggers the heart to pump blood, and a peak blood pressure occurs at some time interval after the peak electrical pulse. In this example, there are five similar peaks in the second waveform 242, with a relatively normal and uniform amplitude, followed by a higher peak, then a lower peak, followed by another four peaks that are near the normal peak amplitude.

[0034] As indicated at the symbol delta, there is a difference in the ABP measured between the sixth and seventh peaks of the second waveform 242, which appears to be a result of the electrical periodic component being classified as an "artifact." In fact, the second waveform 242 presenting this delta as apparently a result of the so-called "artifact" in the first waveform 240 indicates that the artifact is not an artifact at all. Rather, the corresponding anomaly in the second waveform 242 represented by delta indicates a drop in systolic ABP caused by a premature beat, which results in insufficient blood being pumped by the patient's heart into the vasculature. In other words, a physiologically abnormal electrical pulse has caused a physiologically abnormal hemodynamic pulse. Using the techniques described herein, the periodic component of the second waveform 242 indicated at delta can be classified as "abnormal." In some embodiments, the periodic component of the first waveform 240 (i.e., the electrical pulse) that was initially classified as an "artifact" can be reclassified as "abnormal." The features of one or both periodic components can then be included in templates added to the various databases.

[0035] Figure 3Example methods 300, according to various embodiments, can be performed by one or more components of a cardiovascular analysis system 124. In some embodiments, once a patient is connected to one or more health monitoring devices (e.g., Figure 1 Method 300 can be initially executed by generating templates of various types (e.g., normal, various types of abnormalities) associated with the patient (sections 104-112). In some embodiments, method 300 can be stopped when a sufficient number of templates have been generated. For example, once at least a threshold number of templates with a specific classification (e.g., normal, various types of abnormalities) exist, method 300 can be stopped unless a medical professional determines that the templates are defective. In the event of a defective template, the medical professional can activate a “relearning” routine that restarts the generation of templates for the patient. Although described in a specific order... Figure 3 This describes the operations, but it does not imply limitations. In various embodiments, one or more operations may be reordered, omitted, or added.

[0036] At box 302, the classification of periodic components assigned to an electrical waveform (e.g., signal 114 generated by ECG 104) can be identified. For example, analysis engine 126 can identify annotations assigned to the periodic components in signal 114 by initial annotation engine 125. Alternatively, analysis engine 126 can analyze previously recorded signal waveforms with annotations, which may or may not have been added by initial annotation engine 125. At box 304, corresponding periodic components from hemodynamic waveforms of the same patient can be identified. As described above, corresponding periodic components can be causally correlated with the periodic components of the electrical waveform and can be identified as corresponding in many cases because it follows the electrical periodic components through a specific predictable time interval or range of time intervals.

[0037] At box 306, a signal quality index, or "SQI," can be determined for the periodic components of the hemodynamic waveform (or for components associated with the entire hemodynamic waveform or portions thereof comprising multiple periodic components). The SQI can be calculated in various ways. In some embodiments, for example, the SQI associated with an ABP signal can estimate blood pressure at various stages of the cardiac cycle and assign scores based on the physiological reliability of the signal. Some SQIs can measure morphological normality. Others can measure signal degradation due to noise (or artifacts). If the signal represented by the waveform is relatively noisy, the waveform may receive a relatively low SQI. Other SQIs not specifically mentioned herein may also be applied.

[0038] If the SQI determined at block 306 fails to satisfy a particular threshold value T (which can be set to various values, for example), the method 300 can proceed to block 308, where the corresponding periodic component of the hemodynamic waveform originally identified at block 304 can be rejected. However, if the SQI satisfies the threshold value T, the method can proceed to block 310.

[0039] At block 310, it can be determined whether the classification assigned to the periodic component of the electrical waveform (identified at block 302) is "normal" or its particular equivalent variation. If the answer is yes, the method 300 can proceed to block 312. At block 312, the corresponding component of the hemodynamic waveform identified at block 304 can be classified as "normal" (or its semantically equivalent variation).

[0040] Then, at block 314, one or more features extracted from the now-classified corresponding periodic component can be incorporated into the "normal" hemodynamic template stored in the hemodynamic template database 130. In some embodiments, incorporating the one or more features can include merging the now-classified corresponding periodic component with the already-merged periodic components stored in the joint "normal" hemodynamic template. For example, the durations of the two periodic components can be normalized, and then an average of the two periodic components can be determined.

[0041] In some embodiments, a weighted average of the now-classified periodic components having n periodic components represented by the template can be determined. For example, the periodic components of the "normal" hemodynamic template can include an average of the n periodic components added to the template in the past, each of which can be weighted or unweighted according to how long they have been added. The further back in time (e.g., determined according to iterations or pure time) a particular periodic component is incorporated into the template, the smaller weight it can be assigned. In other words, an exponential weight decreasing further in the past can be assigned to the n periodic components merged into the template. In some embodiments, a low-pass filter such as a single-pole filter can be employed to determine the periodic components stored in the joint template.

[0042] Returning to block 310, if the periodic component of the electrical signal is not annotated as "normal," the method 300 can proceed to block 316. At block 316, it can be determined whether the classification assigned to the periodic component of the electrical waveform (identified at block 302) is "artifact," "noise," or some particular equivalent variation thereof. If the answer is yes, the method 300 can proceed to block 318. At block 318, it can be determined whether the corresponding periodic component of the hemodynamic waveform satisfies one or more criteria. For example, in some embodiments, it can be determined whether the aforementioned delta (i.e., the difference between the amplitudes between two adjacent peaks) satisfies a particular threshold. In some embodiments, an equation such as the following can be employed:

[0043]

[0044] where SBP(n) is the systolic pressure at the periodic component under examination, and SBP(n-1) is the systolic pressure at the immediately preceding periodic component. While 5% is used as the threshold in this example, it should be understood that this is not meant to be limiting. Various other thresholds can be selected in various circumstances, depending on the health of the patient, the environment of the patient (e.g., the activities they are engaged in), and so forth.

[0045] If the criteria of block 318 are not satisfied, the method 300 can proceed to block 320, where the corresponding periodic component can be rejected. This can indicate, for example, that the periodic component of the electrical waveform was actually correctly classified as an artifact. On the other hand, if the criteria of block 318 are satisfied, the method 300 can proceed to block 322. At block 322, the periodic component of the electrical waveform can be reclassified, in this example from "artifact" to "abnormal." At block 324, the corresponding periodic component of the hemodynamic waveform can likewise be classified as "abnormal." And as has been discussed, at block 314, one or more features in the now-classified corresponding periodic component of the hemodynamic waveform can be incorporated into the "abnormal" template stored in the database (e.g., 130).

[0046] Returning to block 316, if the periodic component of the electrical signal is not annotated as "noise" or "artifact," the method 300 can proceed to block 326. At block 326, a determination can be made as to whether the classification assigned to the periodic component of the electrical waveform (identified at block 302) is "abnormal" or its particular equivalent variation. If the answer is no, the method 300 can return to block 302. However, if the answer at block 326 is yes, the method 300 can proceed to block 328. At block 328, a similar determination can be made as was made at block 318. For example, the same or similar equation used above can be used again. If the threshold at block 328 is not met, the method 300 can proceed to block 320, where the periodic component of the hemodynamic waveform can be rejected. If the threshold at block 328 is met, the method 300 can proceed to block 324, and then to block 314, as described above.

[0047] In Figure 3 In the above, a single SQI determination was made at block 306, but this is not meant to be limiting. In various embodiments, different SQI determinations can be made depending on the classification assigned to the periodic component of the electrical waveform. For example, a first SQI can be determined if it is classified as normal. A second SQI can be determined if it is classified as abnormal. And so on. While the particular equation described above can be used to determine abnormalities in hemodynamic waveforms, this is not meant to suggest that it can only be used independently, or that alternative equations cannot be used.

[0048] Figure 4 An example method 400 is depicted for classifying unclassified periodic waveform components using templates developed using methods such as Figure 3 In the above, a single SQI determination was made at block 306, but this is not meant to be limiting. In various embodiments, different SQI determinations can be made depending on the classification assigned to the periodic component of the electrical waveform. For example, a first SQI can be determined if it is classified as normal. A second SQI can be determined if it is classified as abnormal. And so on. While the particular equation described above can be used to determine abnormalities in hemodynamic waveforms, this is not meant to suggest that it can only be used independently, or that alternative equations cannot be used. Figure 4 The operations of the method 400 are depicted in a particular order, but this is not meant to be limiting. In various embodiments, one or more operations can be reordered, omitted, or added.

[0049] At block 402, the hemodynamic waveform under consideration can be segmented, e.g., into segments each including a periodic component such as a peak and / or a trough. At block 404, an SQI can be determined for each segment (or the entire waveform), similar to block 306 of the method 300. Figure 3 If the SQI fails to meet the threshold τ, the method 400 can proceed to block 406, where the segment (i.e., the portion of the waveform containing the periodic component) is classified as "noise" or "artifact." However, if the threshold is met, the method 400 can proceed to block 408.

[0050] At block 408, each segment / periodic component can be associated with a template in a template database (e.g., 130). For example, in some embodiments, the segment / periodic component can be correlated with a normal template and a plurality of abnormal templates, each associated with a different type of abnormal classification (e.g., premature beat, atrial fibrillation, etc.). In some embodiments, and as previously described, the segment / periodic component under analysis can be subtracted from the template periodic component(s), and the difference can be compared to one or more thresholds. In other embodiments, techniques such as FFT or covariance shifting can be employed to correlate the periodic component under analysis with one or more template periodic components.

[0051] If, at block 410, the correlation R between the segment / periodic component and the one or more templates fails to satisfy another threshold (e.g., τ2in Equation 2), the method 400 can proceed to block 412, where the periodic component can be classified as indeterminate and / or rejected. However, if the threshold τ2is satisfied, the method 400 can proceed to block 414. In various embodiments, τ2may be an adjustable threshold for evaluating the correlation between a template and a segment / periodic component. In some embodiments, τ2may be set to a value between 0.5 and 1, such as 0.8. In other embodiments, τ2may be learned over time from patient data (e.g., using machine learning techniques or various heuristics). For example, as more templates are added to the template database, closer matches can be obtained, and thus τ2may vary over time. Figure 4

[0052] At block 414, one or more features of the segment (i.e., the periodic component) can be incorporated into a template stored in the template database (e.g., 130). Examples of how the periodic component can be incorporated into a template are described above with respect to block 314. At block 416, the segment / periodic component can be classified accordingly, e.g., for use by one or more downstream components and / or algorithms.

[0053] ​Waveforms whose periodic components are classified / reclassified / annotated using the techniques described herein can be used for various downstream purposes. For example, periodic components (e.g., peaks) of hemodynamic waveforms that are classified as abnormal using the techniques described herein can be used to detect and alert for hemodynamic deterioration. Additionally, other cardiovascular measurements can be made more accurate when viewed in conjunction with annotated periodic components. For example, heart rate oscillations can be more accurately identified based at least in part on classifications / annotations determined using the disclosed techniques. In some embodiments, the techniques described herein can be used with a sleep monitoring system that simultaneously monitors ECG and PLETH signals. Additionally, hemodynamic waveforms annotated using the techniques herein can be used for applications such as clinical decision support algorithms, e.g., to reduce false positive rates by appropriately classifying periodic components as abnormal. As another example, the techniques described herein can be used to retroactively correct ECG artifacts in ECG signals as abnormal.

[0054] Figure 5 Non-limiting examples of how multiple accumulated periodic components (referred to as “beats” in the images) that have been classified as “abnormal” (top) and “premature” (bottom, i.e., a particular type of abnormal) can be merged into a single accumulated periodic component represented by the thick black line are depicted. If an unclassified periodic component is sufficiently similar to Figure 5 those thick black line periodic components depicted in FIGS. 6A-6D, it can be classified accordingly. For example, if a similarity score between a feature vector extracted from the unclassified periodic component and a feature vector extracted from a merged accumulated periodic component shown in FIGS. 6A-6D satisfies one or more thresholds, the unclassified periodic component can be classified as the same. Figure 5

[0055] Figure 6 is a block diagram of an example computer system 610. Computer system 610 typically includes at least one processor 614 which communicates with a number of peripheral devices via bus subsystem 612. As used herein, the term “processor” will be understood to encompass a variety of devices capable of performing the various functions attributed to the CDS system described herein, such as microprocessors, FPGAs, ASICs, other similar devices, and combinations thereof. These peripheral devices can include a data retention subsystem 624 (including, for example, a memory subsystem 625 and a file storage subsystem 626), user interface output devices 620, user interface input devices 622, and a network interface subsystem 616. The input and output devices allow a user to interact with computer system 610. Network interface subsystem 616 provides an interface to an external network and is coupled to corresponding interface devices in other computer systems.

[0056] ​User interface input devices 622 can include a keyboard, pointing devices such as a mouse, trackball, touchpad, or graphics tablet, a scanner, a touchscreen incorporated into the display, audio input devices such as voice recognition systems, microphones, and / or other types of input devices. In general, use of the term "input device" is intended to include all possible types of devices and ways to input information into computer system 610 or onto a communication network.

[0057] User interface output devices 620 can include a display subsystem, a printer, a fax machine, or non-visual displays such as audio output devices. The display subsystem can include a cathode ray tube (CRT), a flat-panel device such as a liquid crystal display (LCD), a projection device, or some other mechanism for creating a visible image. The display subsystem can also provide non-visual display such as via audio output devices. In general, use of the term "output device" is intended to include all possible types of devices and ways to output information from computer system 610 to the user or to another machine or computer system.

[0058] Data retention system 624 stores programming and data structures that provide the functionality of some or all of the modules described herein. For example, data retention system 624 can include logic that performs selected aspects of method 300 or 400 and / or implements one or more components of cardiovascular analysis system 124.

[0059] The software modules described herein are generally executed by processor 614, alone or in combination with other processors. Memory 625 used in the storage subsystem can include a number of memories including a main random access memory (RAM) 630 for storage of instructions and data during program execution and a read only memory (ROM) 632 for storage of fixed instructions, and other types of memories such as flash memory, a second level cache, which can be integrated onto the same chip as the processor 614 or which can be discrete. File storage subsystem 626 can provide persistent (nonvolatile) storage of program and data files, and can include a hard disk drive, a floppy disk drive along with associated removable media, a CD-ROM drive, an optical drive, or removable media cartridges. The modules implementing the functionality of certain implementations can be stored by file storage subsystem 626 in data retention system 624, or in other machines accessible by the processor(s) 614. As used herein, the term "non-transitory computer readable medium" will be understood to include volatile memory (such as DRAM and SRAM), non-volatile memory (such as flash memory, magnetic storage, and optical storage), but not transitory signals per se.

[0060] The bus subsystem 612 provides a mechanism for letting the various components and subsystems of the computer system 610 communicate with each other as intended. Although the bus subsystem 612 is illustrated as a single bus, alternative implementations of the bus subsystem can use multiple buses.

[0061] The computer system 610 can be of various types, including a workstation, server, computing cluster, blade server, server farm, or any other data processing system or computing device. In some embodiments, the computer system 610 can be implemented within a cloud computing environment. Due to the ever-changing nature of computers and networks, Figure 6 The description of the computer system 610 depicted in FIG. 6 is intended only as an example. Many other configurations of the computer system 610 having more or fewer components than the computer system 610 depicted in FIG. 6 are possible. Figure 6 The description of the computer system 610 depicted in FIG. 6 is intended only as an example. Many other configurations of the computer system 610 having more or fewer components than the computer system 610 depicted in FIG. 6 are possible.

[0062] While several embodiments have been described and illustrated herein, a variety of other means and / or structures for performing the functions and / or obtaining the results and / or one or more of the advantages described herein can be utilized, and each of such variations and / or modifications is deemed to be within the scope of the embodiments described herein. More generally, those skilled in the art will readily appreciate that all parameters, dimensions, materials, and configurations described herein are meant to be exemplary and that the actual parameters, dimensions, materials, and / or configurations will depend upon the specific application or applications for which the teachings is / are used. Those skilled in the art will recognize, or be able to ascertain using no more than routine experimentation, many equivalents to the specific embodiments described herein. It is, therefore, to be understood that the foregoing embodiments are presented by way of example only and that other embodiments may be developed without departing from the scope of the disclosure. The inventive embodiments lie in the novel methods, systems, articles, kits, and / or compositions described herein.

[0063] All definitions, as defined and used herein, should be understood to control over dictionary definitions, definitions in documents incorporated by reference, and / or ordinary meanings of the defined terms.

[0064] The words “a” and “an,” as used herein, should be understood to mean “at least one” unless explicitly indicated to the contrary.

[0065] The phrase “and / or,” as used herein in the specification and in claims, should be understood to mean “either or both of” when applied to a list of two or more items, and “at least one of” or “one or more of” when applied to the elements of a “consisting” or “consisting essentially of” group. As used herein in the specification and in the claims, “or” should be understood to have the same meaning as “and / or” as defined above. For example, when separating items in a list, “or” or “and / or” shall be interpreted as being inclusive, i.e., the inclusion of at least one, but also including more than one, of a number or list of elements, and, optionally, additional unlisted items. Only terms clearly expressed as “only one of” or “exactly one of,” or, when used in the claims, “consisting of,” will refer to the inclusion of exactly one element of a number or list of elements. In general, the term “or” as used herein shall only be interpreted as indicating exclusive alternatives (i.e. “one or the other but not both”) when preceded by terms of exclusivity, such as “either,” “one of,” “only one of,” or “exactly one of.” “Consisting essentially of,” when used in the claims, shall have its ordinary meaning as used in the patent law.

[0066] As used herein in the specification and in the claims, “or” should be understood to have the same meaning as “and / or” as defined above. For example, when separating items in a list, “or” or “and / or” shall be interpreted as being inclusive, i.e., the inclusion of at least one, but also including more than one, of a number or list of elements, and, optionally, additional unlisted items. Only terms clearly expressed as “only one of” or “exactly one of,” or, when used in the claims, “consisting of,” will refer to the inclusion of exactly one element of a number or list of elements. In general, the term “or” as used herein shall only be interpreted as indicating exclusive alternatives (i.e. “one or the other but not both”) when preceded by terms of exclusivity, such as “either,” “one of,” “only one of,” or “exactly one of.” “Consisting essentially of,” when used in the claims, shall have its ordinary meaning as used in the patent law.

[0067] As used in the specification and claims, the phrase “at least one,” in reference to a list of one or more elements, should be understood to mean at least one element selected from among the list of elements, but not necessarily including at least one of each and every element specifically listed within the list of elements and not excluding any combinations of elements in the list of elements. This definition also allows that the

[0068] It should also be understood that, unless clearly indicated to the contrary, in any methods claimed herein that include more than one step or act, the order of the steps or acts of the method is not necessarily limited to the order in which the steps or acts are recited.

[0069] In the claims, as well as in the specification above, all transitional phrases such as “comprising,” “including,” “carrying,” “having,” “containing,” “involving,” “holding,” “composed of,” and the like are to be understood to be open-ended, i.e., to mean including but not limited to. Only the transitional phrases “consisting of’ and “consisting essentially of’ shall be closed or semi-closed transitional phrases, as set forth in the United States Patent Office Manual of Patent Examining Procedures, Section 2111.03. It should be understood that the use of certain specific expressions in the claims, such as “consisting of’ and “consisting essentially of’ do not limit the scope of the claims under the Patent Cooperation Treaty (“PCT”) Rule 6.2(b).

Claims

1. A computer-implemented method comprising: identifying (302), by one or more processors, a periodic component of an electrical waveform and a previously assigned classification associated with the periodic component of the electrical waveform, wherein the electrical waveform is representative of electrical activity in a heart of a patient; analyzing (306, 318, 328), by one or more of the processors, a corresponding periodic component of a hemodynamic waveform representative of hemodynamic activity in a cardiovascular system of the patient, wherein the corresponding periodic component is causally related to the periodic component of the electrical waveform; classifying (312, 324), by one or more of the processors, the corresponding periodic component of the hemodynamic waveform with the previously assigned classification in response to determining, based on the analysis, that the previously assigned classification also applies to the corresponding periodic component; updating (314), by one or more of the processors, a hemodynamic template associated with the previously assigned classification in a database (130) of hemodynamic templates to include one or more features of the corresponding periodic component of the hemodynamic waveform; identifying, by one or more of the processors, an artifact classification assigned to another periodic component of the electrical waveform that is deemed to be an artifact; analyzing, by one or more of the processors, another corresponding periodic component of the hemodynamic waveform that is causally related to the other periodic component of the electrical waveform; classifying, by one or more of the processors, the other corresponding periodic component of the hemodynamic waveform with an abnormal classification in response to determining, based on the analysis, that a difference between the other corresponding periodic component of the hemodynamic waveform and another previous periodic component of the hemodynamic waveform satisfies a threshold; and reclassifying, by one or more of the processors, the other periodic component of the electrical waveform with the abnormal classification.

2. The computer-implemented method of claim 1, further comprising: receiving, by one or more of the processors, electrophysiology data associated with the patient, wherein the electrophysiology data includes the electrical waveform and one or more previously assigned classifications associated with one or more periodic components of the electrical waveform; and receiving, by one or more of the processors, hemodynamic data associated with the patient, wherein the hemodynamic data includes the hemodynamic waveform.

3. The computer-implemented method of claim 2, wherein, The electrophysiology data is received from one or more electrodes of an electrocardiogram (104).

4. The computer-implemented method of claim 2, wherein, The hemodynamic data includes signals (116) indicative of at least one of an arterial blood pressure of the patient, a pulmonary blood pressure of the patient, and a central venous pressure.

5. The computer-implemented method of claim 2, wherein, The hemodynamic data includes signals (122) from a plethysmograph (112).

6. The computer-implemented method of claim 1, further comprising: identifying (402), by one or more of the processors, an unclassified periodic component of a same hemodynamic waveform or a different hemodynamic waveform associated with a different patient; matching (408), by one or more of the processors, the unclassified periodic component to a template of the database of hemodynamic templates; and classifying (416), by one or more of the processors, the unclassified periodic component of the hemodynamic waveform with a classification associated with the matched template.

7. The computer-implemented method of claim 6, further comprising: updating (414), by one or more of the processors, the matched template to include one or more features of the now-classified periodic component of the hemodynamic waveform.

8. The computer-implemented method of claim 1, wherein, The previously-assigned classification comprises one of a normal classification, an artifact classification, and an abnormal classification.

9. The computer-implemented method of claim 1, wherein, The previously-assigned classification comprises an abnormal classification, and wherein the assigning comprises assigning the abnormal classification to the corresponding periodic component of the hemodynamic waveform in response to determining, based on the analysis, that a difference between the corresponding periodic component of the hemodynamic waveform and a previous periodic component of the hemodynamic waveform satisfies a threshold.

10. The computer-implemented method of claim 1, wherein, The previously-assigned classification comprises a normal classification, and wherein the assigning comprises assigning the normal classification to the corresponding periodic component of the hemodynamic waveform in response to determining, based on the analysis, that the corresponding periodic component satisfies a signal quality indicator (SQI).

11. The computer-implemented method of claim 1, wherein, The analysis comprises matching (408) the corresponding periodic component to a template of the database of hemodynamic templates.

12. The computer-implemented method of claim 11, wherein, The updating comprises merging the corresponding periodic component with a jointly-matched periodic component stored by the hemodynamic template.

13. A system comprising: one or more processors (614); and memory (624, 625) operatively coupled with the one or more processors, wherein the memory stores a database (130) of hemodynamic templates, and wherein the memory further stores instructions that, in response to execution of the instructions by the one or more processors, cause the one or more processors to: identify (302) a periodic component of an electrical waveform and a previously-assigned classification associated with the periodic component of the electrical waveform, wherein the electrical waveform is representative of electrical activity in a heart of a patient; analyze (306, 318, 328) a corresponding periodic component of a hemodynamic waveform representative of hemodynamic activity in a cardiovascular system of the patient, wherein the corresponding periodic component is causally related to the periodic component of the electrical waveform; classify (312, 324), in response to determining, based on the analysis, that the previously-assigned classification also applies to the corresponding periodic component, the corresponding periodic component of the hemodynamic waveform with the previously-assigned classification; updating (314), in the database, a hemodynamic template associated with the previously assigned classification to include one or more features of the corresponding periodic component of the hemodynamic waveform; identifying, by one or more of the processors, a false artifact classification assigned to another periodic component of the electrical waveform that is considered to be an artifact; analyzing, by one or more of the processors, another corresponding periodic component of the hemodynamic waveform that is causally related to the another periodic component of the electrical waveform; classifying, by one or more of the processors, the another corresponding periodic component of the hemodynamic waveform with an abnormal classification in response to determining, based on the analyzing, that a difference between the another corresponding periodic component of the hemodynamic waveform and another previous periodic component of the hemodynamic waveform satisfies a threshold; and reclassifying, by one or more of the processors, the another periodic component of the electrical waveform with the abnormal classification.

14. At least one non-transitory computer-readable medium comprising instructions that, in response to execution of the instructions by one or more processors, cause the one or more processors to perform the following operations: identifying (302) a previously assigned classification assigned to a periodic component of the electrical waveform, wherein, the electrical waveform is representative of electrical activity in a heart of a patient; analyzing (306, 318, 328) a corresponding periodic component of a hemodynamic waveform representative of hemodynamic activity in a cardiovascular system of the patient, wherein the corresponding periodic component is causally related to the periodic component of the electrical waveform; classifying (312, 324) the corresponding periodic component of the hemodynamic waveform with the previously assigned classification in response to determining, based on the analyzing, that the previously assigned classification is also applicable to the corresponding periodic component; and updating (314), in a database (130) of hemodynamic templates, a hemodynamic template associated with the previously assigned classification to include one or more features of the corresponding periodic component of the hemodynamic waveform; identifying a false artifact classification assigned to another periodic component of the electrical waveform that is considered to be an artifact; analyzing another corresponding periodic component of the hemodynamic waveform that is causally related to the another periodic component of the electrical waveform; classifying the another corresponding periodic component of the hemodynamic waveform with an abnormal classification in response to determining, based on the analyzing, that a difference between the another corresponding periodic component of the hemodynamic waveform and another previous periodic component of the hemodynamic waveform satisfies a threshold; and reclassifying the another periodic component of the electrical waveform with the abnormal classification.

Images