Systems and methods for screening and predicting sepsis
By extracting the hemodynamic data characteristics using the waveform data generated by the sensor and inputting the prediction calculation model to generate a sepsis screening score or probability score, the problem of difficulty in the existing technology is solved, and efficient diagnosis in the emergency environment is achieved.
Patent Information
- Application Number
- CN202380065575.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Priority Date
- 2022-07-15
- Filing Date
- 2023-07-12
- Publication Date
- 2025-05-16
AI Technical Summary
It is difficult to diagnose sepsis quickly and accurately in the prior art, especially in emergency situations. Traditional SIRS standards and SOFA scores sometimes fail to provide a timely basis for diagnosis.
By using the waveform data generated by the sensor, hemodynamic data characteristics, such as heart rate, kurtosis of the pressure distribution, and sample entropy, etc., and input these characteristics into the prediction calculation model to generate a screening score or probability score for sepsis.
It realizes rapid screening and prediction of the risk of sepsis without relying on SIRS standards, which improves the accuracy and timeliness of diagnosis, which is of great significance especially in the emergency environment.
Smart Images
Figure CN120019446A_ABST
Abstract
Description
[0001] CROSS-REFERENCE TO RELATED APPLICATIONS
[0002] This application claims priority to U.S. Provisional Patent Application No. 63 / 389,577, filed on July 15, 2022, entitled “Systems and Methods for Screening and Predicting Sepsis,” the disclosure of which is incorporated herein by reference. Technical Field
[0003] The present disclosure relates generally to systems and methods for screening and predicting sepsis, and more particularly to systems and methods for screening and predicting sepsis using hemodynamic data. Background Art
[0004] Sepsis is a physical condition in which the body overreacts to a pathogenic infection. Sepsis can cause tissue damage, organ failure, and death. Rapid diagnosis and treatment are needed to prevent serious harm.
[0005] When patients present with a set of signs and symptoms associated with a septic response, they are diagnosed with sepsis. Infection alone is not sufficient to diagnose sepsis, as most pathogenic infections do not lead to sepsis, and pathogenic infections are not clearly associated with the onset of sepsis. One way to diagnose sepsis is to assess for the systemic inflammatory response syndrome (SIRS). A positive diagnosis occurs when at least two of the following criteria are met: (1) fever or hypothermia; (2) increased heart rate (tachycardia); (3) increased respiratory rate (tachypnea); and (4) low or high white blood cell count (leukocytosis or leukopenia) or a high ratio of band cells (bandemia).
[0006] Sepsis progresses to severe sepsis when signs of organ dysfunction develop. Organ dysfunction can be assessed using the Sequential Organ Failure Assessment (SOFA). SOFA assesses lung breathing, blood coagulation, liver function, brain function, cardiovascular function, and kidney function, with higher SOFA scores indicating more severe organ dysfunction. A high SOFA score associated with a sepsis diagnosis indicates an urgent need to treat inflammation and infection to mitigate organ damage and prevent death. Summary of the invention
[0007] Systems and methods for assessing sepsis may include utilizing a sensor to generate waveform data corresponding to arterial blood pressure or a signal proportional to or derived from arterial blood pressure. A set of hemodynamic data features may be extracted from the waveform data. The extracted set of hemodynamic data features and / or other clinical information (including but not limited to patient demographics, vital signs, and laboratory test results) may be used in a computational model to screen for sepsis or predict the probability that a patient is experiencing sepsis.
[0008] In some embodiments, a computational method is used to screen for sepsis. The method includes receiving waveform data from a sensor applied to a patient, the waveform data corresponding to arterial blood pressure or a signal proportional to or derived from arterial blood pressure. The method includes extracting a set of hemodynamic data features from the waveform data. The method includes inputting the extracted set of hemodynamic data features into a predictive computational model to generate a screening score for sepsis. The predictive computational model has been trained to screen for sepsis using the extracted set of hemodynamic data features.
[0009] In some embodiments, the set of hemodynamic data features extracted include heart rate, kurtosis of the pressure distribution, and sample entropy of the time to reach systolic MAP.
[0010] In some embodiments, the set of hemodynamic data features extracted include heart rate, arterial tone factor, sample entropy of attenuation area, dynamic arterial elasticity, and approximate entropy of systolic time.
[0011] In some embodiments, a predictive computational model utilizes an equation to generate a screening score for sepsis.
[0012] In some embodiments, the equation is:
[0013]
[0014] where hr is the heart rate, kurt is the kurtosis of the pressure distribution, and sampEn is the sample entropy of the time to reach systolic MAP.
[0015] In some embodiments, the equation is:
[0016]
[0017] where hr is the heart rate, avgK is the arterial tension factor, decAreaSampEn is the sample entropy of the decay area, dynEa is the dynamic arterial elasticity measure, and tSysApEn is the approximate entropy of the systolic period.
[0018] In some embodiments, the screening score for sepsis indicates a risk of developing sepsis. The method further comprises evaluating the patient further for complications of sepsis.
[0019] In some embodiments, the sepsis screening score indicates a risk of developing sepsis. The method further comprises monitoring the patient for sepsis complications over a specified period of time.
[0020] In some embodiments, a computational model is used to predict the probability of a patient experiencing sepsis. The method includes receiving waveform data from a sensor applied to the patient, the waveform data corresponding to, proportional to, or derived from arterial blood pressure. The method includes extracting a set of hemodynamic data features from the waveform data. The method includes inputting the extracted set of hemodynamic data features into a predictive computational model to generate a probability score for sepsis. The predictive computational model has been trained to predict sepsis using the extracted set of hemodynamic data features.
[0021] In some embodiments, the extracted set of hemodynamic features includes diastolic blood pressure, heart rate, entropy of the beat-to-beat interval, time from reaching systolic MAP to dicrotic notch, and entropy of the standard deviation of the decay phase.
[0022] In some embodiments, the extracted set of hemodynamic features includes heart rate, approximate entropy of blood pressure waveform, sample entropy of systolic region, approximate entropy of systolic time, and approximate entropy of systolic decay time.
[0023] In some embodiments, the equation is:
[0024]
[0025] where Dia is diastolic pressure, hr is heart rate, enIBI is the entropy of the interbeat interval, timeMAP is the time from when systolic MAP is reached to the dicrotic notch, and enDecay is the entropy of the standard deviation of the decay phase.
[0026] In some embodiments, the equation is:
[0027]
[0028] where hr is the heart rate, ApEnV is the approximate entropy of the blood pressure waveform, areaSampEn is the sample entropy of the systolic area, tSysApEn is the approximate entropy of the time of systole, and tDecApEn is the approximate entropy of the time of systolic decay.
[0029] In some embodiments, the probability score for sepsis indicates that the patient has sepsis. The method further comprises further evaluating the patient for complications of sepsis to confirm the probability score.
[0030] In some embodiments, the probability score for sepsis indicates that the patient has sepsis. The method further comprises administering a treatment to the patient to treat sepsis.
[0031] In some embodiments, the method further comprises sensing arterial blood pressure using a sensor.
[0032] In some embodiments, the sensor is an intra-arterial catheter and a disposable pressure transducer, a pressurized cuff and a light sensor, or an applanation tonometer.
[0033] In some embodiments, the extracted set of hemodynamic features includes at least one of the following: heart rate, respiratory rate, cardiac output, stroke volume, stroke volume variation, vascular tone, contractility, afterload, systemic vascular resistance, systolic pressure, diastolic pressure, mean arterial pressure (MAP), kurtosis of pressure distribution, left ventricular ejection time, time from reaching systolic MAP to dicrotic notch, sample entropy of time when reaching systolic MAP, entropy of cardiac interval, entropy of standard deviation of decay phase, or
[0034] In some embodiments, the method further comprises inputting patient clinical information into the model.
[0035] In some embodiments, the patient clinical information includes at least one of patient demographics, patient vital signs, and patient laboratory results.
[0036] In some embodiments, a predictive computational model is a regression-based model, a classification-based model, or an ensemble model.
[0037] In some embodiments, a patient monitoring system is used to screen for sepsis via captured waveform data. The patient monitoring system includes a sensor and a computing processing system operably connected to the sensor. The computing processing system includes a processing system and a memory system, the memory system including one or more applications configured to direct the processor system to receive waveform data corresponding to arterial blood pressure or a signal proportional to or derived from arterial blood pressure from a sensor applied to a patient; extract a set of hemodynamic data features from the waveform data; and input the extracted set of hemodynamic data features into a predictive computing model to generate a screening score for sepsis. The predictive computing model has been trained to screen for sepsis using the extracted set of hemodynamic data features.
[0038] In some embodiments, the set of hemodynamic data features extracted include heart rate, kurtosis of the pressure distribution, and sample entropy of the time to reach systolic MAP.
[0039] In some embodiments, the extracted set of hemodynamic data features includes heart rate, arterial tension factor, sample entropy of attenuation area, dynamic arterial elasticity, and approximate entropy of systolic time.
[0040] In some embodiments, the predictive computational model utilizes an equation to generate a screening score for sepsis.
[0041] In some embodiments, the equation is:
[0042]
[0043] where hr is the heart rate, kurt is the kurtosis of the pressure distribution, and sampEn is the sample entropy of the time to reach systolic MAP.
[0044] In some embodiments, the equation is:
[0045]
[0046] where hr is the heart rate, avgK is the arterial tension factor, decAreaSampEn is the sample entropy of the decay area, dynEa is the dynamic arterial elasticity measure, and tSysApEn is the approximate entropy of the systolic period.
[0047] In some embodiments, the one or more applications are further configured to direct the processor system to display a sepsis screening score on a monitor operably connected to the computing processing system.
[0048] In some embodiments, the screening score for sepsis indicates a risk of developing sepsis. The one or more applications are further configured to direct the processor system to provide an alert indicating the risk upon determining that the screening score for sepsis indicates a risk of developing sepsis.
[0049] In some embodiments, a patient monitoring system is used to predict whether a patient is experiencing sepsis via captured arterial pressure. The patient monitoring system includes a sensor and a computing processing system operably connected to the sensor. The computing processing system includes a processor system and a memory system, the memory system including one or more applications configured to direct the processor system to receive waveform data corresponding to arterial blood pressure or a signal proportional to or derived from arterial blood pressure from a sensor applied to the patient; extract a set of hemodynamic data features from the waveform data; and input the extracted set of hemodynamic data into a predictive computing model to generate a probability score for sepsis. The predictive computing model has been trained to predict sepsis using the extracted set of hemodynamic data features.
[0050] In some embodiments, the extracted set of hemodynamic features includes diastolic pressure, heart rate, entropy of the beat-to-beat interval, time from reaching systolic MAP to dicrotic notch, and entropy of the standard deviation of the decay phase.
[0051] In some embodiments, the extracted set of hemodynamic features includes heart rate, approximate entropy of blood pressure waveform, sample entropy of systolic region, approximate entropy of systolic time, and approximate entropy of systolic decay time.
[0052] In some embodiments, a predictive computational model utilizes an equation to generate a probability score for sepsis.
[0053] In some embodiments, the equation is:
[0054]
[0055] where Dia is diastolic pressure, hr is heart rate, enIBI is the entropy of the interbeat interval, timeMAP is the time from when systolic MAP is reached to the dicrotic notch, and enDecay is the entropy of the standard deviation of the decay phase.
[0056] In some embodiments, the equation is:
[0057]
[0058] where hr is the heart rate, ApEnV is the approximate entropy of the blood pressure waveform, areaSampEn is the sample entropy of the systolic area, tSysApEn is the approximate entropy of the time of systole, and tDecApEn is the approximate entropy of the time of systolic decay.
[0059] In some embodiments, the one or more applications are further configured to direct the processor system to display the probability score for sepsis on a monitor operably connected to the computing processing system.
[0060] In some embodiments, the probability score for sepsis indicates a risk of developing sepsis. The one or more applications are further configured to direct the processor system to provide an alert indicating that the patient is experiencing sepsis upon determining that the probability score for sepsis indicates that the patient is experiencing sepsis.
[0061] In some embodiments, the sensor is an intra-arterial catheter and a disposable pressure transducer, a pressurized cuff and a light sensor, or an applanation tonometer.
[0062] In some embodiments, the extracted set of hemodynamic features includes at least one of the following: heart rate, respiratory rate, cardiac output, stroke volume, stroke volume variation, vascular tone, contractility, afterload, systemic vascular resistance, systolic pressure, diastolic pressure, mean arterial pressure (MAP), kurtosis of pressure distribution, left ventricular ejection time, time from reaching systolic MAP to dicrotic notch, sample entropy of time when reaching systolic MAP, entropy of cardiac interval, entropy of standard deviation of decay phase, or
[0063] In some embodiments, the one or more applications are further configured to direct the processor system to input patient clinical information into the model.
[0064] In some embodiments, the patient clinical information includes at least one of patient demographics, patient vital signs, and patient laboratory results.
[0065] In some embodiments, a predictive computational model is a regression-based model, a classification-based model, or an ensemble model. BRIEF DESCRIPTION OF THE DRAWINGS
[0066] The description and claims will be more fully understood with reference to the following drawings and data graphs, which are presented as examples of the present disclosure and should not be construed as complete statements of the scope of the present disclosure.
[0067] Figure 1 An exemplary method for screening for early signs of sepsis or predicting the probability of sepsis based on waveform data is provided.
[0068] Figure 2 Provides a conceptual illustration of a computational processing system for screening for sepsis.
[0069] Figure 3 A conceptual illustration of a computational processing system for predicting the probability that an individual will experience sepsis is provided. DETAILED DESCRIPTION
[0070] The present disclosure details systems and methods for assessing sepsis using hemodynamic data derived from a continuous blood pressure sensor. Hemodynamic data features can be derived from a blood pressure waveform and used to screen for sepsis and / or predict the probability of sepsis. Thus, these systems and methods can screen for early identification of sepsis in a patient, or can predict the probability that a patient has sepsis. In some embodiments, the hemodynamic data features are used in a trained computational model to screen for or predict the probability of sepsis. In some embodiments, the hemodynamic data features are used in an equation to calculate a score for screening for sepsis or for predicting the probability of sepsis.
[0071] Novel systems and methods are provided for screening for sepsis and / or predicting the probability of sepsis using hemodynamic data. Thus, sepsis can be initially screened and / or diagnosed without analyzing SIRS criteria, which may be helpful in situations where analysis of SIRS criteria is not readily available (e.g., in an emergency room). In some cases, patients are screened for potential risk of developing sepsis, and when a high risk is indicated, the patient is monitored and / or further evaluated for sepsis. In some cases, a patient is predicted to have sepsis, and a confirmatory analysis and / or treatment of sepsis is subsequently performed.
[0072] Figure 1 A method for screening for sepsis or predicting the probability of sepsis is provided, which method can be implemented as a computational process. Method 100 measures (101) waveform data corresponding to arterial blood pressure or a signal proportional to or derived from arterial blood pressure. Any method of measuring continuous arterial blood pressure can be used, including non-invasive and invasive methods. Thus, blood pressure can be measured via an intra-arterial catheter with a disposable pressure transducer (e.g., an intra-arterial pressure catheter), via a pressurized cuff and an optical sensor (e.g., a volume clamp method), via applanation tonometry, or any other method that produces an arterial pressure waveform or a signal proportional to or derived from arterial blood pressure.
[0073] The method 100 also extracts (103) hemodynamic data features from the waveform data. Various hemodynamic data features can be used to screen for sepsis or predict the probability of sepsis. In general, any hemodynamic data feature that can provide predictive capabilities can be utilized. Some hemodynamic data features have been found to provide predictive capabilities. Hemodynamic data features that can be extracted and used to predict or screen for sepsis include (but are not limited to) heart rate, respiratory rate, cardiac output, stroke volume, stroke volume variation, vascular tone, contractility, afterload, systemic vascular resistance, systolic pressure, diastolic pressure, mean arterial pressure (MAP), kurtosis of pressure distribution, left ventricular ejection time, time from when systolic MAP is reached to dicrotic notch, sample entropy of time when systolic MAP is reached, entropy of cardiac interval, entropy of standard deviation of decay phase, or
[0074] The method 100 also utilizes hemodynamic data features extracted from the measured waveform data to screen for sepsis or predict (105) the probability of sepsis. In some embodiments, the extracted features are input into a predictive computational model, wherein the model provides a result indicating the risk or probability of developing or having sepsis. In some embodiments, the extracted features are input into an equation, wherein the equation provides a result indicating the risk or probability of developing or having sepsis. In some embodiments, patient clinical information is input into the model. The clinical information may include, but is not limited to, patient demographics, patient vital signs, and patient laboratory test results.
[0075] In order to screen for sepsis or predict the probability of sepsis, a computational model can be trained for hemodynamic data collected from a cohort of patients known to be diagnosed with sepsis. The hemodynamic data of each patient can be associated with the patient's sepsis diagnosis to train the model. Various computational models can be used, including (but not limited to) regression-based or classification-based models. Regression-based models include (but are not limited to) LASSO regression, ridge regression, k-nearest neighbor, elastic net, least angle regression (LAR) and random forest regression. Classification-based models include (but are not limited to) logistic regression, support vector machine (SVM), decision tree, random forest and naive Bayes. In some embodiments, the model is regularized. In some embodiments, the model can be integrated from multiple models of one or more model types listed above.
[0076] To screen for sepsis or predict the probability of sepsis, an equation can be developed using hemodynamic data collected from a cohort of patients known to be diagnosed with sepsis. Weights can be applied to the various hemodynamic data features in the equation to produce a score that provides a diagnostic indication of sepsis.
[0077] In one example, a machine learning model has been developed for screening for early sepsis identification using features extracted from arterial blood pressure waveforms. The machine learning model develops an equation that utilizes the following features: heart rate (hr), kurtosis of the pressure distribution (kurt), and sample entropy of the time to reach systolic MAP (sampEn). In a specific embodiment, the equation is calculated as follows:
[0078]
[0079] In another example, a machine learning model has been developed for screening for early sepsis identification using features extracted from arterial blood pressure waveforms. The machine learning model develops an equation that utilizes the following features: heart rate (hr), arterial tension factor (avgK), sample entropy of attenuation area (decAreaSampEn), dynamic arterial elasticity (dynEa), and approximate entropy of time during systole (tSysApEn). In a specific embodiment, the equation is calculated as follows:
[0080]
[0081] The screening score indicates an early risk of developing sepsis, as provided by a score range from 0 to 100. A higher score means a higher risk of developing sepsis. The selected features, weights of the features, and scaling of the scores are all provided as examples for generating a screening score for sepsis. Therefore, the selected features, weights of the features, and scaling of the scores can be modified as understood in the art.
[0082] In some embodiments, when a patient's calculated screening score indicates a risk of developing sepsis, the patient is further screened for sepsis complications. Further screening may include, but is not limited to, assessment of systemic inflammatory response syndrome (SIRS) criteria, blood lactate concentration, blood culture assessment for bacterial infection, assessment of organ function, and calculation of a sequential organ failure assessment (SOFA) score. In some embodiments, when a patient's calculated screening score indicates a risk of developing sepsis, the patient is monitored by a clinician for a specific period of time.
[0083] In one example, a machine learning model has been developed for predicting the probability of having sepsis using features extracted from an arterial blood pressure waveform. The machine learning model develops an equation that utilizes the following features: heart rate (hr), diastolic pressure (Dia), time from systolic MAP to dicrotic notch (timeMAP), entropy of the beat-to-beat interval (enIBI), entropy of the standard deviation of the decay phase (enDecay). In one specific embodiment, the equation is calculated as follows:
[0084]
[0085] In another example, a machine learning model has been developed for predicting the probability of having sepsis using features extracted from an arterial blood pressure waveform. The machine learning model develops an equation that utilizes the following features: heart rate (hr), approximate entropy of the blood pressure waveform (ApEnV), sample entropy of the systolic area (areaSampEn), approximate entropy of the time of systolic phase (tSysApEn), and approximate entropy of the time of systolic decay (tDecApEn). In a specific embodiment, the equation is calculated as follows:
[0086]
[0087] The probability score indicates the probability that the individual is experiencing sepsis, as provided by a percentage range from 0% to 100%. The higher the percentage, the higher the likelihood of experiencing sepsis. The selected features, weights of the features, and scaling of the scores are all provided as examples for generating a probability score for sepsis. Therefore, the selected features, weights of the features, and scaling of the scores can be modified as understood in the art.
[0088] In some embodiments, when the patient's calculated probability score indicates a high probability of having sepsis, the patient is diagnosed as having sepsis. In some embodiments, when the patient's calculated probability score indicates a high probability of having sepsis, the patient is further screened to confirm the score result. Further screening may include, but is not limited to, assessment of systemic inflammatory response syndrome (SIRS) criteria, assessment of organ function, and calculation of a sequential organ failure assessment (SOFA) score. When the patient's calculated probability score indicates a high probability of having sepsis, treatment for sepsis is administered to the patient. Treatment for sepsis includes, but is not limited to, administration of antibiotics, administration of intravenous fluids, administration of vasopressors, and surgical removal of abscesses, infected tissue, or dead tissue.
[0089] Although specific examples of methods for screening for sepsis or predicting the probability of sepsis are described above, it will be appreciated by those of ordinary skill in the art that the various steps of the method may be performed in a different order, and that certain steps may be optional according to various embodiments. Therefore, it should be clear that the various steps of the method may be used appropriately for the requirements of a specific application. In addition, any of the various methods for screening for sepsis or predicting the probability of sepsis that are suitable for the requirements of a given application may be utilized in various embodiments.
[0090] Feature Selection
[0091] As explained in the previous section, hemodynamic data and / or other clinical information (including but not limited to patient demographics, vital signs, and laboratory test results) are used as features to build a computational model, which is then used to screen for sepsis or predict the probability of sepsis. The features used to train the model can be selected in a variety of ways. In some cases, the features are determined by which data provide a strong correlation with the diagnosis of sepsis. In some cases, the features are determined using a computational model that can determine which features or combinations of features provide good predictive power.
[0092] Features can be identified and / or selected by several methods. In some cases, relevant features are selected based on the clinical significance of sepsis and related diseases. In some cases, features are selected based on a high correlation with effect or performance to predict effect metrics. Therefore, the strength of the relationship between hemodynamic data and sepsis diagnosis (e.g., SIRS criteria) can be determined. Many statistical methods are known to determine the strength of correlation (e.g., correlation coefficient), including linear association (Pearson correlation coefficient), Kendall rank correlation coefficient, and Spearman rank correlation coefficient). In some cases, a computational model can identify features or feature combinations based on its cost function. Computational models for selecting features include (but are not limited to) LASSO, elastic net, and ridge regression, which can identify features based on their performance using weights or coefficients. The computational model for identifying useful features can be a model different from (or the same as) a prediction model for providing early screening of sepsis or predicting the probability that a patient is experiencing sepsis. In some cases, a computational method can search all possible features and identify which features are the most sensitive. Some computational methods for searching for features and identifying sensitivities include, but are not limited to, restricted recursive feature elimination and information gain criteria. In some cases, ensemble methods are used that combine multiple models for feature selection and model development. In any approach, an appropriate computational model can be selected that results in a number of manageable features. For example, building a predictive model from a large number of features may have overfitting problems. Likewise, too few features may result in lower predictive power.
[0093] Computing and monitoring systems
[0094] The computational processing system for screening for sepsis or predicting the probability of sepsis according to the various methods and processes of the present disclosure generally utilizes a processing system including one or more of a CPU, a GPU, and / or a neural processing engine. Waveform data corresponding to arterial blood pressure or a signal proportional to or derived from arterial blood pressure may be recorded by a sensor. Sensors include (but are not limited to) intra-arterial catheters, disposable pressure transducers, pressurized finger cuffs, and optical sensors, and applanation tonometers. In addition, hemodynamic data features may be extracted from the waveform data to screen for sepsis or predict the probability of sepsis.
[0095] The computing and processing system may be housed within the patient monitor in the form of a direct connection between the monitor and / or components (including sensors). Alternatively, the computing and processing system may be housed separately from the patient monitor and / or components, receiving acquired waveform data via a wired or wireless connection (e.g., WiFi, cellular, Bluetooth, etc.). The computing and processing system may be implemented on any suitable computing device (e.g., but not limited to, a patient monitor, a tablet computer, and / or a portable computer).
[0096] An exemplary computing processing system that can be used to perform the various methods and processes of the present disclosure is provided in Figure 2 and Figure 3 Shown in. Figure 2 Describing a computational system for screening for sepsis (e.g., detecting the probability of early onset of sepsis), and Figure 3 Depicting a computing system for predicting the probability that a patient is experiencing sepsis. The computing processing system 110 includes a processor system 112, an I / O interface 114, a memory system 116, and a sensor 118. It can be readily appreciated that the processor system 112, the I / O interface 114, and the memory system 116 can be implemented using any of a variety of components suitable for the requirements of a particular application, including (but not limited to) a CPU, a GPU, an ISP, a DSP, a wireless modem (e.g., WiFi, Bluetooth modem), a serial interface, a volatile memory (e.g., DRAM), and / or a non-volatile memory (e.g., SRAM and / or NANO flash memory).
[0097] The sensor 118 can be applied to the patient to sense waveform data of the patient, which corresponds to the arterial blood pressure or a signal proportional to or derived from the arterial blood pressure. The sensor 118 is operably connected to the monitoring system 110 and the I / O interface 114, which can provide a visual representation of the arterial pressure waveform captured from the sensor. The sensor 118 can be a non-invasive or invasive pressure sensor. Therefore, the sensor 118 can be an intra-arterial catheter with a disposable pressure transducer (e.g., an intra-arterial pressure catheter), a pressurized finger cuff and an optical sensor (e.g., a volume clamp method), an applanation tonometer, or any other pressure sensor that generates an arterial pressure waveform.
[0098] In the illustrated example, the memory system 116 is capable of storing various data and models. It should be understood that the listed data and models are a representative sample of things that can be stored in the memory, and various memory systems can store some or all of the various data and models listed. In addition, any combination of data and models can be stored, and in some embodiments, various data, applications and / or models are temporarily stored.
[0099] In some embodiments, the memory system 116 can store waveform data 200, which can be obtained from the sensor 118. The application can extract hemodynamic data 202 from the waveform data 200, which can also be stored in the memory system 116. The extracted hemodynamic data 202 can be used in a sepsis screening model 204, which can be stored in the memory system 116. The processor system 112 is configured to execute the sepsis screening model 204 to generate a calculated score 206 indicating early screening of sepsis in the patient. In addition, the waveform data 200 and / or the calculated score 206 can be displayed on a monitor or other screen via the I / O interface 114.
[0100] In some embodiments, the memory system 116 can store waveform data 300, which can be obtained from the sensor 118. The application can extract hemodynamic data 302 from the waveform data 300, which can also be stored in the memory system 116. The extracted hemodynamic data 302 can be used in a sepsis probability model 304, which can be stored in the memory system 116. The processor system 112 is configured to execute the sepsis probability model 304 to generate a calculated score 306 indicating a probability that the patient has sepsis. In addition, the waveform data 300 and / or the calculated score 306 can be displayed on a monitor or other screen via the I / O interface 114.
[0101] Based on the calculated score, the monitoring system 110 can provide alerts of screening results and / or sepsis probability results to clinicians. In particular, in the event that an individual is predicted to have sepsis, the alert can enable timely and effective intervention to prevent organ failure or other serious complications associated with sepsis.
[0102] Although the above reference Figure 2 and Figure 3 A specific computing processing system is described, but it should be readily understood that the computing processes and / or other processes for providing sepsis screening or prediction can be implemented on any of a variety of processing devices (including combinations of processing devices). Therefore, the computing device should be understood as not limited to a specific monitoring system, computing processing system, and / or specific applications and models. The computing device can be implemented using any of the combinations of systems described herein and / or modified versions of the systems described herein to perform the processes, combinations of processes, and / or modified versions of processes described herein.
Claims
1. A computational method for screening for sepsis, comprising: receiving waveform data from a sensor applied to the patient, the waveform data corresponding to arterial blood pressure or a signal proportional to or derived from the arterial blood pressure; extracting a set of hemodynamic data features from the waveform data; and The extracted set of hemodynamic data features are input into a predictive computing model to generate a screening score for sepsis, wherein the predictive computing model has been trained to screen for sepsis using the extracted set of hemodynamic data features.
2. The calculation method according to claim 1, further comprising: The arterial blood pressure is sensed using the sensor.
3. The calculation method according to claim 1 or 2, wherein the sensor is an intra-arterial catheter and a disposable pressure transducer, a pressurized finger cuff and a light sensor or an applanation tonometer.
4. The calculation method according to claim 1, 2 or 3, wherein the extracted set of hemodynamic features includes at least one of the following: heart rate, respiratory rate, cardiac output, stroke volume, stroke volume variation, vascular tone, contractility, afterload, systemic vascular resistance, systolic pressure, diastolic pressure, mean arterial pressure MAP, kurtosis of pressure distribution, left ventricular ejection time, time from reaching systolic MAP to dicrotic notch, sample entropy of time when reaching systolic MAP, entropy of heart rate interval, entropy of standard deviation of decay phase, or 5. The computational method according to any one of claims 1 to 4, further comprising inputting patient clinical information into the model.
6. The computational method of claim 5, wherein the patient clinical information comprises at least one of patient demographics, patient vital signs, and patient laboratory results.
7. The calculation method according to any one of claims 1 to 6, wherein the set of features extracted includes heart rate, kurtosis of pressure distribution and sample entropy of time to reach systolic MAP.
8. The calculation method according to any one of claims 1 to 6, wherein the extracted set of Features include heart rate, arterial tension factor, sample entropy of attenuation area, dynamic arterial elasticity, and approximate entropy of systolic duration.
9. The computational method of any one of claims 1 to 8, wherein the predictive computational model utilizes an equation to generate the screening score for sepsis.
10. The calculation method according to claim 9, wherein the equation is: where hr is the heart rate, kurt is the kurtosis of the pressure distribution, and sampEn is the sample entropy of the time to reach systolic MAP.
11. The calculation method according to claim 9, wherein the equation is: where hr is the heart rate, avgK is the arterial tension factor, decAreaSampEn is the sample entropy of the decay area, dynEa is the dynamic arterial elasticity measure, and tSysApEn is the approximate entropy of the systolic period.
12. The computing method according to any one of claims 1 to 11, wherein the predictive computing model is a regression-based model, a classification-based model or an integrated model.
13. The calculation method according to any one of claims 1 to 12, wherein the screening score for sepsis indicates the risk of developing sepsis; the method further comprising: The patient was further evaluated for septic complications.
14. The calculation method according to any one of claims 1 to 13, wherein the screening score for sepsis indicates the risk of developing sepsis; the method further comprising: The patient is monitored for sepsis complications over a specific period of time.
15. A computational method for predicting the probability of a patient experiencing sepsis, comprising: receiving waveform data from a sensor applied to a patient, the waveform data corresponding to, proportional to, or derived from, arterial blood pressure; extracting a set of hemodynamic data features from the waveform data; and The extracted set of hemodynamic data features are input into a predictive computational model to generate a probability score for sepsis, wherein the predictive computational model has been trained to predict sepsis using the extracted set of hemodynamic data features.
16. The calculation method according to claim 15, further comprising: The arterial blood pressure is sensed using the sensor.
17. The calculation method according to claim 15 or 16, wherein the sensor is an intra-arterial catheter and a disposable pressure transducer, a pressurized finger cuff and a light sensor or an applanation tonometer.
18. The calculation method according to claim 15, 16 or 17, wherein the extracted set of hemodynamic features includes at least one of the following: heart rate, respiratory rate, cardiac output, stroke volume, stroke volume variation, vascular tone, contractility, afterload, systemic vascular resistance, systolic pressure, diastolic pressure, mean arterial pressure MAP, kurtosis of pressure distribution, left ventricular ejection time, time from reaching systolic MAP to dicrotic notch, sample entropy of time when reaching systolic MAP, entropy of heart rate interval, entropy of standard deviation of decay phase, or 19. The computational method of any one of claims 15 to 18, further comprising inputting patient clinical information into the model.
20. The computational method of claim 19, wherein the patient clinical information comprises at least one of patient demographics, patient vital signs, and patient laboratory results.
21. The calculation method according to any one of claims 15 to 20, wherein the extracted set of hemodynamic features includes diastolic pressure, heart rate, entropy of cardiac intervals, time from reaching systolic MAP to dicrotic notch, and entropy of standard deviation of decay phase.
22. A calculation method according to any one of claims 15 to 20, wherein the extracted set of hemodynamic features includes heart rate, approximate entropy of blood pressure waveform, sample entropy of systolic region, approximate entropy of systolic time and approximate entropy of systolic decay time.
23. The computational method of any one of claims 15 to 22, wherein the predictive computational model utilizes an equation to generate the probability score for sepsis.
24. The calculation method according to claim 23, wherein the equation is: where Dia is diastolic pressure, hr is heart rate, enIBI is the entropy of the interbeat interval, timeMAP is the time from when systolic MAP is reached to the dicrotic notch, and enDecay is the entropy of the standard deviation of the decay phase.
25. The calculation method according to claim 23, wherein the equation is: where hr is the heart rate, ApEnV is the approximate entropy of the blood pressure waveform, areaSampEn is the sample entropy of the systolic area, tSysApEn is the approximate entropy of the time of systole, and tDecApEn is the approximate entropy of the time of systolic decay.
26. The computational method of any one of claims 15 to 25, wherein the predictive computational model is a regression-based model, a classification-based model, or an integrated model.
27. The computational method of any one of claims 15 to 26, wherein the probability score for sepsis indicates that the patient has sepsis; the method further comprising: The patients were further evaluated for sepsis complications to confirm the probability score.
28. The computational method of any one of claims 15 to 27, wherein the probability score for sepsis indicates that the patient has sepsis; the method further comprising: The patient is administered therapy to treat the sepsis.
29. A patient monitoring system for screening for sepsis via captured waveform data, the system comprising: sensor; and a computing processing system operatively connected to the sensor; The computing and processing system comprises: Processor system; and a memory system comprising one or more applications configured to direct the processor system to: receiving waveform data from a sensor applied to the patient, the waveform data corresponding to arterial blood pressure or a signal proportional to or derived from the arterial blood pressure; extracting a set of hemodynamic data features from the waveform data; and The extracted set of hemodynamic data features are input into a predictive computing model to generate a screening score for sepsis, wherein the predictive computing model has been trained to screen for sepsis using the extracted set of hemodynamic data features.
30. The patient monitoring system of claim 29, the sensor being an intra-arterial catheter and disposable pressure transducer, a pressurized finger cuff and optical sensor, or an applanation tonometer.
31. A patient monitoring system according to claim 29 or 30, wherein the extracted set of hemodynamic features includes at least one of the following: heart rate, respiratory rate, cardiac output, stroke volume, stroke volume variation, vascular tone, contractility, afterload, systemic vascular resistance, systolic pressure, diastolic pressure, mean arterial pressure MAP, kurtosis of pressure distribution, left ventricular ejection time, time from reaching systolic MAP to dicrotic notch, sample entropy of time when reaching systolic MAP, entropy of cardiac interval, entropy of standard deviation of decay phase, or 32. The patient monitoring system of claim 25, 26 or 27, wherein the one or more applications are further configured to direct the processor system to: Patient clinical information was input into the model.
33. The patient monitoring system of claim 32, wherein the patient clinical information comprises at least one of patient demographics, patient vital signs, and patient laboratory results.
34. A patient monitoring system according to any one of claims 29 to 33, wherein the set of features extracted comprises heart rate, kurtosis of pressure distribution and sample entropy of time to reach systolic MAP.
35. A patient monitoring system according to any one of claims 29 to 33, wherein the set of extracted features comprises heart rate, arterial tension factor, sample entropy of attenuation area, dynamic arterial elasticity amount and approximate entropy of time of systole.
36. The patient monitoring system of any one of claims 25 to 29, wherein the predictive computational model utilizes an equation to generate the screening score for sepsis.
37. The patient monitoring system of claim 36, wherein the equation is: where hr is the heart rate, kurt is the kurtosis of the pressure distribution, and sampEn is the sample entropy of the time to reach systolic MAP.
38. The patient monitoring system of claim 36, wherein the equation is: where hr is the heart rate, avgK is the arterial tension factor, decAreaSampEn is the sample entropy of the decay area, dynEa is the dynamic arterial elasticity measure, and tSysApEn is the approximate entropy of the systolic period.
39. A patient monitoring system according to any one of claims 29 to 38, wherein the predictive computational model is a regression-based model, a classification-based model, or an integrated model.
40. A patient monitoring system according to any one of claims 29 to 39, wherein the one or more applications are further configured to direct the processor system to: The screening score for sepsis is displayed on a monitor operably connected to the computing and processing system.
41. The patient monitoring system of any one of claims 29 to 40, wherein the screening score for sepsis indicates a risk of developing sepsis; wherein the one or more applications are further configured to direct the processor system to: Upon determining that the screening score for sepsis indicates a risk of developing sepsis, an alert is provided indicating the risk.
42. A patient monitoring system for predicting whether a patient is experiencing sepsis via captured arterial pressure, the system comprising: sensor; and a computing processing system operatively connected to the sensor; The computing and processing system comprises: Processor system; and a memory system comprising one or more applications configured to direct the processor system to: receiving waveform data from a sensor applied to the patient, the waveform data corresponding to arterial blood pressure or a signal proportional to or derived from the arterial blood pressure; extracting a set of hemodynamic data features from the waveform data; and The extracted set of hemodynamic data is input into a predictive computational model to generate a probability score for sepsis, wherein the predictive computational model has been trained to predict sepsis using the extracted set of hemodynamic data features.
43. A patient monitoring system according to claim 42, wherein the sensor is an intra-arterial catheter and a disposable pressure transducer, a pressurized finger cuff and a light sensor, or an applanation tonometer.
44. A patient monitoring system according to claim 42 or 43, wherein the extracted set of hemodynamic features includes at least one of the following: heart rate, respiratory rate, cardiac output, stroke volume, stroke volume variation, vascular tone, contractility, afterload, systemic vascular resistance, systolic pressure, diastolic pressure, mean arterial pressure MAP, kurtosis of pressure distribution, left ventricular ejection time, time from reaching systolic MAP to dicrotic notch, sample entropy of time when reaching systolic MAP, entropy of cardiac interval, entropy of standard deviation of decay phase, or 45. A patient monitoring system according to claim 42, 43 or 44, wherein the one or more applications are further configured to direct the processor system to: Patient clinical information was input into the model.
46. The patient monitoring system of claim 45, wherein the patient clinical information comprises at least one of patient demographics, patient vital signs, and patient laboratory results.
47. A patient monitoring system according to any one of claims 42 to 46, wherein the extracted set of hemodynamic features includes diastolic pressure, heart rate, entropy of cardiac intervals, time from reaching systolic MAP to dicrotic notch, and entropy of standard deviation of decay phase.
48. A patient monitoring system according to any one of claims 42 to 46, wherein the extracted set of hemodynamic features includes heart rate, approximate entropy of blood pressure waveform, sample entropy of systolic region, approximate entropy of systolic time and approximate entropy of systolic decay time.
49. The patient monitoring system of any one of claims 42 to 48, wherein the predictive computational model utilizes an equation to generate the probability score for sepsis.
50. The patient monitoring system of claim 49, wherein the equation is: where Dia is diastolic pressure, hr is heart rate, enIBI is the entropy of the interbeat interval, timeMAP is the time from when systolic MAP is reached to the dicrotic notch, and enDecay is the entropy of the standard deviation of the decay phase.
51. The patient monitoring system of claim 49, wherein the equation is: where hr is the heart rate, ApEnV is the approximate entropy of the blood pressure waveform, areaSampEn is the sample entropy of the systolic area, tSysApEn is the approximate entropy of the time of systole, and tDecApEn is the approximate entropy of the time of systolic decay.
52. The patient monitoring system of any one of claims 42 to 51, wherein the predictive computational model is a regression-based model, a classification-based model, or an integrated model.
53. A patient monitoring system according to any one of claims 42 to 52, wherein the one or more applications are further configured to direct the processor system to: The probability score for sepsis is displayed on a monitor operably connected to the computing and processing system.
54. The patient monitoring system of any one of claims 42 to 53, wherein the probability score for sepsis indicates a risk of developing sepsis; wherein the one or more applications are further configured to direct the processor system to: Upon determining that the probability score for sepsis indicates that the patient is experiencing sepsis, providing an alert indicating that the patient is experiencing sepsis.