Systems and methods for predicting patient health status

Abstract

To provide systems and methods for treating a patient in cardiogenic shock.SOLUTION: An intravascular heart pump system is inserted into the vasculature of the patient. The heart pump system has a cannula, a pump outlet, a pump inlet, and a rotor. The heart pump system is positioned within the patient such that the cannula extends across the patient's aortic valve, the pump inlet is located within the patient's left ventricle, and the pump outlet is located within the patient's aorta. Data related to time-varying parameters of the heart pump system is acquired from the heart pump system. Multiple features are extracted from the data. A probability of survival of the patient is determined based on the multiple features and using a prediction model. The heart pump system is operated to treat the patient.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] REFERENCE TO RELATED APPLICATIONS This application claims priority to and benefit of U.S. Provisional Patent Application No. 62 / 609,158, entitled "SYSTEMS AND METHODS FOR PREDICTING PATIENT HEALTH STATUS," filed December 21, 2017. The entire contents of the above referenced application are incorporated herein by reference. [Background technology]

[0002] background Acute and chronic cardiovascular conditions reduce quality and longevity of life. A variety of treatments have been developed for cardiac health, ranging from pharmaceuticals to mechanical devices and transplants. Temporary cardiac assist devices, such as cardiac pump systems, provide hemodynamic support and promote cardiac recovery. Some cardiac pump systems, such as the IMPELLA® family of devices (Abiomed, Inc., Danvers MA), can be inserted percutaneously into the heart and operate in parallel with the native heart to supplement cardiac output.

[0003] Currently, it is difficult or impossible for clinicians to track the health status of patients. Clinicians tend to rely on qualitative judgments and indirect estimates of cardiac function to predict the health status of patients, but these processes are inconsistent and unreliable. Judgments of the health status of patients may vary between clinicians. Furthermore, this process is time-consuming for clinicians, and often clinicians are unable to analyze all measurements associated with the patient's cardiac function in time to make informed medical decisions. Summary of the Invention

[0004] overview The systems, devices, and methods described herein use predictive modeling to forecast patient outcomes and track a patient's condition over time, particularly in relation to the cardiac health of patients in cardiovascular distress and / or suffering from cardiogenic shock. In particular, these systems, devices, and methods allow a heart pump system to provide data useful for determining a patient's probability of survival. One way to use the patient's probability of survival is to rank a set of patients from lowest to highest probability, or the patient's probability of survival can be used to assign the set of patients to different strata for different ranges of survival probabilities. In this way, the systems and methods described herein are a quantitative and objective way for clinicians to identify the patients with the most dire conditions and direct the clinician's immediate attention to those patients who need it most. Another way to use the patient's probability of survival to track an individual patient's probability of survival over a period of time is to provide a quantitative assessment of the patient's health over time. In this way, the systems and methods provide a quantitative and objective way for clinicians to identify whether the patient's health is progressing as expected, so that the clinician can update the patient's treatment plan as necessary.

[0005] The patient's probability of survival may be determined based on one or more of a variety of factors including at least continuous and / or discrete measurements of cardiac performance acquired by the heart pump system. For example, one data parameter provided by the heart pump system may include cardiac power output (CPO). The CPO value, along with one or more clinical data parameters, such as lactate concentration measured from the patient, may be used to determine the probability of survival, which may then be used to modify the operation of the heart pump system. Systems and methods for acquiring CPO and lactate concentration are described in detail below. One way to modify the operation of the heart pump system is to increase or decrease the level of cardiac assistance from the heart pump system depending on the probability of survival. For example, if the probability of survival is high, the patient's outlook is good and the heart pump system decreases the level of cardiac assistance. Alternatively, if the probability of survival is low, the patient's outlook is poorer and the heart pump system increases the level of cardiac assistance.

[0006] In some aspects, an intravascular heart pump system is inserted into the patient's vasculature. The heart pump system may be inserted using a minimally invasive procedure. For example, the heart pump system may be inserted through the femoral artery or femoral vein by catheterization. In some embodiments, the heart pump system includes a cannula, a pump inlet, a pump outlet, and a rotor. For example, the intravascular heart pump system is a percutaneous ventricular assist device, such as the IMPELLA® family of devices (Abiomed, Inc., Danvers MA). In some embodiments, the rotor is coupled to a motor. The motor may drive the rotor and pump blood through the pump. In some embodiments, the heart pump system includes one or more sensors. For example, the sensors may be configured to obtain data related to the performance, cardiac function, hemodynamic performance, or any other suitable data of the heart pump system. In some embodiments, the heart pump system includes a controller. For example, the heart pump system may include an Automated Impella Controller (AIC). The controller may be configured to execute instructions, analyze data, calculate values, determine relationships between parameters, or perform any other suitable task. For example, the controller may perform the methods described herein. The controller may include a processor, memory, a user interface, a display screen, a touch screen, user interactive buttons and / or dials, a power source, any other suitable elements, or any combination thereof.

[0007] In some embodiments, the heart pump system is positioned partially within the patient's heart. In some embodiments, the heart pump system is a left ventricular assist device (LVAD). The heart pump system may be positioned within the patient such that the cannula extends beyond the patient's aortic valve, the pump inlet is located within the patient's left ventricle, and the pump outlet is located within the patient's aorta. For example, the heart pump system may be inserted by catheterization through the femoral artery, into the ascending aorta, past the aortic valve, and into the left ventricle. In some embodiments, the heart pump system is a right ventricular assist device (RVAD). For example, the heart pump system may be inserted by catheterization through the femoral vein into the right atrium. Although some embodiments presented herein are directed to a heart pump system implanted beyond the aortic valve and partially residing in the left ventricle, the same concepts may be applied to devices in other regions of the heart, cardiovascular system, or body.

[0008] In some aspects, the systems and methods obtain first data related to time-varying parameters of a cardiac pump system, extract a plurality of features from the first data, and determine a cardiac health index. The cardiac health index represents the cardiac health of a patient and may indicate the functioning of the patient's heart and systemic perfusion that lead to overall patient recovery and outcome. In some embodiments, the cardiac health index represents a value indicative of the likelihood or probability of survival of the patient.

[0009] The systems, devices, and methods presented herein use measurements related to a patient's health to determine a cardiac health index and / or predict patient survival. In some embodiments, the measurements are cardiac parameters related to cardiac function. In some embodiments, a cardiac pump system obtains, measures, processes, or otherwise quantifies the measurements.

[0010] The methods described herein may include obtaining first data or measurements related to a time-varying parameter of the heart pump system (such as any of the measurements described below). The first data may represent continuous or near-continuous measurements obtained through the heart pump system or may represent known quantities such as inputs to the heart pump system. The first data may relate to the operation of the heart pump system or factors measured by the heart pump system and may include data indicative of heart rate, pump pressure, differential pressure, motor current, P-level, motor speed, any other data provided directly by the heart pump system or inferred from data provided directly by the heart pump system, or any suitable combination thereof. From these measurements, information regarding cardiac function and, in some cases, performance of the cardiac assist device (e.g., occurrence of aspiration events, etc.) can be determined. This information regarding cardiac function can be used in a predictive modeling system to predict patient outcomes.

[0011] The first data may be determined from measurements taken by one or more sensors on the heart pump system, on an external system, or both. For example, one or more sensors on the heart pump system may be positioned within the patient's heart, external to the patient's heart, or a combination of both during operation of the heart pump system. In one example, a sensor on the heart pump system measures pressure within the patient's vasculature. That pressure may be used in the calculation of additional parameters, such as cardiac power output, as described below.

[0012] Methods described herein may include processing acquired or known data, such as the first data described above, to determine or estimate other parameters or features related to the patient's health or the operation of the heart pump. In some embodiments, these parameters are determined based in part on hysteresis between pressure and motor current measurements that allows detection of a phase of the cardiac cycle corresponding to a given pair of pressure and current measurements. In some embodiments, a number of features are extracted from the first data. Extracting the features may include processing the first data at the heart pump system or at an external device. These features may include left ventricular end diastolic pressure, LVEF, LVH, LVHV ... pressure (LVEDP), stroke volume, ejection fraction, chamber dilation, chamber hypertrophy, chamber pressure, stroke work, preload state, afterload state, heart rate, cardiac recovery, aortic pressure, differential pressure, motor current, motor speed, pump pressure, left ventricular pressure, end diastolic pressure, aortic pulse pressure, intrinsic cardiac output, cardiac output, CPO, configuration, mean flow, target flow, P level, contractility, relaxation, configuration signal, mean configuration, standard deviation of configuration, mean configuration range, standard deviation of configuration range, mean differential pressure, standard deviation of differential pressure, mean differential pressure range, standard deviation of differential pressure range, left ventricular pressure maximum, left ventricular pressure minimum, pump pressure maximum, pump pressure average, pump pressure minimum, differential pressure maximum, differential pressure minimum, motor current maximum, motor current minimum, motor current average, motor speed average, any other suitable feature related to cardiac function, and any combination thereof. The first data may be obtained during a first time period during which the cardiac pump system is operating, such as one second, one minute, five minutes, ten minutes, one hour, several hours, one day, several days, one week, one month, or any suitable time frame. The representative, average, and minimum values ​​of the features described above may be representative, average, or minimum values ​​of the features during the first time period. The systems and methods described herein may use these multiple features to determine a patient's probability of survival or other cardiac health index, as described below.

[0013] In some embodiments, the methods described herein include obtaining second data related to a physiological parameter of the patient. The second data may be measured from the patient by a clinician or by a device external to the heart pump system, or may be inferred from measurements. The second data may include temperature, weight, height, waist circumference, body surface area (BSA), age, sex, urine volume, creatinine level, potential of Hydrogen (pH), oxygen concentration, carbon dioxide concentration, lactate concentration, or any other suitable measurement or patient sample, such as blood, urine, saliva, plasma, feces, urine, tissue, or any other suitable sample. For example, the clinician may collect and analyze a blood sample from the patient to obtain the second data. In some embodiments, the second data is obtained during the same time period that the first data is obtained. A cardiac health index or probability of survival may be based on the second data.

[0014] The second data may be obtained via one or more sensors on the heart pump system and / or via an external system. The one or more sensors on the heart pump system and / or the external system may be positioned within the patient's heart, external to the patient's heart, or a combination of both. For example, a clinician may measure a lactate concentration value in the patient's blood and then enter the lactate concentration into a user interface on the heart pump system or another system.

[0015] In some embodiments, the heart pump system itself receives and processes both the first data related to cardiac function and the second data related to physiological parameters. The heart pump system then calculates the cardiac health index or survival probability as described herein. In other embodiments, a device separate from the heart pump system, such as a computer, a mobile device, a tablet, or any other suitable device, receives the first data and the second data and determines the cardiac health index or survival probability based on the data, as described below.

[0016] In some embodiments, the methods described herein include determining a cardiac health index indicative of the patient's cardiac health. The cardiac health index may be indicative of the patient's likelihood of recovery, including a cardiac component and a systemic perfusion component. The cardiac component is related to the patient's cardiac health and may include unloading, contractility, or any suitable indicator of the patient's cardiac performance. The system perfusion component is related to the health of the patient's vasculature and may include cardiac output (CO), aortic pressure mean (AoPm), or any suitable indicator of the patient's circulatory system performance. In some aspects, the cardiac health index may be the patient's survival probability. The survival probability is a value indicative of the patient's likelihood of survival or death. In some embodiments, the survival probability is a number, e.g., between 0 and 1. In some embodiments, the survival probability is indicative of survival if the probability is equal to or greater than a threshold value (e.g., 0.5) (e.g., the patient has more than a 50% chance of living given the patient's cardiac health). The survival probability may be based on the features described above. For example, a patient with a low cardiac output, a low maximum pressure, a high minimum pressure, a high standard deviation of the differential pressure, or any suitable combination thereof, may have a low probability of survival, and a patient with a high cardiac output, a high maximum pressure, a low minimum pressure, a low standard deviation of the differential pressure, or any suitable combination thereof, may have a high probability of survival.

[0017] In some aspects, the method includes operating a cardiac pump system to treat a patient, e.g., turning on a pump, adjusting the level of assistance provided by the pump (e.g., by adjusting a motor speed to increase or decrease the level of assistance), or stopping the pump. For example, if a patient has a low CPO and high lactate levels, the pump is turned on or turned on, or the level of assistance is increased while continuing to monitor the patient's health. In a patient with a high CPO and low lactate levels, an already operating pump is stopped or turned off, or the level of assistance is decreased while continuing to monitor the patient's health.

[0018] In some implementations, a pump operating parameter value is selected based on the survival probability. The pump operating parameter may be any factor that affects the operation of the pump. For example, the pump operating parameter may be a pump speed, a P level, a motor current, a target flow rate, or any other suitable parameter. In some implementations, the pump speed is increased based on a cardiac health index, which may be the survival probability (e.g., if the survival probability is low or below a certain threshold). In some implementations, the pump speed is decreased based on a cardiac health index (e.g., if the survival probability is high or above a certain threshold).

[0019] The heart health index may be determined using a predictive model. The predictive model may be a machine learning model. For example, the predictive model may be one of a logistic regression method, a deep learning method, a decision tree method, a random forest method, a naive Bayes method, and a support vector machine method. The heart health index may be based on a plurality of features. The method may further include predicting a patient outcome based on the heart health index. In some aspects, the patient outcome may be the patient's death or survival.

[0020] The method may further include displaying the cardiac health index. For example, the cardiac health index may be displayed on the cardiac pump system using a graphical user interface or remotely on another system. The cardiac health index may be shown as a numerical value, a color representation, a visual indicator, or any other suitable display method. For example, the AIC may display a green color if the patient's survival probability is equal to or greater than a first threshold, a yellow color if the survival probability is between the first threshold and a second threshold lower than the first threshold, and a red color if the survival probability is equal to or less than the second threshold.

[0021] The method may further include obtaining a plurality of cardiac health indices. The plurality of cardiac health indices may include cardiac health indices, each of which may correspond to one of the plurality of time periods. The method may further include determining a change in the patient's health based on the plurality of cardiac health indices. For example, a small change in a patient factor (e.g., CPO, contractility, motor current average, etc.) may appear insignificant when viewed alone, but may indicate a general decline in the patient's health when viewed in combination with other patient factors. These multiple factors may be considered in the cardiac health index. This method of aggregating multiple patient factors into a single value or trend allows a patient or clinician to quickly and easily interpret the patient's health. The method may further include displaying the plurality of cardiac health indices over the plurality of time periods. For example, the plurality of cardiac health indices may be displayed using a graphical user interface (e.g., on an AIC). For example, a clinician may display a graphical representation of the cardiac health index over time to easily visualize trends in the patient's health. In some implementations, if the patient's chance of survival is declining at a constant rate or at a rate above a given threshold, a clinician may be alerted to the patient's declining health. Such notification may include, for example, an audible alarm, a flashing light on a user interface, an email or phone message, or any suitable notification. For example, a clinician may use a cardiac health index to quantitatively determine that the patient's chance of survival is steadily declining over a period of days (or weeks). This determination allows the clinician to intervene in the patient's treatment (e.g., by adjusting the operating parameters of the patient's heart pump) to improve the patient's outlook.

[0022] The method may further include displaying an indication of a relative importance of a first feature of the plurality of features compared to a second feature of the plurality of features. The relative importance may be indicated in a visual display. For example, each feature may be shown as a bar in a bar graph or as a point in a spider plot, with each bar or point in the plot being given a size or placement relative to its importance. In some embodiments, the heart pump system includes a controller including a user interface and a display screen. The relative may be displayed on the display screen. In some embodiments, a clinician may view the indication remotely, for example, via a personal computer or mobile device. For example, the controller may send periodic reports regarding the patient's condition to the clinician automatically or upon request of the clinician.

[0023] In one aspect, a method for measuring a patient's health status may include obtaining a training data set from a database, the training data set including a plurality of data points associated with a time-varying parameter of a heart pump system. For example, a controller or a remote computer system of the heart pump system may be trained with data obtained from a plurality of patient cases in which the patient's outcome (e.g., survival or death) is known. The method may further include pre-processing the data set to determine a plurality of features corresponding to the plurality of data points, and processing the plurality of features to determine a pattern. For example, training the controller or computer system may include determining what patient factors had the greatest and least impact on the patient's outcome. The pattern may include a weight for each feature of a subset of the plurality of features. The method may further include obtaining the patient data and calculating a cardiac health index for the patient based on the patient data and the pattern. By training the controller or computer system with known case data, the computer system can self-correct and "learn" how to accurately predict the patient's probability of survival.

[0024] In one aspect, the heart pump system may include a catheter, a motor, a rotor operably coupled to the motor, a pump housing, at least one sensor, and a controller. The pump housing may at least partially surround the rotor such that activation of the motor drives the rotor and pumps blood into the pump housing. The controller may be configured to perform any of the methods described herein. For example, the controller may obtain first data associated with a time-varying parameter of the heart pump system from the at least one sensor during a first time period, extract a plurality of features from the first data, determine a cardiac health index indicative of the patient's cardiac health using a predictive model and based on the plurality of features, and predict an outcome of the patient based on the cardiac health index.

[0025] In some aspects, an intravascular heart pump system is inserted into the patient's vasculature as described above or throughout the various aspects described herein. The heart pump system may be inserted using a minimally invasive procedure. For example, the heart pump system may be inserted through the femoral artery or femoral vein by catheterization. In some embodiments, the heart pump system is positioned partially within the patient. In some embodiments, the heart pump system is a left ventricular assist device (LVAD). The heart pump system may be positioned within the patient such that the cannula extends past the patient's aortic valve, the pump inlet is located within the patient's left ventricle, and the pump outlet is located within the patient's aorta. For example, the heart pump system may be inserted through the femoral artery, into the ascending aorta, past the aortic valve, and into the left ventricle by catheterization. In some embodiments, the heart pump system is a right ventricular assist device (RVAD). For example, the heart pump system may be inserted through the femoral vein into the right atrium by catheterization.

[0026] In some embodiments, the systems and methods described herein operate or are configured to operate a heart pump system during a first time period to provide a first level of cardiac assistance to a patient. For example, the heart pump system may be operated at a first pump speed, P level, or motor parameter, such as current delivered to the motor, power delivered to the motor, motor speed, etc. In some embodiments, the system operates to provide a constant or near-constant level of assistance to the patient.

[0027] In some embodiments, the systems and methods described herein obtain at least one CPO value derived from measurements provided by the heart pump system. The CPO represents the pumping power of the heart. The CPO is a function of mean arterial pressure and cardiac output, where the mean arterial pressure is a function of systolic and diastolic blood pressure, and the cardiac output is a function of heart rate and stroke volume. Cardiac output can be estimated or measured by various means, such as calculating the area under the volume-pressure curve for the patient's cardiac cycle. In some examples, the CPO may be equal to the mean arterial pressure multiplied by the cardiac output divided by 451. In some embodiments, a cardiac power index (CPI) is used instead of or in addition to the CPO. The CPI represents the pumping power of the heart normalized by body surface area. In some embodiments, the CPO is calculated from pressure measurements taken by one or more sensors of the heart pump system. In some embodiments, obtaining at least one CPO value includes determining cardiac output over time from sensors of the heart pump system. For example, the controller of the heart pump system may determine the CPO from systolic, diastolic, and / or differential pressure measurements taken during operation of the pump system within the patient's vasculature. In some implementations, the CPO is calculated each time a pressure measurement is updated at a sensor or an updated pressure measurement is received by the controller. Alternatively, the CPO may be calculated only when an updated pressure measurement is received that differs from the previous measurement by a certain amount. In some implementations, the CPO is regularly updated at a fixed time interval following a first period of time. For example, this first time interval may be 0.01 seconds, 0.1 seconds, 0.5 seconds, 1 second, 5 seconds, 10 seconds, 1 minute, 10 minutes, 15 minutes, 30 minutes, 1 hour, or any suitable time interval.

[0028] In some embodiments, the systems and methods described herein obtain at least one lactate concentration value measured from the patient. The lactate concentration represents the balance between lactate production and clearance in the patient. The lactate concentration may be measured via the patient's blood. For example, a clinician may measure the lactate concentration by drawing blood from the patient. In some embodiments, the lactate concentration is manually entered by the clinician or other user into a user interface connected to the heart pump system or another device. In some embodiments, the lactate concentration is imported via an electronic wired or wireless connection. For example, the patient's lactate concentration may be stored in a remote memory location that communicates with the heart pump system to provide the physiological parameter value for processing. In some embodiments, the lactate concentration value is regularly updated at a fixed time interval following the first period. For example, this second time interval may be 1 hour, 3 hours, 5 hours, 7 hours, 10 hours, 1 day, 1 week, or any suitable time interval.

[0029] In some embodiments, the systems and methods described herein determine a prediction of a patient outcome. The patient outcome may be based on at least one CPO value and at least one lactate concentration value. In some embodiments, the prediction of the patient outcome represents the likelihood of the patient surviving or dying. For example, the prediction of the patient outcome may be a value between 0 and 1, with 1 representing a high likelihood of the patient surviving and 0 representing a low likelihood of the patient surviving.

[0030] In some embodiments, the systems and methods described herein operate a cardiac pump system to treat a patient. For example, pump operation may be altered based on a predictive value of a patient outcome. In particular, altering pump operation may include adjusting an operating parameter of the cardiac pump system to provide a second level of cardiac assistance during a second time period following the first time period. The second level of cardiac assistance may be the same as the first level of cardiac assistance, or the second level of cardiac assistance may be different from the first level of assistance. In one example, adjusting the operating parameter of the cardiac pump system includes adjusting a pump speed (e.g., by increasing or decreasing) based on changes in cardiac power output, lactate concentration, or both. It may be desirable to increase the pump speed when at least one CPO value falls below a first threshold, when at least one lactate concentration value rises above a second threshold, or both. For example, the first threshold value may be 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1.0 W, etc., and the second threshold value may be 1, 2, 3, 4, 5, 6, 7 mmol / L, etc. A low CPO value and a high lactate value may indicate that the patient has a relatively low chance of survival. Because the patient is in poor physical condition, the clinician may attempt to increase the level of cardiac assistance provided by the pump, for example, by increasing the pump speed. It may be desirable to decrease or not change the pump speed when at least one CPO value is above the first threshold, when at least one lactate concentration value is below the second threshold, or both. A high CPO value and a low lactate value may indicate that the patient has a relatively high chance of survival. Because the patient is in good physical condition, the clinician may decide not to change the parameters of the pump's operation. Alternatively, the clinician may attempt to decrease the level of cardiac assistance provided by the pump, or turn the pump off entirely. [The present invention 1001] 1. A method for treating a patient with cardiogenic shock, comprising: inserting an intravascular heart pump system, including a cannula, a pump inlet, a pump outlet, and a rotor, into the patient's vasculature; positioning the cardiac pump system within the patient such that a cannula extends past the patient's aortic valve, a pump inlet is located within the patient's left ventricle, and a pump outlet is located within the patient's aorta; obtaining first data from the cardiac pump system related to a time-varying parameter of the cardiac pump system; extracting a plurality of features from the first data; determining a probability of survival for said patient based on said plurality of features using a predictive model; and operating the cardiac pump system to treat the patient. A method comprising: [The present invention 1002] The method of claim 1001, wherein the survival probability includes a cardiac component representative of the functioning of the patient's heart and a system perfusion component representative of the functioning of the patient's circulatory system. [The present invention 1003] acquiring second data related to a physiological parameter of the patient; wherein determining the survival probability is based on the second data. The method of the present invention 1001 or 1002. [The present invention 1004] The method of claim 1003, wherein the second data includes at least one of age, sex, body surface area (BSA), urine volume, creatinine level, pH, oxygen concentration, carbon dioxide concentration, and lactate concentration. [The present invention 1005] The method of any one of claims 1004 to 12, wherein the first plurality of features includes cardiac power output and the second plurality of features includes lactate concentration. [The present invention 1006] Any of the methods of the preceding invention, wherein pump operating parameter values ​​are selected based on said survival probability. [The present invention 1007] The method of claim 1006, wherein the pump operating parameter is a pump speed. [The present invention 1008] The method of any one of claims 10 to 15, wherein the pump speed is increased based on the survival probability. [The present invention 1009] obtaining a plurality of survival probabilities including the survival probabilities, each survival probability corresponding to one of a plurality of time periods; and determining a change in the patient's health based on the plurality of survival probabilities. Any of the methods of the present invention further comprising: [The present invention 1010] Any of the methods of the invention, wherein the plurality of features comprises at least one of aortic pressure, differential pressure, motor current, motor speed, pump pressure, left ventricular pressure, end diastolic pressure, aortic pulse pressure, intrinsic cardiac output, cardiac output, cardiac power output, configuration, mean flow, target flow, P level, contractility, relaxation, configuration signal, mean configuration, standard deviation of configuration, mean configuration range, standard deviation of configuration range, mean differential pressure, standard deviation of differential pressure, mean differential pressure range, standard deviation of differential pressure range, left ventricular pressure maximum, left ventricular pressure minimum, pump pressure maximum, pump pressure average, pump pressure minimum, differential pressure maximum, differential pressure minimum, motor current maximum, motor current minimum, motor current average, and motor speed average. [The present invention 1011] Any of the methods of the present invention, wherein the predictive model is a machine learning model that is one of a logistic regression method, a deep learning method, a decision tree, a random forest method, a naive Bayes method, and a support vector machine method. [The present invention 1012] displaying an indication of the relative importance of a first feature of the plurality of features compared to a second feature of the plurality of features. Any of the methods of the present invention further comprising: [The present invention 1013] determining the survival probability, obtaining a training data set from a database, the training data set including a plurality of data points related to time-varying parameters of the heart pump system; pre-processing the data set to determine a third plurality of features corresponding to the plurality of data points; processing the third plurality of features to determine a pattern including a weight for each feature of the subset of the third plurality of features; acquiring patient data; and calculating a patient's probability of survival based on said patient data and said pattern. Any of the methods of the present invention comprising: [The present invention 1014] Any of the methods of the present invention, wherein said survival probability is indicative of death or survival of said patient. [The present invention 1015] 1. A method for treating a patient with cardiogenic shock, comprising: inserting an intravascular heart pump system, including a cannula, a pump inlet, a pump outlet, and a rotor, into the patient's vasculature; positioning the cardiac pump system within the patient such that a cannula extends past the patient's aortic valve, a pump inlet is located within the patient's left ventricle, and a pump outlet is located within the patient's aorta; operating the cardiac pump system to provide a first level of cardiac assistance to the patient; obtaining at least one cardiac power output value derived from measurements provided by the cardiac pump system and at least one lactate concentration value measured from the patient; determining a predictive value of patient outcome based at least in part on the at least one cardiac power output value and the at least one lactate concentration value; and operating the cardiac pump system to treat the patient. A method comprising: [The present invention 1016] The method of the present invention 1015 further comprises adjusting operating parameters of the cardiac pump system based on the predicted value of the patient's outcome to provide the patient with a second level of cardiac assistance different from the first level. [The present invention 1017] The method of any one of claims 1015 to 1016, wherein adjusting an operating parameter of the cardiac pump system comprises adjusting a pump speed based on changes in cardiac output or lactate concentration. [The present invention 1018] Adjusting the pump speed comprises: increasing the pump speed when the at least one cardiac power output value falls below a first threshold and the at least one lactate concentration value exceeds a second threshold. The method of the present invention 1017, comprising: [The present invention 1019] Adjusting the pump speed comprises: decreasing the pump speed when the at least one cardiac power output value is above a first threshold and the at least one lactate concentration value is below a second threshold. The method of the present invention 1018, comprising: [The present invention 1020] the cardiac pump system providing the first level of cardiac assistance to the patient during a first period of time; updating the at least one cardiac power output value after a first time interval following the first time period; and updating the at least one lactate concentration value after a second time interval following the first period. Any of the methods of 1015 to 1019 of the present invention further comprising: [The present invention 1021] Any of the methods of claims 1015 to 1020, wherein the at least one lactate concentration value is measured from the patient by a clinician and provided to the cardiac pump system on a user interface. [The present invention 1022] A method according to any one of claims 1015 to 1021, wherein said step of obtaining at least one cardiac power output value comprises determining cardiac output over time from a sensor of said cardiac pump system. [The present invention 1023] The method of any of claims 1015 to 1022, wherein said predictive value of a patient outcome is indicative of death or survival of said patient. [The present invention 1024] a pump including a pump housing, a rotor, and an opening positioned proximally of the pump housing, the cannula having a proximal end; a cannula including a proximal end that connects with the distal end of the pump housing and a distal end with at least one distal opening, the pump configured to be operated by the motor; an elongate catheter extending proximally of the pump housing; At least one sensor; and A controller configured to execute any one of the methods of the present invention 1001 to 1023 23. A heart pump system comprising: [The present invention 1025] The heart pump system of the present invention 1024 further comprising a flexible protrusion extending distally from said distal end of said cannula. [The present invention 1026] The heart pump system of any one of claims 1024 to 1025, wherein the elongate catheter is coupled at its distal end to the pump housing, the pump further comprising a drive cable extending through the elongate catheter. [The present invention 1027] a pump including a pump housing, a rotor, and an opening positioned proximally of the pump housing, the cannula having a proximal end; a cannula including a proximal end that connects with the distal end of the pump housing and a distal end having at least one distal opening, the pump being configured to be operated by a motor; an elongate catheter extending proximally of the pump housing; at least one sensor configured to acquire first data related to a time-varying parameter; and a controller configured to determine a probability of survival of the patient based on the first data. 23. A heart pump system comprising: [The present invention 1028] The heart pump system of claim 1027 further comprising a flexible protrusion extending distally from said distal end of said cannula. [The present invention 1029] The heart pump system of any one of claims 1027 to 1028, further comprising an integral motor sized and configured for insertion within the patient's vasculature. [The present invention 1030] A heart pump system according to the present invention 1029, wherein the motor is configured to maintain a substantially constant rotor speed during operation of the rotor. [The present invention 1031] The heart pump system of any one of claims 1027 to 1028, wherein the elongated catheter is coupled at its distal end to the pump housing, and the pump further comprises a drive cable extending through the elongated catheter. [The present invention 1032] Any of the cardiac pump systems of inventions 1027 to 1031, wherein the pump is configured such that a cannula extends beyond the aortic valve of the patient, the distal end is positioned within the left ventricle of the patient, and the proximal end is positioned within the aorta of the patient. [The present invention 1033] A heart pump system of any one of claims 1027 to 1032, wherein the controller is configured to execute any one of the methods of claims 1001 to 1023.

[0031] These and other objects and advantages will become apparent from the following detailed description considered in conjunction with the accompanying drawings, in which like reference characters refer to like parts throughout. [Brief description of the drawings]

[0032] [Figure 1] 1 is a flow chart illustrating a method for patient condition monitoring. [Diagram 2] 1 is a flow chart illustrating a method for predicting a patient outcome. [Diagram 3] FIG. 1 is a block diagram illustrating log data mining. [Figure 4] Two scatter plots with decision boundaries used for training and classification. [Diagram 5] Two scatter plots with decision boundaries used for training and classification. [Figure 6] A scatter plot that is not suitable for decision boundary separation. [Figure 7] A scatter plot that is not suitable for decision boundary separation. [Figure 8-1] 1 is a bar chart ranking the importance of features. [Figure 8-2] 1 is a bar chart ranking the importance of features. [Figure 8-3] 1 is a bar chart ranking the importance of features. [Figure 9] 1 is a spider plot showing feature ratings. [Figure 10] FIG. 2 shows an example of characteristics of a particular patient, Patient X. [Figure 11] FIG. 1 shows an example of characteristics of a particular patient, Patient Y. [Figure 12] FIG. 1 illustrates patient-related characteristics over time in a set of four graphs. [Figure 13] FIG. 1 shows an example of characteristics of a particular patient, Patient Z. [Figure 14] 1 is a flow chart illustrating the determination of a patient's probability of survival. [Figure 15] 1 is a flow chart showing the determination of a predictive value of patient outcome based on CPO and lactate concentration. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0033] Detailed Description To provide an overall understanding of the systems, methods, and devices described herein, certain exemplary aspects will be described. Although the aspects and features described herein are specifically described for use in connection with a patient's cardiac health, it will be understood that all components and other features outlined below may be combined with each other in any suitable manner and may be adapted and applied to other types of medical therapy and patient health.

[0034] The systems, devices, and methods described herein, particularly in relation to cardiac health of patients suffering from cardiovascular distress and / or cardiogenic shock, use predictive modeling to predict patient outcomes and track the progress of the patient's condition over time. The predicted patient outcome may be based on a cardiac health index, which may be used interchangeably with health index throughout this specification. The cardiac health index may include a cardiac component and a systemic perfusion component and may be indicative of the patient's health. In particular, these systems, devices, and methods enable the heart pump system to provide data useful for determining the patient's outcome, i.e., the patient's probability of survival. The cardiac health index may be determined based on one or more of a variety of factors including at least continuous and / or discrete measurements of cardiac performance obtained by the heart pump system. For example, one data parameter provided by the heart pump system may include cardiac power output (CPO). The CPO value, together with one or more clinical data parameters, such as lactate concentration, may be used to determine the patient's probability of survival, which may then be used to modify the operation of the heart pump system.

[0035] The systems and methods described herein also provide classification models of patient outcomes using cardiac parameters. There is an unmet need in the medical field to monitor patient health through predictive modeling, especially for patients with cardiac health issues, such as patients with heart pump systems. Existing predictive modeling systems generally relate to physiological data, but do not take into account the effects of the heart pump system implanted in the patient that may affect the patient's health, nor do existing modeling systems take into account the data provided by such implantable heart pump systems. Furthermore, these systems do not provide a solution for monitoring patient health over time, nor for assisting clinicians in determining which patients are most in need of immediate attention.

[0036] The systems and methods described herein may rely on physiological data measured by the heart pump system or from other sources, as well as data related to the operation of the patient's heart pump, such as motor current to the pump, to predict patient outcome. Such physiological data may include age, sex, body surface area (BSA), and clinical values ​​such as lactate concentration, urine volume, creatinine level, pH, O2 concentration, CO2 concentration, etc. The physiological data and other characteristics used to predict outcome may be manually entered into the systems described herein or automatically pulled from electronic medical records.

[0037] The systems and methods described herein enhance the clinician's ability to quantitatively and objectively determine a patient's cardiac health by incorporating data already available when predicting a patient's outcome. By predicting a patient's probability of survival, the system may alert the clinician to a patient's health issue before it is otherwise detected, thereby providing the clinician more time than standard of care to treat the patient. For example, a patient's health may be slowly declining, but the individual features monitored by the clinician (e.g., heart rate, CPO, stroke volume, etc.) may be too slight or subtle to alert the clinician; however, the cardiac health index may collate these seemingly insignificant changes and clearly represent the overall decline in the patient's health. By representing multiple features within a single metric, the cardiac health index provides an index of the patient's health that is easily interpreted by the clinician efficiently. If the cardiac health index or probability of survival falls below a threshold or is declining rapidly, the clinician may be notified of the patient's declining health and may be able to treat the patient quickly. Furthermore, the prediction of a patient's probability of survival can be used to rank a set of patients or sort them into strata with different ranges of survival probabilities. In this way, clinicians can quantitatively and objectively identify those patients who are most in need of their immediate attention. Again, this is an improvement over the standard of care, where clinicians simply round on their own patients in no particular order.

[0038] The operation of the heart pump system may be modified based on the cardiac health index or the predicted patient outcome. One way to modify the operation of the heart pump system is to increase or decrease the level of cardiac assistance from the heart pump system depending on the survival probability. For example, if the survival probability is high, the patient's outlook is good, and the heart pump system may be updated to maintain or decrease the level of cardiac assistance, or the clinician may even attempt to stop the heart pump. Alternatively, if the survival probability is low, the patient's outlook is poor, and the heart pump system may increase the level of cardiac assistance. The heart pump system may auto-adjust the operation of the pump, or the clinician may manually adjust the operation of the pump.

[0039] In some embodiments, the system trains a machine learning method (classification model) to match the patient's physiological signal with a label of the patient's outcome (survival or death). The patient's physiological signal may include cardiac parameters such as aortic pressure, differential pressure, left ventricular pressure, or any suitable signal derived from physiological measurements from the patient. Features are extracted from the signal and used in the classification model. Suitable features may include statistics of the physiological signal over a period of time, such as, for example, the mean and standard deviation of the raw signal. The classification model is used to classify patients as high or low risk, and the classification model can be used to track the patient's condition over time. The classification model can be a logistic regression method, a deep learning method, a decision tree, a random forest method, a naive Bayes method, a support vector machine, or any suitable model. The methods and systems described herein use this model to predict the patient's health status (e.g., probability of survival) using a previous window of the signal to represent the patient's condition in real time. Such systems and methods allow a user (e.g., a clinician, caregiver, etc.) to track a patient's health status or health index and see changes in predicted outcomes for the patient so that "at-risk" patients (e.g., patients with declining health indexes) can receive closer attention. For example, the health status may be displayed on an interface that is connected to or part of the heart pump system, such as an automatic impeller controller (AIC). Raw features may be extracted from the signals of the heart pump system. Such features may include the mean or standard deviation of raw signals such as aortic pressure, differential pressure, and left ventricular pressure. Feature engineering may be used to find trends in the signals, find sudden changes in the signals, and generate signals such as contractility from these raw signals.

[0040] FIG. 1 shows a flow diagram of a method 100 for monitoring a patient's condition. The method shown in FIG. 1 may determine a cardiac health index as described above. At step 102, the system acquires training log ("X") data over a period of time. The period may be one hour, two hours, one day, or any suitable length of time (as shown in FIG. 1). The log data X may correspond to a most recent period of multiple periods of patient condition monitoring. At step 104, the system acquires training patient outcome ("Y") data. The patient outcome data associates a patient with survival or death. The patient outcome data may be associated with the period of log data such that the patient outcome data is indicative of the patient's condition at the end of the period of log data. Alternatively, the patient outcome data may not be associated with the period of log data, but instead is indicative of the patient's condition at a point after the end of the period. The log data X and the patient outcome data Y are acquired from a number N of patients, where N is large enough to adequately train a model for accurate prediction. The log data may be measured and / or aggregated by one or more cardiac pump systems. The heart pump system may be at least partially inserted into the heart of N patients. For example, the heart pump system may extend through the patient's aorta to the patient's left ventricle. The one or more heart pump systems may be the same or different types of heart pump systems. A heart pump system compatible with the present disclosure is disclosed in U.S. Patent Application Serial No. 15 / 709,080 to Edelman et al. (U.S. Patent Application Publication No. 2018 / 0078159 A1, published March 22, 2018), the entire contents of which are incorporated herein by reference. In general, a heart pump system or any other system for acquiring physiological data from a patient may be used in conjunction with the present disclosure.

[0041] In step 106, the system builds a classification model, which may be a machine learning model. The model is trained with training log data X and patient outcome data Y. The model may be stored in a database and may include mathematical rules for classification of features using a learning method. The learning method may be logistic regression, decision trees, deep learning, naive Bayes, or any other suitable technique.

[0042] For example, logistic regression is based on an equation that is used to represent a predictive model with coefficients learned from training data. The representation of the model may be stored in a database as an example of coefficients, each corresponding to a weight indicating the relative importance of a particular feature, and can be used to calculate probabilities, such as the probability of a patient's survival. The probability of death is calculated as (1+exp(-x)) -1 where x is equal to α*Feature_α+β*Feature_β+γ*Feature_γ+... for any number of features and associated coefficients.

[0043] In another example, decision tree learning uses decision trees as predictive models that go from observations about an item to a conclusion about a target value for that item. Tree depth can be a hyperparameter in decision tree learning. Hyperparameters are values ​​that cannot be estimated from the data used in the model. Hyperparameters are often used to aid in the estimation of model parameters and can be tuned for a given predictive modeling problem. Accuracy can be used as a performance metric for the predictive model. By determining the maximum accuracy of a decision tree through tuning of hyperparameters such as tree depth, the system can provide an optimized machine learning model (such as machine learning model 106) and therefore better provide predictions (such as patient outcomes in step 110, described below). Receiver Operating Characteristic (ROC) and Area Under Curve (AUC) can also be used as metrics to compare predictive algorithms. In some implementations, steps 102, 104, 106, and any combination thereof are optional. For example, the method may start with step 108, described below. In some implementations, the classification model is periodically updated with new patient information. In some embodiments, the classification model is collated, developed, or executed by a system separate from the heart pump system. For example, a third party system may collate data from multiple different heart pumps and build a machine learning model. The machine learning model may be used to enable steps 108-112 in some examples.

[0044] At step 108, the system obtains new log data for the particular patient over the period. The particular patient may be one of the N patients from which the new log data was received, or may be a new patient not included in the N patients. At step 110, the new log data is input into the model trained at step 106 to predict a patient outcome for the particular patient. The patient outcome may be a binary value representing survival or death.

[0045] At step 112, the model, along with the new log data, is used to predict the patient's condition as a health index over time. In some embodiments, the health index is displayed for patient monitoring. For example, the health index may be displayed on the heart pump system or viewed via a computer system, a mobile device, a tablet, or any other suitable device. The health index over time is found via a sliding window process. A health index is calculated for a particular patient at a first one of a plurality of times over a time window. A health index is then calculated for a particular patient at a second one of a plurality of times over that time window. For example, at 2:00 pm, the system may calculate a health index for patient W using the past hour's log data (1:00 pm to 2:00 pm). At 2:15 pm, the system may again calculate a health index for patient W using the past hour's log data (1:15 pm to 2:15 pm). Thus, the window (one hour period) is "slid" in time in 15 minute increments to provide an updated time-varying patient health index. The time between calculations (15 minutes in the above example) may be any suitable time increment, such as 1 hour, 30 minutes, 1 minute, 20 seconds, etc. The health index may be a cardiac health index that indicates the cardiac health or survival probability of a particular patient. In some embodiments, the health index is graphed over time and displayed to the clinician so that the clinician can see trends in the health index over time.

[0046] In some embodiments, the health index includes a cardiac component and a systemic perfusion component. The health index may be indicative of overall patient recovery and survival probability (i.e., patient outcome). The cardiac component may include unloading, contractility, or any suitable indicator of the performance of the patient's heart. The system perfusion component may include cardiac output (CO), aortic pressure mean (AoPm), or any suitable indicator of the performance of the patient's circulatory system.

[0047] FIG. 2 shows a flow diagram of a method 200 for predicting a patient's outcome in a manner similar to that described in connection with FIG. 1. Steps 202, 204, 206, and 208 are identical to steps 102, 104, 106, and 108 of FIG. 1, respectively. Steps 202, 204, and 206 are optional, and any combination of steps 202, 204, and 206 may be omitted from the method described herein. For example, the model may be already formulated or may be imported from an external system. At step 210, the model is used together with new log data to predict the patient's condition as a health index over time. In some implementations, the health index is displayed for patient monitoring. The health index may be a cardiac health index indicating the cardiac health or survival probability of a particular patient. At step 212, the health index is used to predict the patient's outcome. The patient's outcome may be a binary value representing survival or death. For example, the health index may represent the patient's probability of survival as a value x, where x is between 0 and 1 (boundary values ​​included). The health index may be used to determine a binary output. For example, if the health index is greater than 0.5 (or any other suitable threshold), the health index may indicate a patient outcome of survival, corresponding to a binary value of 1. The patient outcome may be displayed to a clinician. In FIG. 1, the system predicts a patient outcome at step 110 and then provides patient status monitoring over time at step 112 using a sliding window. In contrast, in FIG. 2, the system provides patient health status monitoring over time at step 210 using a machine learning model and log data. The system then provides a prediction of the patient's outcome at step 212 using the last value (health index) calculated during the patient status monitoring.

[0048] The data used as the log data X or the patient outcome data Y described in connection with FIG. 1 and FIG. 2 may be stored in a database system. The database system may include multiple databases or a single database. For example, the database may include a clinical data database, a device registry, and an AIC log. The database system may include data of over thousands of cases. Each case may correspond to a separate patient or may correspond to the implantation of a heart pump system. The data may represent different case times or may be from a period specific to each database in the database system. The features described by the data stored in the database system (e.g., the AIC log) may include pump type, pressure signal, P level, flow rate of the heart pump system, impeller flow rate, motor current, alarms, outcomes, or any other suitable features. The P level is the performance level of the heart pump system and is associated with the flow control of the system. As the P level increases, the flow rate and revolutions per minute associated with the heart pump system increases. The data stored in the database system and the data storage included in the database system may be used to train the models described herein.

[0049] The predictive modeling system and predictive modeling methods described herein use machine learning to make predictions on a number of features according to data science techniques. A data science project begins with data input into the system. The data is preprocessed and feature engineered. Preprocessing challenges may arise when processing the log data 102 and patient outcome data 104 of FIG. 1 prior to use in training the machine learning model 106. Problems may include too many missing values ​​and unreliable data. Unreliable data may include incorrect or incomplete data. An example of unreliable data is when a procedure outcome is listed as "death" in a database system, but the actual outcome at the end of intensive care unit (ICU) support is listed as "survival." Incorrect data, incomplete data, and a high rate of missing data may each affect the performance of a predictive model. "Bad" data may be labeled as "unreliable", "too many missing values" data, or with any other type of label indicating that the data is unreliable and needs to be removed or supplemented from the training dataset to provide a better prediction of patient outcomes and thus improve patient health monitoring. Other examples of preprocessing may include reformatting the data, removing unusable features, handling outliers, supplementing missing values, encoding categorical features, scaling, and any suitable measures to resolve data issues. Once raw data, such as the log data 102 and patient outcome data 104 of FIG. 1, has been preprocessed, feature data may be extracted. Examples of feature data are shown in Table 1.

[0050] [Table 1]

[0051] In some implementations, the feature data is split into training data and cross-validation data and used to build a machine learning method. The machine learning method can be applied to the new unlabeled data to make predictions about the new data, such as predicting a health state of a patient associated with the new data. This prediction, along with the feature data, can be further analyzed for visualization.

[0052] The systems and methods described in connection with Figures 1 and 2, as well as other processes described herein, may use some or all of the data held in the databases described above. For example, machine learning methods may be trained using only the data stored in the AIC log.

[0053] FIG. 3 illustrates a data flow diagram 300 for AIC log data, e.g., held in an AIC log, as described above in connection with the database system. Long term (LT) log data 302 includes, e.g., a position signal. LT may correspond to one sample per minute or any other suitable sampling rate. Real time (RT) log data 304 uses information sampled at a higher sampling rate than that corresponding to the LT data to generate features, such as heart rate. RT may correspond to 25 Hz (25 samples per second) or any other suitable sampling rate. The combination of LT and RT log data is processed as raw and generated features, which are then input as X to log data mining 306. IQ database 310 includes patient outcomes, which are also input as Y to log data mining 306. Log data mining 306 may be used to extract relevant and / or important features for use in training a machine learning model, such as machine learning model 106 of FIG. 1, and thus provide more efficient patient outcome predictions.

[0054] First data related to cardiac or cardiac pump parameters and / or second data related to physiological parameters, such as the first and second data described above, may be used to predict patient outcomes. In some embodiments, the first and / or second data of a plurality of patients are used to build a predictive modeling system as described above in connection with Figs. 1-3. The first and / or second data may be included in a modeling training set or may be for feature extraction as described above in connection with Fig. 3. The methods and systems described herein may test for significant separation between two groups of data corresponding to patient outcomes as a function of one or more patient parameters. For example, Figs. 4-7, described below, show two different patient parameters graphed against each other to determine whether the parameters show separation along with patient outcomes. Although only a few examples are described herein, such testing may be performed for any number of patient parameters of the first and / or second data. As shown in Figs. 4-5, some data shown herein show significant separation. Pairs of parameters that show significant separation (e.g., represented by boundaries 412, 422, 532) may be more predictive of patient outcome than other parameters. Other parameters that show less or no significant separation (e.g., shown in Figures 6-7, described below) may be less predictive of patient outcome than other parameters.

[0055] 4 and 5 show exemplary feature plots used for training and classification for features suitable for decision boundary separation. In graphs 410, 420, and 530, open dots represent survival cases and filled dots represent death cases. In each graph, the machine learning results are represented by a linear decision boundary that is used to separate the filled dots from the open dots with the best tradeoff using logistic regression. The dots collectively represent patient outcome training data, such as data Y, described above in connection with step 104 of FIG. 1. The placement of the dots in graphs 410, 420, and 530 is determined by the associated log data training, such as data X, described above in connection with step 102 of FIG. 1.

[0056] The graph 410 represents a calculated boundary 412 of the mean placement signal (PS), which may also be referred to as the mean placement level. The placement signal may be the aortic pressure in the Impella CP / 2.5 case or the differential pressure in the Impella 5.0 / LD case. The x-axis of the graph 410 represents the mean value of the mean placement signal (PS_mean), and the y-axis represents the standard deviation of the PS_mean. For example, a hollow dot may represent a patient who survived (from the patient outcome training data 104), who is also associated with a set of mean placement level data (from the training log data 102). The system may calculate the mean placement level and standard deviation of the mean placement level for that patient and graph the associated hollow dot accordingly. Once the dots are graphed for the patients in the training data, a machine learning model (such as the machine learning model 106 of FIG. 1 ) calculates a linear decision boundary 412 according to the mean placement level and standard deviation of the placement level for the mean of the dots. A linear decision boundary may be represented by a set of coefficients tied to the training data, as described above. When the system receives new data related to a new patient (e.g., new log data 108 of FIG. 1), the system may determine the mean alignment of the new patient and the standard deviation of the mean alignment. Depending on where these values ​​"place" the patient's dot on the graph 410, a predicted patient outcome may be determined based on the location of the dot relative to the decision boundary 412. For example, if a patient has a mean alignment level of 100 and a standard deviation of the mean alignment level of 30, then according to the graph 410, the predicted patient outcome will be survival. This is because the patient's dot will fall to the right of the boundary 410 and is therefore more strongly associated with surviving patients (open dots). Such a calculation (where the new patient falls relative to the decision boundary) may constitute a prediction of the patient's outcome, as described above in relation to FIG. 1. Thus, the decision boundary may represent a threshold for predicting the patient's outcome. The decision boundary may be calculated differently in different machine learning instances. For example, the decision boundary 412 may shift to the right or left, may have a different slope, or may be non-linear.

[0057] Similarly, graph 420 represents calculated bounds 422 of the mean differential pressure. The x-axis of graph 420 represents the average value of the mean differential pressure, and the y-axis represents the standard deviation of the mean differential pressure.

[0058] Graph 530 represents calculated bounds 532 for maximum left ventricular pressure (LVP). The x-axis of graph 530 represents the mean maximum LVP and the y-axis represents the standard deviation of maximum LVP.

[0059] 6 and 7 show two example feature plots that are not suitable for decision boundary separation. In graphs 610, 720, open dots represent survival cases and filled dots represent death cases. Without boundary separation, the information distributions shown by graphs 610, 720 may be less useful in predicting patient outcomes than those shown in FIGS. 4 and 5, since the system is not provided with clear boundaries (or equations with a set of coefficients) to classify new patient data. The x-axis of graph 610 represents the slope of the linear regression of PS_mean over time, and the y-axis represents the coefficient of determination (r-squared or r 2 The data represented in graph 610 is non-separable because the solid and hollow dots have the same pattern. The x-axis of graph 720 represents Slp(PS_mean) and the y-axis represents the coefficient of determination of the linear regression of PS_mean. The data represented in graph 720 is quasi-separable because the solid and hollow dots have different patterns but may be "too weak" to be separated.

[0060] FIG. 8 shows an exemplary bar graph ranking the importance of features as an exemplary result of the machine learning methods described herein. The importance of features may correspond to, for example, coefficients of a logistic regression model used as the machine learning model 106 of FIG. 1 to predict patient outcomes. A higher coefficient in the model may correlate to a higher importance of the feature to overall patient health. Knowing the importance of features may be particularly helpful to clinicians when determining treatment methods in response to a decline in a patient's cardiac health (as may be indicated by the patient status monitoring 112 of FIG. 1). In general, features may include aortic pressure, differential pressure, motor current, left ventricular pressure, end diastolic pressure, aortic pulse pressure, intrinsic cardiac output, cardiac output, CPO, configuration, flow, P level, contractility, and relaxation. These features may be processed to determine further features such as mean configuration, standard deviation of configuration, mean configuration range, standard deviation of configuration range, mean differential pressure, standard deviation of differential pressure, mean differential pressure range, standard deviation of differential pressure range, left ventricular pressure maximum and left ventricular pressure minimum. The importance of these features can be determined by ranking the features using different calculations.

[0061] Graph 810 shows feature ranking using F-1. F-1 is a statistical term defined as 2*Precision*Recall / (Precision+Recall). Precision is equal to TP / (TP+FP), and recall is equal to TP / (TP+FN), where TP stands for true positive, FP stands for false positive, and FN stands for false negative. Graph 820 shows feature ranking using precision. Graph 830 shows feature ranking using recall. In all three graphs 810, 820, 830, the most important feature (the feature with the highest importance) is the average LVP maximum level, suggesting that this feature is beneficial in understanding a person's health condition compared to the other features shown in FIG. 8. In graphs 810 and 830, the average mean alignment level is ranked as the second most important feature. However, in graph 820, the average mean alignment level is ranked third. The differences in feature rankings between graphs 810, 820, and 830 illustrate that different metrics used to calculate feature importance (such as precision, recall, or F-1 score) can affect the outcome of which features are deemed most important and given the greatest weight in the model.

[0062] Displaying a visual representation of the feature importance may be helpful to clinicians. A graphical representation may allow clinicians to interpret the feature importance more quickly or easily compared to a numerical representation. Specifically, the feature importance may be represented by a bar graph as shown in FIG. 8 or a spider plot as shown in FIG. 9. In some embodiments, the visual representation shows the patient's health status and / or the feature importance at a single time point, or an average value over multiple time points. In some embodiments, the visual representation is updated periodically, at regular intervals, or in real time, such that the visual representation appears to a viewer as a video stream.

[0063] FIG. 9 shows an exemplary spider plot 900 illustrating relative feature scores for a patient. The patient's cardiac performance index 902 is 0.58 in this case, which is an example of a cardiac health index, as described above. The CPO 904 associated with the patient is 0.75. CPO is a function of mean arterial pressure (MAP) and CO. CPO may be used as a predictor of patient outcome and may be a component of a cardiac health index. FIG. 9 shows the patient's features and scores at a single point in time when the CPO was 0.75 and the cardiac performance index was 0.58. These values ​​may be updated over time. In one example, CPO may be a time-varying feature used in calculating the patient's likelihood of survival.

[0064] The spider plot 900 visually displays the relative impact of the five features on the patient's health. Each feature is given a score on a scale of 1 to 5 that represents the patient's status of that feature. In some embodiments, the scores are on another scale such as 0 to 1, 1 to 10, 1 to 50, 1 to 100, 1 to 1000, or any other suitable scale. A score of 1 for left ventricular (LV) contractility indicates a dP / dt max value (which may be a ventricular contractility assessment) of greater than 200 mmHg / sec, a score of 2 indicates a value greater than 400 mmHg / sec, a score of 3 indicates a value greater than 600 mmHg / sec, a score of 4 indicates a value greater than 600 mmHg / sec, and a score of 5 indicates a value greater than 1000 mmHg / sec. A rating of 1 for LVEDP indicates a deviation of 20 mmHg, a rating of 2 indicates a deviation of 15 mmHg, a rating of 3 indicates a deviation of 10 mmHg, a rating of 4 indicates a deviation of 5 mmHg, and a rating of 5 indicates an LVEDP within the target range of 10-15 mmHg, with deviations being measured as deviations from this target range. A rating of 1 for LV relaxation indicates a dP / dt max of less than 1000 mmHg / sec, a rating of 2 indicates less than 800 mmHg / sec, a rating of 3 indicates less than 600 mmHg / sec, a rating of 4 indicates less than 400 mmHg / sec, and a rating of 5 indicates less than 200 mmHg / sec. A rating of 1 for AoPm indicates 60 mmHg, a rating of 2 indicates 70 mmHg, a rating of 3 indicates 80 mmHg, a rating of 4 indicates 90 mmHg, and a rating of 5 indicates 100 mmHg. A rating of 1 for CO indicates 2 L / min CO, a rating of 2 indicates 3 L / min CO, a rating of 3 indicates 4 L / min CO, a rating of 4 indicates 5 L / min CO, and a rating of 5 indicates 6 L / min CO, where CO measurements are a function of heart rate and stroke volume. For example, LV relaxation 906 is 5, LVEDP 908 is 5, LV contractility 910 is 3, CO 912 is 3, and AoPm 914 is 2. In this example, the AoPm is low relative to the other features, and therefore, this patient's AoPm is poor relative to the other features of this patient. In this way, by displaying the feature data on a uniform rating scale across all features, clinicians can quickly review patient data and understand which features may need to be addressed to improve the patient's overall health.In this example, a clinician may look at spider plot 904 and decide to address the patient's AoPm first. After addressing the patient's AoPm by clinical means, the clinician may then follow the patient's progress by observing the patient's cardiac health status, as updated at that time, on spider plot 904 via patient status monitoring (step 112 of FIG. 1 ).

[0065] The features displayed in the spider plot 904 may be weighted by their relative feature importance. For example, CO may have a weight of 0.4, LV contractility may have a weight of 0.2, LVEDP may have a weight of 0.3, AoPm may have a score of 0.2, and LV relaxation may have a weight of 0.1. In this example, AoPm may still have the lowest unweighted score, but CO may have the lowest weighted score due to its relative importance and low score. In another example, each feature may be weighted equally.

[0066] FIG. 10 shows an example of characteristics of a particular patient, Patient X. Graph 1010 shows a decision boundary 1012. As described above, open dots represent patient survival and filled dots represent patient death. Dot 1014 represents Patient X who survived. Thus, dot 1014 is located to the right of decision boundary 1012 as described above. Decision boundary 1012 may represent a line between whether a patient is more likely to survive or more likely to die. For example, patients to the right of boundary 1012 are more likely to survive and patients to the left of boundary 1012 are more likely to die. The x-axis of graph 1010 represents the mean value of the mean position signal (PS) and the y-axis represents the standard deviation of the mean PS. FIG. 10 shows an example of a user interface for a clinician where the clinician interprets the patient's cardiac health by looking at the position signal, motor current, P-levels, and likelihood of survival. Specifically, the graph 1010 allows the clinician to see a graphical representation of the likelihood of a patient's survival in relation to a boundary between the standard deviation and the mean of the alignment signal. For example, the graph 1010 allows the clinician to visualize the distance between the boundary 1012 and the data representing the patient 1014. The distance from the boundary or separation line may indicate to the clinician how likely the patient is to survive. For example, patients farther to the right of the decision boundary 1012 (e.g., patients with a higher mean alignment signal) are more likely to survive than patients closer to the decision boundary (e.g., patients with a lower mean alignment signal).

[0067] Graph 1020 shows a positioning signal 1022. The y-axis of graph 1020 represents pressure in mmHg. Graph 1030 shows a motor current signal 1032. The y-axis of graph 1030 represents motor current in mA. Graph 1040 shows a P-level 1142. The y-axis of graph 1040 represents P-level. The x-axis of graphs 1020, 1030, and 1040 represent time. Graphs 1020, 1030, and 1040 are shown for the same time period, on the same time scale, and for the same patient, patient X, represented by dot 1014. The positioning signal 1022, motor current signal 1032, and P-level 1042 may represent first data related to a time-varying parameter of the cardiac pump system, as described above. The positioning signal 1022, the motor current signal 1032, and the P level 1042 may be some of the features used to determine a patient's cardiac health index. In some aspects, the cardiac health index may be a patient's survival probability, which may be used to predict a patient's outcome. In this case, the patient's outcome (represented at 1014) is survival. As shown in graph 1040, the P level is gradually decreased over time. In one example, a clinician may review the likelihood of patient X's survival to determine when and how much to change the P level. In the example shown in FIG. 10, the P level is gradually decreased, thereby decreasing the speed of the pump.

[0068] 11 shows another example for a particular patient, Patient Y. Graph 1110 shows a decision boundary 1112. As mentioned above, open dots represent surviving patients and solid dots represent deceased patients. Dot 1114 represents Patient Y who has died. Dot 1114 is located across boundary line 1112 from the majority of open "surviving" dots. The x-axis of graph 1110 represents the mean value of the mean PS and the y-axis represents the standard deviation of the mean PS.

[0069] Graph 1120 shows a positioning signal 1122. The y-axis of graph 1120 represents pressure in mmHg. Graph 1130 shows a motor current signal 1132. The y-axis of graph 1130 represents motor current in mA. Graph 1140 shows a P-level 1142. The y-axis of graph 1140 represents P-level. The x-axis of graphs 1120, 1130, and 1140 represent time. Graphs 1120, 1130, and 1140 are shown for the same time period, with the same time scale, and for the same patient, patient Y, represented by dot 1122. The positioning signal 1122, motor current signal 1132, and P-level 1142 may represent first data related to a time-varying parameter of the cardiac pump system, as described above. The positioning signal 1122, the motor current signal 1132, and the P level 1142 may be some of the features used to determine a cardiac health index for the patient. In some aspects, the cardiac health index may be a patient's probability of survival, which may be used to predict a patient's outcome. In this case, the patient's outcome (represented by the dot 1114) is death. The graph 1140 shows the variation of the target P level, the minimum P level, and the maximum P level over time. In some implementations, the clinician varies the target P level to increase or decrease pump operation to treat the patient. For example, the clinician may increase or decrease the target P level if the patient's probability of survival decreases. As an example, the graph 1140 shows the target P level at which the cardiac pump system is set to treat the patient.

[0070] The difference between Figures 10 and 11 shows how measured characteristics representative of a patient, such as placement signals, motor currents, and P levels, can be correlated to survival of a patient's death. For example, patient X (represented in Figure 10) survived and patient Y (represented in Figure 11) died. The systems and methods described herein predicted this outcome based on the boundary 1412 calculated by the training data (log data 102 and patient outcomes 104 in Figure 1), the standard deviation of PS_mean and the mean PS_mean (new logged data 108 in Figure 1) combined with the boundary 1112 (machine learning model 106 in Figure 1).

[0071] FIG. 12 illustrates a set of four graphs 1200 depicting patient-related characteristics over time. Three graphs 1210, 1220, 1230 depict measured patient-related characteristics over the same time frame. Graph 1210 depicts pump pressure maximum 1212, pump pressure average 1214, pump pressure minimum 1216, differential pressure maximum 1219, and differential pressure minimum 1218. The y-axis of graph 1210 represents millimeters of mercury (mmHg) and the x-axis represents time. Graph 1220 depicts motor current maximum 1222, average 1224, and minimum 1226 of a cardiac pump system disposed within a patient. The y-axis of graph 1220 represents milliamps (mA) and the x-axis represents time on the same scale and over the same period as the x-axis of graph 1210. Graph 1230 depicts motor speed (MS) average 1232. The y-axis of graph 1230 represents revolutions per minute (rpm) and the x-axis represents time on the same scale and over the same time period as the x-axis of graphs 1210 and 1220. Graph 1240 shows a calculated measure of the patient's cardiac health expressed as a survival probability 1242. The y-axis of graph 1230 represents the percentage of the patient's survival probability (defined as the health state) and the x-axis represents the same time period as described above in connection with graphs 1210, 1220, and 1230.

[0072] The pump pressure, differential pressure, motor current, and motor speed may represent first data related to time-varying parameters of the cardiac pump system, as described above. The pump pressure maximum 1212, pump pressure average 1214, pump pressure minimum 1216, differential pressure maximum 1219, differential pressure minimum 1218, motor current maximum 1222, motor current average 1224, motor current minimum 1226, and motor speed average 1232 may be some of the features used to determine a patient's cardiac health index. In some aspects, the cardiac health index may be a patient's survival probability 1242, which may be used to predict a patient's outcome. The patient's outcome prediction may change over time.

[0073] The survival probability 1242 may be calculated by the methods and systems described above. The values ​​shown in the graphs 1210, 1220, 1230 may be obtained from a cardiac pump system. Instead of predicting the survival probability (H-index) using only the most recent hour, the survival probability may be calculated over a sliding window from the beginning to the end of the patient's case. Such a process allows the system to monitor the patient's health. The survival probability 1242 is determined, at least in part, by the pump pressure, motor current, and motor speed in this example. For example, the pump pressure, motor current, and motor speed all drop in value shortly after time marker 1000. At the same time, there is a corresponding drop in the survival probability 1242. Such a drop may indicate a decline in the patient's cardiac health. By predicting a decline in cardiac health, the system may alert a clinician to a patient's health problem before it can normally be determined by the clinician, thereby providing the clinician more time to treat the patient.

[0074] The results of the predictive modeling system are shown in Table 1. The model was tested on 13 shock cases. The shock case data, in this example, was provided by Henry Ford Hospital as a "third-party" test data set. The data for the 13 cases are shown in Table 2.

[0075] [Table 2]

[0076] A separate, unrelated test dataset (a subset of the data held in the database system described above) was tested. As shown in Table 3, the results of this test yielded an accuracy of 81.4%, with N (the number of patient data used in the test) being 80.

[0077] [Table 3]

[0078] FIG. 13 illustrates the characteristics of an exemplary case over time with a set of graphs 2000. Three graphs 1310, 1320, 1330 illustrate characteristics related to a patient over time. The positioned signal graph 1310 has a y-axis showing pressure in millimeters of mercury (mmHg) and an x-axis showing time. The flow rate graph 1320 has a y-axis showing flow rate in liters per minute (L / min) and an x-axis showing time. The health index graph 1330 has a y-axis showing cardiac health index and an x-axis showing time. The time scales of the graphs 1310, 1320, and 1330 are the same, and the graphs 1310, 1320, and 1330 are positioned such that the x-axes of the three graphs are aligned.

[0079] The positioning signal and the flow rate may represent first data related to time-varying parameters of the cardiac pump system, as described above. The positioning signal (shown in graph 1310), the average flow rate (shown in graph 1320) and the target flow rate may be some features of a plurality of features used to determine a patient's cardiac health index. The index shown in graph 1330 may be a percentage of the patient's probability of survival, which may be used to predict the patient's outcome. The patient's outcome prediction may change over time.

[0080] In the example shown in FIG. 13, Patient Z was found unresponsive at home. Patient Z's spouse administered CPR until emergency services arrived. Patient Z was placed with a heart pump system for 34 hours of support. The x-axis of graphs 1310, 1320, and 1330 begins at approximately the time of placement of the heart pump system. The average flow rate 1322 of the heart system was approximately 3 L / min at P-7, which was good. Marker 1340 represents a first time point, specifically, 5:30 AM on March 12, 2016. At marker 2040, Patient Z was hemodynamically stable. Marker 1350 represents a second time point, specifically, 5:45 PM on March 12, 2016. At marker 1350, Patient Z became pale. Patient Z's oxygen saturations (O2 sats) dropped and Patient Z went into ventricular tachycardia or ventricular fibrillation (VT / VF). Physicians were unable to resuscitate Patient Z.

[0081] The time between marker 1340 and marker 1350 shows a significant disruption in the cardiac system's average flow rate 1322 and the positioning signal 1312. As can be seen from graph 1330, the significant disruption in the positioning signal 1312 and the average flow rate 1322 would result in a dramatic change in patient Z's cardiac health index 1332 between marker 1340 and marker 1350. A clinician could review the cardiac health index and determine that the patient is at risk. Prior to marker 1340, graphs 1310 and 1320 are relatively constant (compared to the high fluctuations shown between marker 1340 and marker 1350). However, graph 1330 shows a gradual but steady decline in the patient's health prior to marker 1340. A clinician could review the decline in the cardiac health index and determine that the patient's health is deteriorating. By identifying a decline in the cardiac health index, a clinician could intervene before the time represented by marker 1340 (i.e., before the patient's flow and positioning signals show significant disturbances). In some cases, early intervention can be highly beneficial to the patient's health and is a determining factor in the patient's survival. In some instances, if the cardiac health index is declining (e.g., as shown before marker 1340 in graph 1330), a clinician may be alerted to the patient's declining health so that the clinician may intervene in the patient's treatment. The cardiac health index graph 1330 may also be used in post-mortem case analysis after the patient has died.

[0082] FIG. 14 shows a flow diagram for determining a patient's survival probability. At step 1400, an intravascular heart pump system is inserted into the patient's vasculature. The heart pump system includes a cannula, a pump inlet, a pump outlet, and a rotor. In some embodiments, the heart pump system is a left ventricular assist device (LVAD). In some embodiments, the heart pump system is a right ventricular assist device (RVAD). In some embodiments, the cannula, the pump inlet, the pump outlet, and the rotor are optional. At step 1402, the heart pump system is positioned within the patient such that the cannula extends past the patient's aortic valve, the pump inlet is located within the patient's left ventricle, and the pump outlet is located within the patient's aorta. In some embodiments, step 1402 is optional and the heart pump system is only partially positioned within the patient. For example, the heart pump system may be inserted by catheterization through the femoral artery, into the ascending aorta, past the aortic valve, into the left ventricle, or through the femoral vein into the right atrium.

[0083] At step 1403, first data is obtained. The first data is related to a time-varying parameter of the heart pump system. The first data may represent continuous or near-continuous measurements obtained through the heart pump system or may represent known quantities, such as inputs to the heart pump system. The first data is related to the operation of the heart pump system or factors measured by the heart pump system, i.e., the first data would not be known without the heart pump system. The first data may include data indicative of heart rate, pump pressure, differential pressure, motor current, P-level, and / or motor speed. From these measurements, important information regarding cardiac function and, in some cases, information regarding the performance of the cardiac assist device, including the occurrence of aspiration events, can be determined. As described below in connection with step 1408, this information regarding cardiac function can be used to predict the patient's probability of survival.

[0084] In some embodiments, one or more sensors on the heart pump system acquire the first data. In some embodiments, the first data is acquired via an external system. In some embodiments, one or more sensors on the heart pump system acquire the first data. The one or more sensors on the heart pump system may be positioned within the patient's heart, external to the patient's heart, or a combination of both during operation of the heart pump system. For example, the sensors on the heart pump system may measure pressure within the patient's vasculature. The measured pressure may be used in the calculation of additional parameters, such as cardiac power output, as described above.

[0085] A plurality of features are extracted from the first data at step 1404. Extracting the features may include processing the first data at the heart pump system or at an external device. The multiple features may include left ventricular end diastolic pressure (LVEDP), stroke volume, ejection fraction, chamber dilation, chamber hypertrophy, chamber pressure, stroke work, preload state, afterload state, heart rate, cardiac recovery, aortic pressure, differential pressure, motor current, motor speed, pump pressure, left ventricular pressure, end diastolic pressure, aortic pulse pressure, intrinsic cardiac output, cardiac output, CPO, configuration, mean flow, target flow, P level, contractility, relaxation, configuration signal, mean configuration, standard deviation of configuration, mean configuration range, standard deviation of configuration range, mean differential pressure, standard deviation of differential pressure, mean differential pressure range, standard deviation of differential pressure range, left ventricular pressure maximum, left ventricular pressure minimum, pump pressure maximum, pump pressure average, pump pressure minimum, differential pressure maximum, differential pressure minimum, motor current maximum, motor current minimum, motor current average, motor speed average, any other suitable feature, and any combination thereof.

[0086] In some implementations, the first data is obtained during a first time period during which the heart pump system is operating, such as one second, one minute, five minutes, ten minutes, one hour, several hours, one day, several days, one week, one month, or any suitable time frame. The representative, average, and minimum values ​​of the features described above may be the representative, average, or minimum values ​​of the features during the first time period.

[0087] At step 1408, a survival probability of the patient is determined. The survival probability is a value indicating the likelihood of the patient's survival or death. In some embodiments, the survival probability is a number, e.g., between 0 and 1. In some embodiments, if the probability is equal to or greater than a threshold value (e.g., 0.5), the survival probability indicates survival (e.g., the patient has a greater than 50% chance of living given the patient's cardiac health). The survival probability is based on the plurality of features extracted at step 1404, as described above, and is determined using a predictive model. In some embodiments, the predictive model is a machine learning model. For example, the predictive model can be one of a logistic regression method, a deep learning method, a decision tree, a random forest method, a naive Bayes method, and a support vector machine method, or any other suitable model.

[0088] At step 1410, the cardiac pump system is operated to treat the patient. In some embodiments, step 1410 is optional. In some embodiments, the cardiac pump system may be operated to provide a constant or near-constant level of assistance to the patient. In some embodiments, the pump operation is altered based on a predictor of the patient's outcome. In particular, altering the pump operation may include adjusting an operating parameter of the cardiac pump system to provide a different level of assistance than was provided during the period during which the first data was acquired. In one example, adjusting an operating parameter of the cardiac pump system includes adjusting a pump speed (e.g., by increasing or decreasing) based on changes in cardiac power output, lactate concentration, or both. It may be desirable to increase the pump speed when one feature value falls below a first threshold and when a second feature value exceeds a second threshold, or both, as described below in connection with FIG. 15.

[0089] 15 shows a flow diagram for determining a predicted value of a patient's outcome based on CPO and lactate concentration. Steps 1500 and 1502 are the same as steps 1400 and 1402 described above. At step 1504, the heart pump system operates for a first time period to provide a first level of cardiac assistance to the patient. Examples of providing a level of cardiac assistance include operating the pump at a P level or motor speed, providing current to the pump motor, turning on the pump, directing flow through the pump or the patient's heart, or any other suitable method of assistance. The level at which to operate the heart pump system may be provided by the controller, for example, via a user command entered via a user interface.

[0090] At least one CPO value is obtained in step 1506. The at least one CPO value is derived from measurements provided by the cardiac pump system. Optionally, the at least one CPO value represents the CPO at at least one time point within a first time period described above in connection with step 1504. In some implementations, the CPO is regularly updated at a regular time interval following the first time period. For example, the first time interval may be 0.01 seconds, 0.1 seconds, 0.5 seconds, 1 second, 5 seconds, 10 seconds, 1 minute, 10 minutes, 15 minutes, 30 minutes, 1 hour, or any suitable time interval. At least one lactate concentration value is also obtained in step 1506. The at least one lactate concentration value may be measured from the patient. For example, the lactate concentration value may be manually entered into the cardiac pump system or retrieved from an external database.

[0091] At step 1508, a predicted outcome for the patient is determined. The predicted value is based at least in part on the at least one cardiac power output value and the at least one lactate concentration value obtained at step 1506. It may be desirable to increase the pump speed when the at least one CPO value falls below a first threshold, when the at least one lactate concentration value exceeds a second threshold, or both. For example, the first threshold may be a value of 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1.0 W, etc., and the second threshold may be a value of 1, 2, 3, 4, 5, 6, 7 mmol / L, etc. A low CPO value and a high lactate value may indicate a relatively low probability of survival for the patient. Due to the poor physical condition of the patient, the clinician may attempt to increase the level of cardiac assistance provided by the pump, for example, by increasing the pump speed. It may be desirable to reduce or not change the pump speed when at least one CPO value is above a first threshold, when at least one lactate concentration value is below a second threshold, or both. A high CPO value and a low lactate value may indicate a relatively high probability of survival for the patient. Because the patient is in good physical condition, the clinician may decide not to change the parameters of the pump's operation. Alternatively, the clinician may attempt to reduce the level of cardiac assistance provided by the pump or turn the pump off entirely. Step 1510 is the same as step 1410 described above in connection with FIG. 14.

[0092] The foregoing is merely illustrative of the principles of the present disclosure, and the device may be practiced in other than the described manner, which is presented for purposes of illustration and not limitation. It should be understood that while the device disclosed herein is shown for use in percutaneous insertion of a heart pump, it may also be adapted for use in devices for other applications requiring hemostasis.

[0093] Variations and modifications will occur to those skilled in the art after reviewing this disclosure. The features disclosed may be implemented in any combination and subcombination (including multiple subcombinations and subcombinations) with one or more other features described herein. The various features described or illustrated above may be combined or integrated in other systems, including any combination thereof. Furthermore, certain features may be omitted or not implemented.

[0094] The described systems and methods may be implemented locally on the heart pump system or on a controller of the heart pump system, such as an AIC. The heart pump system may include a data processing device. The systems and methods described herein may be implemented remotely on a separate data processing device. The separate data processing device may be connected directly to the heart pump system or indirectly via a cloud application. The heart pump system may communicate in real time (or near real time) with the separate data processing device.

[0095] In general, aspects of the subject matter and functional operations described herein can be implemented as digital electronic circuitry, or as computer software, firmware, or hardware, including the structures disclosed herein and their structural equivalents, or as a combination of one or more of them. Aspects of the subject matter described herein can be implemented as one or more computer program products, i.e., one or more modules of computer program instructions encoded on a computer-readable medium for execution by or controlling the operation of a data processing apparatus. The computer-readable medium can be a machine-readable storage device, a machine-readable storage substrate, a memory device, a composition affecting a machine-readable propagated signal, or one or more combinations thereof. The term "data processing apparatus" encompasses all apparatus, devices, and machines for processing data, including, for example, a programmable processor, a computer, or multiple processors or computers. An apparatus can include, in addition to hardware, code that creates an execution environment for the computer program in question, such as code constituting a processor firmware, a protocol stack, a database management system, an operating system, or one or more combinations thereof. A propagated signal is an artificially generated signal, such as a machine-generated electrical, optical, or electromagnetic signal, generated to encode information for transmission to an appropriate receiving device.

[0096] A computer program (also known as a program, software, software application, script, or code) can be written in any type of programming language, including compiled or interpreted languages, and can be implemented in any form, including a stand-alone program, a module, a component, a subroutine, or other unit suitable for use in a computing environment. A computer program may correspond to a file in a file system. A program can be stored in part of a file that holds other programs or data (e.g., one or more scripts stored in a markup language document), in a single file dedicated to the program in question, or in multiple associated files (e.g., a file that stores one or more modules, subprograms, or portions of code). A computer program can be implemented to be executed on one computer or on multiple computers that are located at one site or distributed across multiple sites and interconnected by a communications network.

[0097] The processes and logic flows described herein may be performed by one or more programmable processors executing one or more computer programs that perform functions by operating on input data to generate output. The processes and logic flows may also be performed by, and an apparatus may be implemented as, special purpose logic circuitry, such as an FPGA (field programmable gate array) or an ASIC (application specific integrated circuit).

[0098] Processors suitable for the execution of a computer program include, by way of example, both general purpose and special purpose microprocessors, and any one or more processors of any kind of digital computer. Typically, a processor receives instructions and data from a read-only memory or a random access memory, or both. The essential elements of a computer are a processor for executing instructions and one or more memory devices for storing instructions and data. Typically, a computer will include one or more mass storage devices, such as magnetic disks, magneto-optical disks, or optical disks, for storing data, or will be operatively coupled to receive data from or transfer data to the mass storage device, or both. However, a computer need not necessarily have such devices.

[0099] Examples of changes, substitutions, and alterations can be ascertained by one skilled in the art and made without departing from the scope of the information disclosed herein. All references cited herein are incorporated by reference in their entirety and made a part of this application.

[0100] Exemplary embodiments: 1. A method for predicting a patient outcome, comprising: acquiring first data from the heart pump system during a first time period, the first data relating to a time-varying parameter of the heart pump system; Extracting a plurality of features from the first data; determining a cardiac health index indicative of the cardiac health of the patient based on the plurality of features using the predictive model; and Predicting patient outcomes based on a cardiac health index A method comprising: 2. The method of embodiment 1, wherein the cardiac health index further indicates the likelihood of the patient's recovery. 3. The method of embodiment 1 or 2, wherein the cardiac health index represents a cardiac component and a systemic perfusion component. 4. The method of any one of aspects 1-3, wherein the heart pump system is at least partially inserted within the patient's heart. 5. acquiring second data related to a physiological parameter of the patient; and Extracting a second plurality of features from the second data. Further comprising: and determining the cardiac health index is further based on the second plurality of features. The method according to any one of aspects 1 to 4. 6. The method of embodiment 5, wherein the second plurality of characteristics comprises at least one of age, sex, body surface area (BSA), urine volume, creatinine level, hydrogen ion level (pH), oxygen concentration, carbon dioxide concentration, and lactate concentration. 7. The method of any one of aspects 1-6, further comprising the step of displaying the heart health index. 8. obtaining a plurality of heart health indexes, each of which corresponds to a period of a plurality of time periods; and determining a change in the patient's health based on a plurality of cardiac health indices; The method of any one of embodiments 1 to 7, further comprising: 9. The method of embodiment 8, further comprising displaying a plurality of cardiac health indices over a plurality of time periods. 10. The method of any one of aspects 1-9, wherein the plurality of features includes at least one of aortic pressure, differential pressure, motor current, motor speed, pump pressure, left ventricular pressure, end diastolic pressure, aortic pulse pressure, intrinsic cardiac output, cardiac output, cardiac power output, configuration, flow, P level, contractility, relaxation, mean configuration, standard deviation of configuration, mean configuration range, standard deviation of configuration range, mean differential pressure, standard deviation of differential pressure, mean differential pressure range, standard deviation of differential pressure range, left ventricular pressure maximum, and left ventricular pressure minimum. 11. The method of any one of aspects 1-10, wherein the predictive model is a machine learning model. 12. The method of embodiment 11, wherein the machine learning model is one of a logistic regression method, a deep learning method, a decision tree method, a random forest method, a naive Bayes method, and a support vector machine method. 13. Displaying a measure of the relative importance of a first feature of the plurality of features compared to a second feature of the plurality of features. 13. The method of any one of embodiments 1 to 12, further comprising: 14. The step of determining a heart health index comprises: obtaining a training data set from a database, the training data set including a plurality of data points related to a time-varying parameter of the heart pump system; preprocessing the dataset to determine a third plurality of features corresponding to the plurality of data points; processing the third plurality of features to determine a pattern including a weight for each feature of the subset of the third plurality of features; Obtaining patient data; Calculating a patient's cardiac health index based on patient data and patterns 14. The method of any one of embodiments 1 to 13, comprising: 15. The method of any one of aspects 1-14, wherein the first data includes data indicative of the configuration signal, the motor current, and the P level over time. 16. The method of any one of aspects 1-15, wherein the plurality of features includes the configuration signal, the motor current, the P level, and a standard deviation of the configuration signal. 17. The method of any one of aspects 1-16, wherein the first data includes data indicative of pump pressure, differential pressure, motor current, and motor speed over time. 18. The method of any one of aspects 1-17, wherein the plurality of features includes a pump pressure maximum value, a pump pressure average value, a pump pressure minimum value, a differential pressure maximum value, a differential pressure minimum value, a motor current maximum value, a motor current minimum value, a motor current average value, and a motor speed average value. 19. The method of any one of aspects 1-18, wherein the first data includes data indicative of the location signal and flow rate over time. 20. The method of any one of aspects 1-19, wherein the plurality of features includes a positioning signal, an average flow rate, and a target flow rate. 21. The method of any one of aspects 1-20, wherein the cardiac health index is the patient's probability of survival. 22. The method of any one of embodiments 1-21, wherein the patient outcome is patient death. 23. The method of any one of embodiments 1-21, wherein the patient outcome is patient survival. 24. A catheter; A motor; a rotor operatively connected to the motor; a pump housing at least partially surrounding the rotor such that actuation of the motor drives the rotor to pump blood into the pump housing; At least one sensor; A controller configured to execute any one of the methods of aspects 1 to 23. 23. A heart pump system comprising: 25. A system including a controller configured to execute any one of the methods of aspects 1 to 23.

Claims

1. a pump including a pump housing, a rotor, and an opening positioned proximally of the pump housing; a cannula including a proximal end that connects with the distal end of the pump housing and a distal end having at least one distal opening, the pump being configured to be operated by a motor; an elongate catheter extending proximally of the pump housing; At least one sensor; and controller 1. A heart pump system comprising: The controller obtaining at least one cardiac power output value derived from measurements provided by the cardiac pump system and at least one lactate concentration value measured from the patient; determining a predictive value of an outcome for the patient based at least in part on the at least one cardiac power output value and the at least one lactate concentration value; and configured to adjust a pump speed of the cardiac pump system based on the predicted value of a patient outcome; adjusting a pump speed of the cardiac pump system includes (a) increasing the pump speed when the at least one cardiac power output value is below a first threshold and the at least one lactate concentration value is above a second threshold, or (b) decreasing the pump speed when the at least one cardiac power output value is above a first threshold and the at least one lactate concentration value is below a second threshold. Heart pump system.

2. Adjusting the pump speed increasing a pump speed when the at least one cardiac power output value falls below a first threshold and the at least one lactate concentration value exceeds a second threshold.

2. The heart pump system of claim 1, comprising:

3. Adjusting the pump speed decreasing a pump speed when the at least one cardiac power output value is above a first threshold and the at least one lactate concentration value is below a second threshold.

2. The heart pump system of claim 1, comprising:

4. The controller: updating the at least one cardiac power output value after a first time interval following a first time period; and updating the at least one lactate concentration value after a second time interval following the first period. A heart pump system according to any one of claims 1 to 3.

5. The cardiac pump system of any one of claims 1 to 4, wherein the at least one lactate concentration value is measured from the patient by a clinician and provided to the cardiac pump system on a user interface.

6. 6. The cardiac pump system of claim 1, wherein obtaining the at least one cardiac power output value comprises determining cardiac output over time from a sensor of the cardiac pump system.

7. The heart pump system of any one of claims 1 to 6, wherein the predicted value of a patient outcome indicates death or survival of the patient.

8. a pump including a pump housing, a rotor, and an opening positioned proximally of the pump housing; a cannula including a proximal end that connects with the distal end of the pump housing and a distal end having at least one distal opening, the pump being configured to be operated by a motor; an elongate catheter extending proximally of the pump housing; At least one sensor; and controller 1. A heart pump system comprising: The controller obtaining at least one cardiac power output value derived from measurements provided by the cardiac pump system and at least one lactate concentration value measured from the patient; determining a predictive value of outcome for the patient based at least in part on the at least one cardiac power output value and the at least one lactate concentration value; updating the at least one cardiac power output value after a first time interval following a first time period; updating the at least one lactate concentration value after a second time interval following the first period; and Operate the cardiac pump system to treat the patient. It is configured as follows: Heart pump system.

9. The heart pump system of claim 8 , wherein the controller adjusts operating parameters of the heart pump system based on the predicted value of patient outcome.

10. 10. The heart pump system of claim 9, wherein adjusting an operating parameter of the heart pump system comprises adjusting a pump speed based on a change in cardiac output or lactate concentration.

11. Adjusting the pump speed increasing a pump speed when the at least one cardiac power output value falls below a first threshold and the at least one lactate concentration value exceeds a second threshold.

11. The heart pump system of claim 10, comprising:

12. Adjusting the pump speed decreasing a pump speed when the at least one cardiac power output value is above a first threshold and the at least one lactate concentration value is below a second threshold.

12. The heart pump system of claim 11, comprising:

13. A cardiac pump system according to any one of claims 8 to 12, wherein the at least one lactate concentration value is measured from the patient by a clinician and provided to the cardiac pump system on a user interface.

14. 14. The cardiac pump system of claim 8, wherein obtaining the at least one cardiac power output value comprises determining cardiac output over time from a sensor of the cardiac pump system.

15. The heart pump system of any one of claims 8 to 14, wherein the predicted value of a patient outcome indicates death or survival of the patient.

Images