Hemodynamic sensor system for predicting and diagnosing hypotension endotype

Abstract

A system for determining a hypotension profile of a patient can include a hemodynamic sensor and a transducer. The system can receive a simulated haemodynamic sensor signal from the patient from the haemodynamic sensor. The system can use a converter to convert the analog hemodynamic sensor signal into an arterial pressure signal waveform and extract a plurality of cardiac health parameters from the arterial pressure signal waveform. Using a deep learning model, the system can encode a plurality of cardiac health parameters into one or more submerged cardiac health parameters. The system can generate a position of an arterial pressure signal waveform in the submerged space and determine a relative position of the arterial pressure signal waveform in the submerged space. Based on the relative position, the system can determine a hypotension profile of the patient and display an alert indicative of the hypotension profile.

Description

[0001] Cross-references to related applications This application claims priority to U.S. Provisional Patent Application No. 63 / 589,939, filed October 12, 2023, and U.S. Provisional Patent Application No. 63 / 621,900, filed January 17, 2024, the disclosures of which are incorporated herein by reference in their entirety. Technical Field

[0002] This disclosure generally relates to hemodynamic monitoring, including using monitored hemodynamic data to determine hypotensive endotypes in patients (e.g., human or veterinary subjects). Background Technology

[0003] Monitoring a patient's hemodynamic variables allows for improved patient care. Hemodynamic variables may include cardiac health parameters, such as cardiac output. Monitoring such cardiac health parameters allows systems to diagnose hypotension and provide interventions for patients with hypotension or potential hypotension. The systems and methods described in this article offer potential space rescue solutions. Summary of the Invention

[0004] On one hand, a system for determining the hypotensive endotype of a patient (e.g., a human or veterinary subject) includes a hemodynamic sensor and a converter. The system receives simulated hemodynamic sensor signals from the patient using the hemodynamic sensor. The system uses the converter to convert the simulated hemodynamic sensor signals into an arterial pressure signal waveform and extracts multiple cardiac health parameters from the arterial pressure signal waveform. Using a fully connected deep learning model, the system encodes the multiple cardiac health parameters into one or more latent space cardiac health parameters. The system generates the position of the arterial pressure signal waveform in the latent space and determines its relative position within the latent space. Based on the relative position, the system determines the patient's hypotensive endotype and displays an alert indicating the hypotensive endotype.

[0005] In another aspect, a hemodynamic sensor system can determine and display an intralow blood pressure pattern of a patient. The system can include a hemodynamic sensor that generates an analog hemodynamic sensor signal representative of an arterial pressure signal waveform of the patient, an analog-to-digital converter that converts the analog hemodynamic sensor signal into the arterial pressure signal waveform, a non-transitory memory having stored thereon executable instructions, and an electronic hardware processor in communication with the non-transitory memory. The system can receive the analog hemodynamic sensor signal from the patient from the hemodynamic sensor and can convert the analog hemodynamic sensor signal into the arterial pressure signal waveform using the analog-to-digital converter. The system can determine the intralow blood pressure pattern of the patient based on the arterial pressure signal waveform and generate data for display of an alert indicating the intralow blood pressure pattern of the patient based on the determined intralow blood pressure pattern of the patient.

[0006] In another aspect, the system can extract a plurality of cardiac health parameters from the arterial pressure signal waveform and encode the plurality of cardiac health parameters into one or more latent space cardiac health parameters using a fully connected deep learning model. In another aspect, the system can generate a location of the arterial pressure signal waveform in a latent space using the one or more latent space cardiac health parameters and determine a relative position of the arterial pressure signal waveform in the latent space based on the location of the arterial pressure signal waveform in the latent space.

[0007] In another aspect, the system can determine a set of clusters from the reference arterial pressure signal waveforms and determine a cluster associated with the arterial pressure signal waveform in the latent space based on the set of clusters. Based on the cluster of the arterial pressure signal waveform, the system can determine an intralow blood pressure pattern of the patient. The alert can further indicate at least one of the cardiac health parameters.

[0008] In another aspect, a hemodynamic sensor system can receive analog hemodynamic sensor signals from a patient from a hemodynamic sensor and convert the analog hemodynamic sensor signals to arterial pressure signal waveforms using an analog-to-digital converter. The system can extract a plurality of cardiac health parameters from the arterial pressure signal waveforms and encode the plurality of cardiac health parameters into one or more latent space cardiac health parameters using a fully connected deep learning model. The system can generate a location of the arterial pressure signal waveforms in a latent space using the one or more latent space cardiac health parameters. The system can determine a relative position of the arterial pressure signal waveforms in the latent space based on the location of the arterial pressure signal waveforms in the latent space. The system can obtain a plurality of reference arterial pressure signal waveforms from a plurality of patients and extract a plurality of cardiac health parameter reference value sets from the plurality of reference arterial pressure signal waveforms. The system can combine each of the plurality of cardiac health parameter reference value sets into a corresponding one or more reference latent space cardiac health parameters and generate a reference location of each of the plurality of reference arterial pressure signal waveforms in the latent space based on the one or more reference latent space cardiac health parameters of each of the plurality of cardiac health reference value sets. The system can determine a clustering evaluation metric of the reference locations in the latent space. The clustering evaluation metric system can indicate a clustering goodness of the reference locations. Based on the clustering evaluation metric, the system can determine an optimal number of clusters associated with the reference locations. The system can associate each of the reference arterial pressure signal waveforms with a corresponding cluster in the latent space based on the optimal number of clusters to determine a set of clusters and determine a cluster associated with the arterial pressure signal waveforms in the latent space based on the set of clusters. Based on the cluster of the arterial pressure signal waveforms, the system can determine a hypotensive profile of the patient and generate data for displaying an alert indicating the hypotensive profile of the patient based on the determined hypotensive profile of the patient.

[0009] In another aspect, the system can decode the one or more latent space cardiac health parameters into the plurality of cardiac health parameters in a raw space using the fully connected deep learning model. The system can determine a reference location of each of the plurality of reference arterial pressure signal waveforms in the raw space based on the decoded plurality of cardiac health parameters in the raw space and validate the determined optimal number of clusters associated with the reference locations by comparing the determined reference locations in the latent space with the determined reference locations in the raw space.

[0010] In another aspect, the plurality of cardiac health parameters can include one or more of a first cardiac health parameter including a stroke volume index (SVI), a second cardiac health parameter including a heart rate (HR), a third cardiac health parameter including a cardiac index (CI), a fourth cardiac health parameter including a systemic vascular resistance index (SVRI), and / or a fifth cardiac health parameter including a stroke volume variation (SVV). The system can generate data for displaying at least one of the first, second, third, fourth, or fifth cardiac health parameters based on the determined hypotensive profile of the patient.

[0011] In another aspect, the clustering evaluation metric includes at least one of: a silhouette metric, a Calinski-Harabasz index, a Davies-Bouldin index, and / or a Gaussian mixture model. The fully connected deep learning model includes an autoencoder configured to automatically convert the plurality of cardiac health parameters to one or more latent space cardiac health parameters. The relative position of the arterial pressure signal waveform in the latent space includes a cluster associated with the arterial pressure signal waveform in the latent space.

[0012] In another aspect, the hypotension intra-type includes at least one of: a vasodilation intra-type, a myocardial depression intra-type, a bradycardia intra-type, and / or a hypovolemia intra-type. The optimal number of clusters associated with the reference position can be exactly four clusters.

[0013] In another aspect, a hemodynamic sensor system can receive an analog hemodynamic sensor signal from a patient from a hemodynamic sensor and convert the analog hemodynamic sensor signal to an arterial pressure signal waveform using an analog-to-digital converter. The system can extract a plurality of cardiac health parameters from the arterial pressure signal waveform and encode the plurality of cardiac health parameters to one or more latent space cardiac health parameters using a fully connected deep learning model. The system can generate a position of the arterial pressure signal waveform in a latent space using the one or more latent space cardiac health parameters and determine a relative position of the arterial pressure signal waveform in the latent space based on the position of the arterial pressure signal waveform in the latent space. The system can determine a set of clusters from reference arterial pressure signal waveforms and determine a cluster associated with the arterial pressure signal waveform in the latent space based on the set of clusters. Based on the cluster of the arterial pressure signal waveform, the system can determine a hypotension intra-type of the patient and generate data for display indicating an alert of the hypotension intra-type of the patient based on the determined hypotension intra-type of the patient.

[0014] In another aspect, the system can decode the one or more latent space cardiac health parameters to a plurality of cardiac health parameters in a raw space using the fully connected deep learning model. The system can determine a reference position of each of a plurality of reference arterial pressure signal waveforms in the raw space based on the decoded plurality of cardiac health parameters in the raw space and validate the determined cluster associated with the arterial pressure signal waveform in the latent space by comparing the determined reference positions in the latent space to the determined reference positions in the raw space.

[0015] In another aspect, the plurality of cardiac health parameters include one or more of: a first cardiac health parameter including a stroke volume index (SVI), a second cardiac health parameter including a heart rate (HR), a third cardiac health parameter including a cardiac index (CI), a fourth cardiac health parameter including a systemic vascular resistance index (SVRI), and / or a fifth cardiac health parameter including a stroke volume variation (SVV). The system can generate data for display of at least one of the first, second, third, fourth, or fifth cardiac health parameters based on the determined hypotension subtype of the patient.

[0016] In another aspect, the system can determine a cluster evaluation metric for the reference arterial pressure signal waveform, the cluster evaluation metric configured to indicate a cluster goodness of the reference location and determine the optimal number of clusters associated with the reference location based on the cluster evaluation metric.

[0017] In another aspect, the cluster evaluation metric includes at least one of: a silhouette metric, a Calinski-Harabasz index, a Davies-Bouldin index, and / or a Gaussian mixture model. The optimal number of clusters associated with the reference location can be exactly four clusters. The fully connected deep learning model includes an autoencoder configured to automatically convert the plurality of cardiac health parameters to one or more latent space cardiac health parameters.

[0018] In another aspect, the relative location of the arterial pressure signal waveform in the latent space includes a cluster associated with the arterial pressure signal waveform in the latent space. The hypotension subtype includes at least one of: a vasodilatory subtype, a cardiomyocyte depression subtype, a bradycardia subtype, and / or a hypovolemia subtype.

[0019] In another aspect, the hemodynamic sensor system can receive, from the hemodynamic sensor, a plurality of analog hemodynamic sensor signals from a plurality of reference patients and convert, using an analog-to-digital converter, the plurality of reference analog hemodynamic sensor signals to a corresponding plurality of reference arterial pressure signal waveforms. The system can extract a corresponding set of reference cardiac health parameters from the plurality of reference arterial pressure signal waveforms and encode, using a fully connected deep learning model, the plurality of sets of reference cardiac health parameters to a plurality of corresponding one or more reference latent space cardiac health parameters. The system can generate, based on the plurality of one or more reference latent space cardiac health parameters for each of the plurality of sets of reference cardiac health parameters, a reference location in the latent space for each of the plurality of reference arterial pressure signal waveforms. The system can determine a clustering evaluation metric for the reference locations in the latent space, the clustering evaluation metric configured to indicate a clustering goodness of the reference locations and determine, based on the clustering evaluation metric, an optimal number of clusters associated with the reference locations. The system can associate, based on the optimal number of clusters, each of the reference arterial pressure signal waveforms with a corresponding cluster in the latent space to determine a set of clusters and determine, based on the set of clusters and the reference location in the latent space for each of the plurality of reference arterial pressure signal waveforms, a cluster associated with each of the plurality of reference arterial pressure signal waveforms in the latent space. The system can determine, based on the cluster associated with each of the reference arterial pressure signal waveforms, a hypotensive profile for each of the reference patients and decode, using the fully connected deep learning model, the plurality of one or more reference latent space cardiac health parameters to the corresponding plurality of sets of reference cardiac health parameters. The system can generate a training model for determining a hypotensive profile for a patient.

[0020] In another aspect, each of the sets of the plurality of reference cardiac health parameters includes one or more of: a first cardiac health parameter including a stroke volume index (SVI), a second cardiac health parameter including a heart rate (HR), a third cardiac health parameter including a cardiac index (CI), a fourth cardiac health parameter including a systemic vascular resistance index (SVRI), and / or a fifth cardiac health parameter including a stroke volume variation (SVV). The system can generate, based on the determined hypotensive profile for the patient, data for displaying at least one of the first, second, third, fourth, or fifth cardiac health parameters.

[0021] On the other hand, clustering evaluation metrics include at least one of the following: contour metric, Calinski-Harabasz index, Davies-Bouldin index, and / or Gaussian mixture model. The optimal number of clusters associated with the reference location is exactly four clusters. The fully connected deep learning model includes an autoencoder configured to automatically convert multiple cardiac health parameters into one or more latent space cardiac health parameters. On the other hand, the hypotensive subtype includes at least one of the following: vasodilatory subtype, myocardial depression subtype, bradycardia subtype, and / or hypovolemic subtype. Attached Figure Description

[0022] FIG. 1 A block diagram of a hemodynamic sensing system that can determine the hypotensive pattern of a patient based on hemodynamic signal data, according to some embodiments.

[0023] FIG. 2 A graph illustrating the example arterial pressure signal waveform.

[0024] FIG. 3 A perspective view of an example hemodynamic sensor that can be coupled (e.g., attached) to a patient according to some embodiments for sensing hemodynamic data representing the patient's arterial pressure.

[0025] FIG. 4 A perspective view of an example hemodynamic sensor for sensing hemodynamic data representing a patient's arterial pressure, according to some embodiments.

[0026] FIG. 5A A block diagram illustrating an example automatic encoder model of an internal model analysis system according to some embodiments.

[0027] FIG. 5B An example autoencoder model using fully connected encoding and decoding layers is shown according to some embodiments.

[0028] FIG. 5C This shows an example autoencoder model that uses convolutional encoding and decoding layers.

[0029] FIG. 6A A graph showing example results of calculating the clustering quality of the latent space in the range of 2 to 14 clusters.

[0030] FIG. 6B The graph shows an example evaluation of the clustering goodness of a set of locations in the original space using a clustering evaluation metric.

[0031] FIG. 6C A graph showing example results of clustering arterial pressure signal waveforms.

[0032] FIG. 6DMultiple box-and-whisker plots show example results of a trained machine learning model configured to diagnose or determine the hypotension pattern in a test patient.

[0033] FIG. 6E-6J This section describes a sample user interface that depicts endogenous trends associated with patients.

[0034] FIG. 7 An example of a method for determining a patient's hypotensive endotype is shown.

[0035] FIG. 8 An example of a method for generating a training model to determine a patient's hypotension pattern is shown.

[0036] FIG. 9 A block diagram illustrating a computer system on which various implementation schemes can be carried out. Detailed Implementation

[0037] SUMMARY Low blood pressure, or hypotension, is a physiological condition characterized by blood pressure readings below acceptable threshold levels. For most people, the acceptable blood pressure threshold is generally considered to be at least 90 / 60 mm Hg (e.g., a systolic threshold of 90 mm Hg and a diastolic threshold of 60 mm Hg). For most people, readings below this threshold are considered low blood pressure. The definition of low blood pressure can vary from person to person, so some people may be born with low blood pressure without experiencing any symptoms. Like its counterpart, high blood pressure, low blood pressure can pose health risks and requires serious medical attention. Blood pressure is a vital component of cardiovascular health, and deviations from acceptable ranges can affect organ perfusion and overall health.

[0038] Traditionally, the diagnosis of hypotension involves measuring blood pressure with a sphygmomanometer. Systolic blood pressure below a systolic threshold level (e.g., approximately 90 mm Hg) and / or diastolic blood pressure below a diastolic threshold level (e.g., approximately 60 mm Hg) can be considered indicative of hypotension. However, this conventional approach may overlook the underlying complexities causing this condition. Identifying one or more hypotension endotypes in a patient can be an important step in the proper treatment of patients with hypotension or potential hypotension. An endotype refers to a different subtype or mechanism that causes or induces hypotension in an individual patient. Hypotension endotypes generally refer to a specific type of hypotension characterized by different physiological or molecular features.

[0039] The system described herein can identify and diagnose various hypotensive endotypes. Each endotype can include different causes of hypotension. These endotypes can include, for example, vasodilation, hypovolemia, myocardial depression, bradycardia, and / or other endotypes. The system described herein can additionally or alternatively identify multiple appropriate endotype classifications based on a large sample of hypotensive patients. Several embodiments of the invention are particularly advantageous because they include one, several, or all of the following benefits: (i) reducing or preventing errors in diagnosing hypotensive endotypes, (ii) allowing real-time (including during emergency situations) diagnosis of hypotensive endotypes, (iii) using a combination of fully connected layers and unsupervised learning algorithms to overcome the functional limitations of computers in diagnosing hypotensive endotypes, and / or (iv) generating real-time output to local and / or remote computing devices based on real-time updated data, and / or verifying the effectiveness of encoded data by decoding the data and comparing it with unencoded data.

[0040] Hemodynamic sensing or monitoring systems can be used to diagnose a patient's hypotensive endotype in real time. Such systems can be more accurate and / or faster than human diagnosis. Correct diagnosis (or, in some cases, prediction) of hypotensive endotype can lead to better (e.g., more effective or faster) treatment. For example, vasodilation can be characterized by abnormal widening or dilation of blood vessels. Therefore, using vasopressors to constrict blood vessels may be appropriate. In contrast, hypovolemia can be caused by a large loss of bodily fluids, such as blood. Treatment for hypovolemia may include providing the patient with intravenous fluids, such as saline or colloids.

[0041] Hemodynamic sensing systems can acquire analog arterial pressure signals. This can be analog hemodynamic sensor signals (e.g., analog hemodynamic signals), which can be converted into signals in different forms (e.g., digital forms), such as arterial pressure signal waveforms. Hemodynamic sensing systems can use machine learning to extract a set of parameters, such as cardiac health parameters, from a patient's arterial pressure. As described herein, "cardiac health parameters" can have its simple and general meaning and can generally refer to health parameters associated with cardiovascular health (e.g., vascular health, blood health, etc.) without being specific to the heart. Hemodynamic sensing systems can use input feature sets to determine one or more hypotensive patterns in a patient while they are visiting a primary care physician's office, in an emergency care setting, and / or in any other patient care setting. In some embodiments, hemodynamic sensing systems can even be obtained "without a prescription" for patients to use at home.

[0042] Based on the severity of a specific hypotensive pattern detected by the hemodynamic sensing system, the hemodynamic sensing system can signal or alert healthcare workers and / or patients to provide an alert that the patient requires attention (e.g., immediate emergency attention).

[0043] Hemodynamic sensing system FIG. 1 This is a block diagram of a hemodynamic sensing system 100, according to some embodiments, capable of determining the hypotensive type of a patient based on hemodynamic signal data. (See diagram below.) FIG. 1 As illustrated, the hemodynamic sensing system 100 includes a hemodynamic sensor 108 coupled to a patient 104, a signal converter 112, an end-type analysis system 114, and / or a graphical user interface 132. Additionally or alternatively, the hemodynamic sensing system 100 may include a pump 110 coupled to the patient 104. In some embodiments, the hemodynamic sensing system 100 includes a remote computing device 140 connected via a network 136. The end-type analysis system 114 may include one or more processors 116, a hemodynamic data interface 118, and / or a memory 120. The memory 120 may include instructions (e.g., software instructions) stored thereon for performing one or more steps described herein. Additionally or alternatively, the memory 120 may include a machine learning model 124 and / or an unsupervised training module 128. The hemodynamic sensing system 100 may be implemented in patient care environments such as intensive care units (ICUs), operating rooms (ORs), and / or other patient care environments.

[0044] For example, in some embodiments, the hemodynamic sensing system 100 includes a hemodynamic sensor 108, a signal converter 112, a memory 120, and one or more processors 116. The one or more processors 116 are configured to execute instructions stored in the memory 120 to receive analog hemodynamic sensor signals from the patient via the hemodynamic sensor 108. The one or more processors 116 may enable the hemodynamic sensing system 100 to use the signal converter 112 to convert the analog hemodynamic sensor signals into arterial pressure signal waveforms and determine the patient's hypotensive pattern based on the arterial pressure signal waveforms.

[0045] One or more processors 116 may be one or more hardware and / or electronic processors. Processor 116 may include one or more of a microprocessor, controller, digital signal processor (DSP), application-specific integrated circuit (ASIC), field-programmable gate array (FPGA), or other equivalent discrete or integrated logic circuitry. In some embodiments, one or more processors 116 may include one or more graphics processing units (GPUs). One or more GPUs may be configured to perform linear algebraic calculations on matrices. For example, machine learning model 124 and / or unsupervised training module 128 may use one or more GPUs to perform the following operations.

[0046] Memory 120 may include a computer-readable storage medium. In some examples, the computer-readable storage medium may include a non-transitory medium. The term "non-transitory" may indicate that the storage medium is not embodied in a carrier wave or propagating signal. In some examples, the non-transitory storage medium may store data that may change over time (e.g., in RAM or cache). Memory 120 may include volatile and non-volatile computer-readable memory. Examples of volatile memory may include random access memory (RAM), dynamic random access memory (DRAM), static random access memory (SRAM), and other forms of volatile memory. Examples of non-volatile memory may include, for example, magnetic hard disks, optical disks, flash memory, or electrically programmable memory (EPROM) or electrically erasable programmable memory (EEPROM).

[0047] In some embodiments, the hemodynamic sensing system 100 may include a remote computing device 140, or communicate with such a remote computing device via a network 136. The remote computing device 140 may include a computing system, such as a client terminal in a hospital or clinic. The remote computing device 140 may include a computer in an ICU or OR. Additionally or alternatively, the remote computing device 140 may include mobile electronic devices, such as laptops, smartphones, health monitors, and / or other electronic devices. In some embodiments, the remote computing device 140 may include a display interface. The remote computing device 140 may be configured to receive (e.g., via network 136) and / or display data transmitted from the endogram analysis system 114, such as identified hypotensive endograms, one or more cardiac health parameters described herein, images and / or other data associated with the cardiac health of patient 104. The network 136 may include wireless and / or wired network connections.

[0048] Hemodynamic sensor Hemodynamic sensor 108 may include one or more sensors coupled to patient 104 (e.g., attached to patient, inserted into patient body, etc.). Hemodynamic sensor 108 may acquire (e.g., receive, sense) a hemodynamic signal representing an arterial pressure waveform of patient 104 (e.g., see arterial pressure signal waveform 200). Signal converter 112 may convert the sensed signal into data (e.g., digital data).

[0049] Hemodynamic sensor 108 can be inserted into a patient via a femoral artery catheter inserted into the patient's leg. Additionally or alternatively, hemodynamic sensor 108 may include a minimally invasive hemodynamic sensor that can be attached to the patient via, for example, a radial artery catheter inserted into the patient's arm (see, for example, hemodynamic sensor 300). For example, in some embodiments, hemodynamic sensor 108 includes a non-invasive hemodynamic sensor that can be attached to the patient via one or more finger cots configured to sense data representing the patient's arterial pressure. For example, hemodynamic sensor 108 may include an inflatable finger cot and a cardiac reference sensor (see, for example, [link to relevant documentation]). FIG. 3 In some embodiments, hemodynamic sensor 108 does not include any invasive hemodynamic sensor (e.g., an in-line hemodynamic sensor, such as hemodynamic sensor 300). In some embodiments, hemodynamic sensor 108 includes a wireless (e.g., infrared) or wired connection to signal converter 112.

[0050] Hemodynamic sensor 108 can perform regular hemodynamic signal measurements on patient 104. Measurements can be performed approximately every 5 seconds, approximately every 10 seconds, approximately every 20 seconds, approximately every 30 seconds, approximately every 45 seconds, approximately every 1 minute, approximately every 2 minutes, approximately every 5 minutes, approximately every 10 minutes, approximately every 15 minutes, approximately every 20 minutes, approximately every 30 minutes, approximately every hour, or any value falling within a range with endpoints therein. For example, in some embodiments, measurements are performed approximately every 15 minutes. The measurement rate can be determined in part by the patient's determined cardiac health, such as whether patient 104 is currently experiencing hypotension, how many types of hypotension have been identified for 104, whether one or more cardiac health parameters exceed one or more corresponding thresholds, etc. The determined measurement rate can be determined automatically or can be set by the user. Although the hemodynamic sensor 108 can monitor the arterial pressure of the patient 104 over an extended period of time, the hemodynamic sensor 108 may only need to monitor the arterial pressure of the patient 104 for a few minutes (e.g., 5 minutes) to generate enough data for the endotype analysis system 114 to determine the hypotensive endotype of the patient 104.

[0051] Signal converter Signal converter 112 may include an analog-to-digital converter (ADC) and / or a digital-to-analog converter (DCA). Signal converter 112 may include hardware and / or software converters. Signal converter 112 can send the converted data to internal analysis system 114. See below for reference. FIG. 3 and FIG. 4 Describe the example signal converter 112.

[0052] Intrinsica analysis system The endogram analysis system 114 can receive converted data from the signal converter 112 via the hemodynamic data interface 118. In some embodiments, the signal converter 112 is configured to convert the data into a form (e.g., a format) that can be read and / or accepted by the hemodynamic data interface 118. Once the converted data is obtained through the endogram analysis system 114, the processor 116 can execute instructions stored in the memory 120 to analyze the converted data.

[0053] The data to be converted may include one or more health parameters, such as heart health parameters (see also...). FIG. 2 Cardiac health parameters are highly predictive of potential (e.g., future) or actual (e.g., current) hypotension in patient 104. These cardiac health parameters can be derived from digital hemodynamic waveform data. Hemodynamic sensing system 100 can utilize some or all of the cardiac health parameters to diagnose one or more hypotension endotypes and / or a hypotension probability index (hereinafter referred to as “HPI”) corresponding to the probability of future hypotension events and / or associated endotypes in patient 104. For example, endotype analysis system 114 can determine the HPI corresponding to the probability of patient 104 developing one or more hypotension endotypes. The probability can be used to perform one or more actions described herein, such as determining a treatment plan for the patient and / or generating commands to deliver treatment to patient 104 via an infusion pump. For example, endotype analysis system 114 can determine a specific hypotension endotype based on determining that the HPI is above a threshold HPI (e.g., 90, 95, 100, etc.). The endotype analysis system 114 can use cardiac health parameters to determine endotype and / or determine HPI by comparing cardiac health parameters with reference values ​​such as a numerical table (e.g., a lookup table), as described herein.

[0054] As described herein, one or more cardiac health parameters may be sent to a graphical user interface (GUI) 132 for display. The GUI 132 may issue alerts and / or recommend actions to address a specific hypotensive pattern to a user (e.g., a healthcare professional or the patient themselves). Such alerts help ensure timely warnings to users of potential acute hypotensive events. Furthermore, by enabling users to access the GUI 132, which displays or shows one or more cardiac health parameters identified as indicating a current or future hypotensive pattern, the GUI 132 provides detailed diagnostic information to allow users to identify the most probable cause of the hypotensive pattern and / or the best medical interventions for preventing or treating that specific pattern.

[0055] Machine learning model Further reference FIG. 1The endotype analysis system 114 can be configured to identify one or more cardiac health parameters associated with the identification of a hypotensive endotype and / or to determine the hypotensive endotype itself. Cardiac health parameters may include health parameters that can be measured from arterial pressure signal waveforms and / or that can be used to identify (e.g., diagnose) a hypotensive endotype. Many potential cardiac health parameters can be extracted from arterial pressure signal waveforms, but they may include, for example, cardiac output (CO), stroke volume (SV), stroke volume variability (SVV), diastolic blood pressure (DIA), pulse rate (PR), stroke volume index (SVI), systemic vascular resistance (SVR), mean arterial pressure (MAP), HPI, and / or other parameters. Additionally or alternatively, cardiac health parameters may include systemic vascular resistance index (SVRI), cardiac index (CI), and / or systolic blood pressure (SYS).

[0056] The endogenous modeling system 114 can receive converted data (e.g., cardiac health parameters) from the signal converter 112 via a hemodynamic data interface 118. The endogenous modeling system 114 can then send the cardiac health parameters to a machine learning model 124 and / or an unsupervised training module 128. The machine learning model 124 can be configured to receive the cardiac health parameters and encode them into different health parameters. Before encoding, the cardiac health parameters can be considered to be in a “raw space.” The raw space can refer to the mathematical or analytical space of values ​​before encoding and / or after decoding. Once encoded, the different health parameters can be cardiac health parameters in different spaces (e.g., latent spaces). The “latent space” can refer to the mathematical or analytical space of values ​​after encoding or compression (e.g., by an autoencoder described below) and / or before decoding. Values ​​in the latent space can be more compact (e.g., lower dimensionality) than values ​​in the raw space. Latent space values ​​are not directly observable. In some embodiments, the machine learning model 124 is not configured to encode cardiac health parameters into any space other than the latent space. Additionally or alternatively, the machine learning model 124 can be configured to decode the latent space into the original space rather than a higher-dimensional space. These different cardiac health parameters can be fed from the machine learning model 124 to the unsupervised training module 128.

[0057] In some embodiments, the machine learning model 124 may further decode latent space cardiac health parameters back to corresponding cardiac health parameters. The machine learning model 124 may include a fully connected deep learning model, a convolutional deep learning model, and / or some other type of learning model. In some embodiments, one or more of these cardiac health parameters may be sent to a graphical user interface 132 for display to healthcare professionals.

[0058] Machine learning model 124 may include an autoencoder model (or "autoencoder"). An autoencoder model may receive raw or unencoded data (e.g., cardiac health parameters) as input and output encoded (and / or decoded) parameters based on the cardiac health parameters. The autoencoder model may be a deep learning-based model. The autoencoder model may be trained to transform unencoded cardiac health parameters into encoded cardiac health parameters, as discussed in more detail below.

[0059] Unsupervised learning module The unsupervised training module 128 can receive different cardiac health parameters to generate corresponding locations of these parameters in different spaces based on the encoded cardiac health parameters. As described below, these locations can be in latent space or other spaces. The unsupervised training module 128 can use one or more criteria or sets of criteria to determine how these cardiac health parameters should be located.

[0060] Unsupervised training module 128 can identify the number of clusters for the model and determine the location of each position within a particular cluster in a different space (e.g., it may be a reference position). In some embodiments, additionally or alternatively, unsupervised training module 128 can identify the number of clusters for the model and determine the location in the original space. Clusters in the original space can be used to validate clusters in different spaces (e.g., latent spaces). In some embodiments, unsupervised training module 128 is not constrained by the number of clusters. For example, unlike supervised training modules, unsupervised training module 128 may be allowed to compute any number of latent clusters without restriction. In some embodiments, endotype analysis system 114 can compare one or more cardiac health parameters with reference values ​​to determine endotype. Reference values ​​may include a range of values ​​for the particular cardiac health parameter (e.g., see...). FIG. 6D The internal modeling system 114 may use reference values ​​of multiple associated indices (e.g., SVI, HR, CI, SVRI, SVV, etc.) to determine internal models and / or the probability of internal model applicability. Reference values ​​may include tables (e.g., lookup tables). Additionally or alternatively, reference values ​​may include one or more thresholds, such as a maximum threshold and / or a minimum threshold.

[0061] Graphical user interface The unsupervised training module 128 can receive different cardiac health parameters to generate corresponding locations of these parameters in different spaces based on the encoded cardiac health parameters. As described below, these locations can be in latent space or other spaces. The unsupervised training module 128 can use one or more criteria or sets of criteria to determine how these cardiac health parameters should be located.

[0062] The graphical user interface 132 provides a user interface including one or more control elements to enable user interaction and / or input. User input can be sent to the endotype analysis system 114. The graphical user interface 132 can provide sensory alerts based on a determined endotype and / or based on measurement data from arterial pressure signal waveforms (e.g., from one or more extracted cardiac health parameters). Sensory alerts can be configured to warn medical personnel based on whether the hypotensive endotype causes an emergency. Sensory alerts may additionally or alternatively include instructions on how to treat the determined hypotensive endotype, the urgency of the treatment, and / or relevant cardiac health parameters that should be addressed based on the hypotensive endotype. For example, a hypovolemic endotype may require emergency infusion of saline solution. Sensory alert 158 ​​can be implemented as one or more of visual, auditory, tactile, and / or other types of sensory alerts. For example, a sensory alert can be invoked as any combination of flashing and / or colored graphics shown in the graphical user interface 132. Additionally or alternatively, the graphical user interface 132 may display a defined hypotension pattern, such as an alarm sound or a repetitive tone, and a tactile alarm configured to vibrate a hemodynamic monitor (not shown) or otherwise deliver a physiological impulse perceptible to a healthcare worker or other user. The tactile alarm signal may be transmitted wirelessly via network 136.

[0063] The graphical user interface 132 may include a liquid crystal display (LCD), a light-emitting diode (LED) display, an organic light-emitting diode (OLED) display, or other display devices suitable for providing information to a user in graphical form. The graphical user interface 132 may include one or more touch-sensitive and / or presence-sensitive elements, such as a touch-sensitive display. In some embodiments, user input may be received in the form of gesture input, such as touch gestures, scrolling gestures, zoom gestures, or other gesture inputs. In some examples, the graphical user interface 132 may include one or more physical control elements, such as physical buttons, keys, knobs, a mouse, a keyboard, or other physical control elements configured to receive user input to interact with components of the hemodynamic sensing system 100. However, in some embodiments, the graphical user interface 132 does not allow user selection (e.g., does not accept user input) and instead only provides output to the user.

[0064] Pump Further reference FIG. 1The hemodynamic sensing system 100 may include a pump 110 capable of providing treatment to a patient 104. An endogenous analysis system 114 may communicate with the pump 110 (e.g., via a data interface 118). The pump 110 may include an infusion pump. For example, the pump 110 may include a gravity infusion pump, a syringe infusion pump, an elastomeric infusion pump, a volumetric infusion pump, a patient-controlled analgesia (PCA) pump, an enteral infusion pump, an insulin pump, a portable infusion pump, and / or any other type of infusion pump.

[0065] Gravity infusion pumps utilize gravity to deliver fluid to a venous (IV) line and ultimately into the patient. The infusion container can be positioned above the patient to allow gravity to work. Syringe infusion pumps can be built into or directly connected to a pump system. Syringe infusion pumps can provide small-volume infusions and / or offer enhanced control over fluid delivery.

[0066] Elastomer infusion pumps can be disposable pumps that operate without an external power source. Such pumps 110 may include one or more elastic balloons filled with fluid, which can create positive pressure to expel fluid along the IV tubing. In some embodiments, pump 110 may allow patient 104 to at least partially control the administration of treatment, such as in a PCA pump.

[0067] Pump 110 may receive instructions from endogenous pattern analysis system 114, such as via data interface 118. Endogenous pattern analysis system 114 may generate control signals to instruct pump 110 to deliver treatment to patient 104, such as intravenous therapy using an intravenous therapeutic agent. The treatment provided may be based on an endogenous pattern determined by endogenous pattern analysis system 114. Additionally or alternatively, the treatment may be based on a treatment protocol, such as a treatment protocol determined by endogenous pattern analysis system 114. The treatment protocol may include instructions related to one or more therapeutic agents, the amount of therapeutic agent, the duration of time associated with the delivery of the therapeutic agent, and / or any other treatment described herein.

[0068] Treatment may include intravenous infusion of one or more fluids to expand plasma volume, which may help increase blood pressure. The type of fluid used (e.g., crystalloid and / or colloid) may be based on the patient's condition (e.g., intracellular and / or the likelihood of intracellularity) and the desired effect. Fluid therapy may correct volume depletion and / or provide additional oxygen. Additionally or alternatively, treatment may include one or more vasopressors, which may include drugs for constricting blood vessels. This increases vascular resistance and blood pressure. Vasopressors stimulate receptors in the cardiovascular system, thereby causing vasoconstriction. Additionally or alternatively, pump 110 may deliver positive inotropic agents: these drugs enhance myocardial contractility and improve cardiac output. Positive inotropic agents improve myocardial contractility and / or heart rate, which in turn supports better blood pressure control. In some embodiments, treatment may include fluid resuscitation, which may increase blood volume and / or improve tissue perfusion. The therapeutic agent may initially be titrated with a low dose, possibly gradually increasing the dose over time to reduce the likelihood of adverse reactions.

[0069] Arterial pressure signal waveform FIG. 2 A graph illustrating an example arterial pressure signal waveform 200 is provided. The arterial pressure signal waveform 200 may include multiple attributes, each indicating one or more aspects of blood flow, cardiac output, and / or other aspects of the cardiac health parameters described herein. The arterial pressure signal waveform 200 may correspond to a single heartbeat of patient 104. Hemodynamic sensor 108 may be configured to detect new arterial pressure signal waveforms 200 for multiple heartbeats within a specific time range. For example, hemodynamic sensor 108 may be able to generate a new arterial pressure signal waveform 200 for each heartbeat.

[0070] Arterial pressure signal waveform 200 is an example waveform corresponding to hemodynamic data sensed by hemodynamic sensor 108 and converted by signal converter 112. Arterial pressure signal waveform 200 (represented via digital hemodynamic data) may include various markers indicative of the cardiac health of patient 104. As discussed herein, various cardiac health parameters can be extracted via end-type analysis system 114. Before extracting cardiac health parameters, a heartbeat detector can identify the start and end of each heartbeat corresponding to each waveform. The heartbeat detector can be a hardware detector (e.g., a resistive detector, an inductive detector, an optical detector, etc.) or a software detector, such as a software heartbeat detector. The heartbeat detector may identify the start of a heartbeat based on maximum arterial pressure, minimum arterial pressure, maximum or minimum rate of change of arterial pressure, and / or the second derivative of arterial pressure with respect to time. Based on heartbeat identification within the arterial pressure signal waveform 200, various cardiac health parameters can be extracted from the waveform on a continuous (e.g., real-time), beat-by-beat basis. In some embodiments, the internal model analysis system 114 can obtain any necessary input data from a single arterial pressure signal waveform 200 without relying on multiple such arterial pressure signal waveforms.

[0071] The start indicator 224 of the arterial pressure signal waveform 200 corresponds to the start of a heartbeat. The maximum systolic indicator 226 of the arterial pressure signal waveform 200 corresponds to the maximum systolic pressure, marking the end of the systolic rise. The notch indicator 228 of the arterial pressure signal waveform 200 corresponds to the presence of the dicrotic notch and the corresponding pressure, marking the end of the systolic decay. The minimum diastolic indicator 230 of the arterial pressure signal waveform 200 corresponds to the minimum diastolic pressure of the patient's heartbeat 104. Furthermore, the arterial pressure gradient, or the pressure difference between points on the arterial pressure signal waveform 200, can be used to identify cardiac health parameters. For example, the pulmonary artery pressure 232 represents the difference between the minimum diastolic pressure (minimum diastolic indicator 230) and the maximum systolic pressure (maximum systolic indicator 226). As shown, the slope S2 is the slope of the arterial pressure signal waveform 200, which can also be used to determine cardiac health parameters. The slope S2 is depicted at one location, but represents multiple slopes that can be determined at multiple locations along the arterial pressure signal waveform 200. For example, the maximum and / or minimum time derivatives of the arterial pressure signal waveform 200 can be used to calculate cardiac health parameters.

[0072] Additional indicators from the arterial pressure signal waveform 200 can be used to calculate cardiac health parameters. For example, the interval between the systolic maximum indicator 226 and the notch indicator 228 can be extracted from the arterial pressure signal waveform 200. Additionally or alternatively, the interval between the start indicator 224 and the diastolic minimum indicator 230 can be extracted from the arterial pressure signal waveform 200. The end-type analysis system 114 can use these and / or other indicators from the arterial pressure signal waveform 200 to identify additional cardiac health parameters from the arterial pressure signal waveform 200. For example, the end-type analysis system 114 can determine systolic rise (e.g., between the start indicator 224 and the systolic maximum indicator 226), systolic decay (e.g., between the systolic maximum indicator 226 and the notch indicator 228), systolic duration (e.g., between the start indicator 224 and the notch indicator 228), diastolic duration (e.g., between the notch indicator 228 and the diastolic minimum indicator 230), and / or heart rate intervals (between consecutive start indicators 224). Such markers may include the mean arterial pressure during one of the aforementioned intervals. The area under the curve of the arterial pressure signal waveform 200 and the standard deviation of the arterial pressure signal waveform 200 determined for the aforementioned intervals may also be used as inputs for cardiac health parameters and / or for calculating such cardiac health parameters.

[0073] Example hemodynamic sensor FIG. 3 A perspective view of an example hemodynamic sensor 300 that can be coupled (e.g., attached) to a patient 104 for sensing hemodynamic data representing the patient's arterial pressure. FIG. 1 The hemodynamic sensor 108 may include the hemodynamic sensor 300 and / or include one or more of its features. The hemodynamic sensor 300 is an example of a minimally invasive hemodynamic sensor that may be attached to the patient 104 via, for example, a radial artery catheter inserted into the patient's arm. In other examples, the hemodynamic sensor 300 may be attached to the patient 104 via a femoral artery catheter inserted into the patient's leg.

[0074] As illustrated, the hemodynamic sensor 300 includes a housing 318, a fluid inlet port 320, a catheter-side fluid port 322, and an I / O cable 324. The fluid inlet port 320 is configured to connect to a fluid source, such as a saline bag or other fluid input source, via conduit or other fluid (e.g., hydraulic) connection. The catheter-side fluid port 322 is configured to connect to a catheter (e.g., a radial artery catheter or a femoral artery catheter) inserted into a patient's arm (i.e., a radial artery catheter) or leg (i.e., a femoral artery catheter) via conduit or other fluid connection. The I / O cable 324 may be configured to connect to a hemodynamic monitor (e.g., a graphical user interface 132) via one or more I / O connectors in a monitor. The housing 318 of the hemodynamic sensor 300 may contain (e.g., encapsulate) one or more pressure transducers, communication circuitry, processing circuitry, and / or corresponding electronic components to sense fluid pressure corresponding to the patient's arterial pressure. One or more of these may be transmitted via the I / O cable 324 to the hemodynamic monitor (e.g., the graphical user interface 132).

[0075] In operation, a fluid column (e.g., a saline solution) is introduced from a fluid source (e.g., a saline bag) via a fluid inlet port 320 through a hemodynamic sensor 300 toward a catheter-side fluid port 322 inserted into the patient. Arterial pressure is transmitted through the fluid column to a pressure sensor located within a housing 316, which senses the pressure of the fluid column. The hemodynamic sensor 300 converts the sensed fluid column pressure into an electrical (e.g., analog) signal via a pressure transducer and outputs a corresponding electrical signal. Thus, the hemodynamic sensor 300 can send analog sensor data (or a digital representation of analog sensor data) to an internal analysis system 114 (e.g., via a signal converter 112), representing real-time monitoring of the patient's arterial pressure, and possibly even beat-by-beat monitoring.

[0076] FIG. 4This is a perspective view of an example hemodynamic sensor 426 used to sense hemodynamic data representing a patient's arterial pressure. Hemodynamic sensor 426 is an example of a non-invasive hemodynamic sensor that can be attached to a patient via one or more finger cots to sense data representing the patient's arterial pressure. Hemodynamic sensor 426 includes an inflatable finger cot 428 and a cardiac reference sensor 430. Inflatable finger cot 428 may include an inflatable blood pressure cuff configured to inflate and deflate under the control of a pressure controller (not shown), which is pneumatically connected to inflatable finger cot 428. Inflatable finger cot 428 may additionally or alternatively include an optical (e.g., infrared) transmitter and / or optical receiver electrically connected to the pressure controller (not shown). The optical transmitter and optical receiver can measure changes in arterial volume under the finger cot. The optical transmitter and optical receiver can be positioned to transmit and receive light between them via the inflatable blood pressure cuff.

[0077] In operation, the pressure controller continuously adjusts the pressure within the finger sleeve to maintain a constant volume of artery in the finger (e.g., the empty volume of the artery), as measured via an optical transmitter and optical receiver of the inflatable finger sleeve 428. The pressure applied by the pressure controller to continuously maintain the empty volume represents the blood pressure in the finger and can be transmitted by the pressure controller to an internal model analysis system (e.g., internal model analysis system 114). The cardiac reference sensor 430 measures the hydrostatic height difference between the level at which the finger is held and a reference level (typically the level of the heart) used for pressure measurement. Thus, the hemodynamic sensor 426 transmits sensor data representing a basic continuous beat-by-beat monitoring of the patient's arterial pressure waveform. As described above, this sensor data can be used to extract the arterial pressure signal waveform 200 (e.g., arterial pressure signal waveform 200) and / or extract one or more cardiac health parameters from this arterial pressure signal waveform.

[0078] Internal shape determined The endotype analysis system 114 can be configured to identify one or more cardiac health parameters associated with the identification of a hypotensive endotype and / or to determine the hypotensive endotype itself. Cardiac health parameters may include health parameters that can be measured from an arterial pressure signal waveform (e.g., arterial pressure signal waveform 200) and / or used to identify (e.g., diagnose) a hypotensive endotype. Many potential cardiac health parameters can be extracted from an arterial pressure signal waveform, but they may include, for example, cardiac output (CO), stroke volume (SV), stroke volume variability (SVV), diastolic blood pressure (DIA), pulse rate (PR), stroke volume index (SVI), systemic vascular resistance (SVR), mean arterial pressure (MAP), HPI, and / or other parameters. Additionally or alternatively, cardiac health parameters may include systemic vascular resistance index (SVRI), cardiac index (CI), and / or systolic blood pressure (SYS).

[0079] The endogenous model analysis system 114 can receive converted data (e.g., cardiac health parameters in digital form) from the signal converter 112 via the hemodynamic data interface 118. The endogenous model analysis system 114 can then send the cardiac health parameters to the machine learning model 124 and / or the unsupervised training module 128.

[0080] Machine learning model 124 can be configured to receive cardiac health parameters and encode them into different health parameters. For example, the encoded health parameters can be health parameters in different spaces, such as latent spaces. As an example, the endogenous analysis system 114 can encode multiple cardiac health parameters into one, two, three, four, or more latent space cardiac health parameters. In some embodiments, the endogenous analysis system 114 can convert cardiac health parameters into multiple latent space cardiac health parameters, the number of which is less than or at most half the number of cardiac health parameters in the original space. The endogenous analysis system 114 can determine or generate the location of those one or more latent space cardiac health parameters. This location can be within the latent space. This location can be based on the encoding of the original cardiac health parameters. For two latent space cardiac health parameters, the two-dimensional location can correspond to the coordinate positions of the two latent space cardiac health parameters within a graph.

[0081] Machine learning model 124 can determine the relative positions of latent space cardiac health parameters. The relative positions of latent space cardiac health parameters may correspond to and / or represent specific latent space cardiac health parameter clusters corresponding to a particular determined latent space cardiac health parameter. In some embodiments, machine learning model 124 may additionally or alternatively determine the relative positions of raw (e.g., decoded or uncoded) cardiac health parameters in the raw space.

[0082] To determine the relative position of latent space cardiac health parameters corresponding to the received arterial pressure signal waveform in the latent space, machine learning model 124 may feed the latent space cardiac health parameters to a learning algorithm, such as unsupervised training module 128. Unsupervised training module 128 may determine the position (e.g., in the latent space) of other arterial pressure signal waveforms from multiple patients, such as reference patients that can serve as the intrinsic model analysis system 114.

[0083] The end-type analysis system 114 can obtain multiple reference cardiac health parameters from corresponding reference arterial pressure signal waveforms. For example, the end-type analysis system 114 can receive multiple reference simulated hemodynamic sensor signals from multiple reference patients. One or more signal converters (e.g., signal converter 112) can convert the multiple reference simulated hemodynamic sensor signals into multiple corresponding reference arterial pressure signal waveforms. The end-type analysis system 114 can extract corresponding sets of reference cardiac health parameters from the multiple reference arterial pressure signal waveforms. Each set of reference cardiac health parameters may include one or more cardiac health parameters described herein, such as stroke volume index (SVI), heart rate (HR), cardiac index (CI), systemic vascular resistance index (SVRI), and / or stroke volume variation (SVV). The end-type analysis system 114 can use a machine learning model 124 to encode the multiple sets of cardiac health parameter reference values ​​into multiple (e.g., first and second) reference latent space cardiac health parameters.

[0084] Autoencoding data in latent space As described above, the internal analysis system 114 may include a machine learning model (e.g., a trained machine learning model), such as an autoencoder. FIG. 5A This is a block diagram illustrating an example autoencoder model 500 of an internal model analysis system (e.g., internal model analysis system 114) according to some embodiments. The autoencoder model 500 may include input data 550, one or more encoding filters 552, a latent space 554, one or more decoding filters 556, and output data 558. The encoder 560 may include input data 550 and / or encoding filters 552. Additionally or alternatively, the decoder 562 may include decoding filters 552 and / or output data 558.

[0085] The autoencoder model 500 may be a machine learning model. For example, the autoencoder model 500 may be a deep learning-based model trained or learned to use a neural network architecture to convert cardiac health parameters into latent space cardiac health parameters. The input data 550 of the autoencoder model 500 may include one or more cardiac health parameters (e.g., in digital form) and / or one or more arterial pressure signal waveforms. The input data 550 may be received at regular and / or irregular intervals. The input data 550 may be received approximately every 5 seconds, approximately every 10 seconds, approximately every 20 seconds, approximately every 30 seconds, approximately every 45 seconds, approximately every 1 minute, approximately every 2 minutes, approximately every 5 minutes, approximately every 10 minutes, approximately every 15 minutes, approximately every 20 minutes, approximately every 30 minutes, approximately every hour, or any value falling within a range having endpoints therein. For example, in some embodiments, the input data 550 is received approximately every 15 minutes.

[0086] Input data 550 can be encoded by one or more encoding filters 552 of the autoencoder model 500 into a condensed version of data in a different data space, such as latent space 554. Latent space 554 can store data in a condensed form or format. The condensed data in latent space 554 can then be decoded by one or more decoding filters 552 of the autoencoder model 500 into output data 558. Output data 558 can be the original one or more cardiac health parameters and / or one or more arterial pressure signal waveforms. The autoencoder model 500 can compare the output data 558 with the input data 550 to confirm that no relevant data was lost during filtering by any of the encoding filters 552 and / or decoding filters 556. The autoencoder model 500 can be tuned to remove irrelevant or redundant data that is unnecessary for determining hypotension patterns. Encoder 560 may refer to a portion of the autoencoder model 500 that uses one or more encoding filters 552 to encode or compress the input data 550 into a smaller number of variables that include all the most relevant information from the input data 550. Decoder 562 can use one or more decoding filters 556 to obtain compressed encoded information and expand it to create output data 558.

[0087] FIG. 5B Example autoencoder model 500 using fully connected encoding and decoding layers is shown. FIG. 5C An example autoencoder model 500 using convolutional encoding and decoding layers is shown. Unless otherwise specified, reference will be made to these two figures. These are the encoding filter 552 of encoder 560 and the decoding filter 556 of decoder 562, respectively. The encoding filter 552 of encoder 560 may include multiple encoding filters 552. Additionally or alternatively, the decoding filter 552 may include multiple decoding filters 552.

[0088] Encoding filter 552 performs mathematical operations to transform input data 550 into compressed data. Data samples of each element of input data 550 (e.g., cardiac health parameters) are reduced by each of the encoding filters 552 according to a target encoding factor until fully compressed data reaches the latent space 554. Decoding filter 556 is applied to the compressed data in latent space 554 to produce output data 558. Data samples are expanded by each of the decoding filters 556 according to a target decoding factor (e.g., different from the target encoding factor) until output data 558 is obtained. Algorithms within autoencoder model 500 can automatically generate the architecture of encoding filter 552 and / or decoding filter 556.

[0089] like FIG. 5BAs shown, the architecture may include fully connected layers, wherein each of one or more encoding filters 552 and / or each of one or more decoding filters 556 receives the same number of data points as the data points output from the previous layer. In some embodiments, the autoencoder model 500 includes only fully connected layers and / or does not include any convolutional layers. FIG. 5B The fully connected deep learning autoencoder model 500 may be referred to as a dense or feedforward neural network. As shown, each node in a layer may be connected to each node in the next layer. Additionally or alternatively, the output from each node may be used as the input to each node in the next layer. Each node may represent a corresponding input value, such as a cardiac health parameter or a feature thereof. The autoencoder model 500 may include one or more hidden layers between input data 550 and latent space 554. Additionally or alternatively, the autoencoder model 500 may include one or more hidden layers between latent space 554 and output data 558. In one or more of these hidden layers, each node may be fully connected to nodes in the previous and / or next layer. Each layer may apply weights or weightings to the data of each node. These weights may be learned during the training phase of the model. Each hidden layer may apply one or more activation functions to temporary data received from the previous layer (e.g., temporary encoded data 551, temporary decoded data 557). One or more of the activation functions may be nonlinear functions, such as trigonometric functions. In some embodiments, the activation function includes a hyperbolic tangent (tanh) function.

[0090] During training of the autoencoder model 500, the weights of each connection (and / or activation function and / or its value itself) between layers can be adjusted via backpropagation and / or gradient descent. The autoencoder model 500 can adjust these connection weights based on the goal of reducing or even minimizing the difference between the predicted output and the actual output.

[0091] like FIG. 5C As shown, the autoencoder model 500 may include a convolutional neural network with automatic selection of convolutions and pooling based on input data 550 and target output data 558. Complex nonlinear relationships may exist in the filtered data at each of one or more layers of the encoding filter 552 and / or decoding filter 556. FIG. 5C In the example, the encoder filter 552 and decoder filter 556 have an 8-56-32 architecture. The first-layer encoder filter 552 has 8 filters, the second-layer encoder filter 552 has 56 filters, and the third-layer encoder filter 552 has 32 filters. As shown in the figure, the first-layer decoder filter 552 has 32 filters, the second-layer decoder filter 552 has 56 filters, and the third-layer decoder filter 552 has 8 filters. FIG. 5CIn the example, the autoencoder model 500 has 12,000 parameters and requires 112 convolution operations. The parameters can be determined when the autoencoder model 500 is trained or becomes trainable. Once trained, the encoding filter 552 can compress and encode the input data 550 (e.g., cardiac health parameters, arterial pressure signal waveform). Additionally or alternatively, once trained, the decoding filter 552 can expand and decode the compressed data (e.g., latent space cardiac health parameters) into output data 558 (e.g., cardiac health parameters, arterial pressure signal waveform).

[0092] The autoencoder model 500 can be used in conjunction with an unsupervised training module (e.g., unsupervised training module 128) to further analyze the latent space data (e.g., latent space cardiac health parameters) in the latent space 554 before the latent space data is decoded into output data 558.

[0093] Latent space position determination The unsupervised training module 128 can receive these different multiple reference latent space cardiac health parameters to generate a reference location in the latent space for each of the multiple reference arterial pressure signal waveforms. The unsupervised training module 128 can use one or more criteria (e.g., clustering evaluation criteria or clustering evaluation measures) or such a set of criteria to determine how these latent space cardiac health parameters should be located. Using clustering evaluation criteria or clustering evaluation measures can include determining the clustering goodness among multiple data points in the latent space. In some embodiments, the unsupervised training module 128 can use one or more of the same clustering evaluation criteria or clustering evaluation measures to determine the clustering goodness among multiple data points in the original space corresponding to the latent space data points. For example, this can be used to verify the clustering goodness in the latent space.

[0094] For example, in some embodiments, the unsupervised training module 128 may use contour criteria (e.g., contour scores) to determine reference locations for latent space cardiac health parameters. The unsupervised training module 128 may determine contour criteria to identify the clustering goodness of latent space cardiac health parameters and / or their reference locations. Contour criteria may include measures for evaluating the effectiveness and quality (e.g., goodness of clustering) of a clustering algorithm. Each reference location may obtain a contour score based on the contour criteria. The contour score indicates how similar a reference location is to its own cluster compared to other clusters. Contour scores may range from -1 to 1, where high values ​​indicate a good match between the reference location and its own cluster, and a poor match with nearby or adjacent clusters.

[0095] Contour Score Can be given as in Let represent the average distance from the i-th data point to other data points in the same cluster, and This represents the minimum average distance from the i-th data point to data points in different clusters, which is minimized in the clustering process. The overall profile score of the cluster can be the average of the profile scores for each instance.

[0096] A score close to 1 indicates a good match between the reference location and its own cluster, but a poor match with neighboring clusters. A score around 0 indicates overlapping clusters. A score close to -1 indicates that the reference location may have been assigned to an incorrect cluster. The unsupervised training module 128 can use contour criteria to evaluate how well-defined and distinct the clusters corresponding to latent space cardiac health parameters are. In some embodiments, the endogenous analysis system 114 can automatically test different numbers of clusters and use contour scores to identify the best or optimal number of clusters (e.g., the maximum number of clusters) associated with the reference location. FIG. 6A The graph shows the results of calculating the clustering goodness of clusters in the latent space, ranging from 2 to 14 clusters. As shown, for the input data used (e.g., input data 550), the maximum score, or maximum number of clusters, was determined to be 4 clusters. This maximum score indicates that the optimal number of clusters is likely 4.

[0097] The unsupervised training module 128 can associate each of the reference arterial pressure signal waveforms with a corresponding cluster in the latent space. This can be done based on the previously identified optimal number of clusters. In this way, the unsupervised training module 128 can determine a set of clusters, or reference clusters, which can be used to test future latent space cardiac health parameters.

[0098] The unsupervised training module 128 can determine the cluster associated with each of the reference arterial pressure signal waveforms in the latent space based on a set of clusters and / or the reference position of each of the multiple reference arterial pressure signal waveforms in the latent space. The endogenous pattern analysis system 114 can then use the clustering information associated with the reference arterial pressure signal waveforms to determine the corresponding hypotensive endogenous pattern for each of the reference patients. In some embodiments, the unsupervised training module 128 does not determine any clustering evaluation criteria other than profile criteria.

[0099] In some embodiments, the unsupervised training module 128 may additionally or alternatively calculate the Calinski-Harabasz index to evaluate the goodness of clustering in the clustering algorithm. The Calinski-Harabasz (CH) index, or variance ratio criterion, describes the ratio of between-cluster variance to within-cluster variance.

[0100] The CH index can be given by the following formula. A higher CH index indicates more defined and independent clusters. The internal model analysis system 114 calculates the CH index for different cluster numbers (k values) at a reference location and selects the number of clusters that maximizes the index. Peaks in the CH index may indicate the optimal number of clusters. In some embodiments, the internal model analysis system 114 compares the results of the profile criteria and the CH index to determine the optimal or best number of clusters.

[0101] In some embodiments, the unsupervised training module 128 may additionally or alternatively calculate the Davies-Bouldin index (DBI) to generate a clustering validation metric. The DBI can also be used to evaluate the clustering quality between reference locations. The DBI may include indicators of the tightness and separation between clusters. Generally, a lower Davies-Bouldin index indicates better clustering.

[0102] The DBI can be calculated by finding the cluster centroid of each reference location cluster. Each cluster has its own centroid. The centroid can be considered as the average of all data points in that cluster. The unsupervised training module 128 can calculate the similarity between each cluster and every other cluster. This similarity can be based on a distance metric (e.g., Euclidean distance) and can represent how close or similar the clusters are.

[0103] For each cluster, it is possible to compute It is given by the following formula in Clustering The density, Clustering The tightness, and Clustering and clustering The separation degree between them. Therefore, DBI can be given by the following formula. The unsupervised training module 128 can use any combination of the above-described criteria and / or other criteria to determine the clustering goodness of the reference location and determine the optimal number of clusters for the reference location based on the goodness of ...

[0104] In some embodiments, the unsupervised training module 128 may additionally or alternatively use a Gaussian mixture model (GMM). A GMM is a probabilistic model that can represent a mixture of multiple Gaussian (e.g., normal) distributions. Like the other criteria and indices described above, GMMs can be used to estimate cluster size or goodness and—or density information associated with each cluster. The unsupervised training module 128 may compute the overall distribution of the data as a combination of several Gaussian distributions, each associated with a cluster at a different reference location.

[0105] Unsupervised training module 128 can identify the number of clusters for the model and compute the mean, covariance, and / or weights of each cluster. These can correspond to the center, shape, and relative importance of each Gaussian distribution. Unsupervised training module 128 can then generate Expectation Maximization (EM). To do this, unsupervised training module 128 can determine the probability that each reference location belongs to a specific cluster by computing the Bayesian posterior probability. Unsupervised training module 128 can maximize the parameters (e.g., mean, covariance, and weights) of each cluster to maximize the probability of reference locations in the latent space. Unsupervised training module 128 can perform a weighted average of the reference locations based on the obtained posterior probabilities. This process can be iterated until convergence is reached. Convergence can be determined when the parameter changes between iterations fall below a convergence threshold and / or after reaching the maximum number of iterations. Based on the degree of convergence, unsupervised training module 128 can assign each location (e.g., a reference location) in the latent space to a specific cluster based on the highest posterior probability.

[0106] In some embodiments, the endogenous pattern analysis system 114 (e.g., machine learning model 124) may further decode latent space cardiac health parameters back to corresponding cardiac health parameters. Machine learning model 124 may include fully connected deep learning models, convolutional deep learning models, and / or some other type of learning model. In some embodiments, one or more of these cardiac health parameters may be sent to a graphical user interface 132 for display to healthcare professionals.

[0107] In some embodiments, the intrinsic type analysis system 114 can verify the clustering goodness in the latent space by determining the clustering goodness in the original space and comparing two scores. If the clustering goodness in the original space and the latent space indicates the same optimal or best number of clusters, then the intrinsic type analysis system 114 can verify the best number of clusters in the latent space. The intrinsic type analysis system 114 can verify the best number of clusters in the latent space by applying one or more of the clustering evaluation metrics described herein to determine the best number of clusters in the original space. Therefore, the latent space cardiac health parameters can be decoded into cardiac health parameters in the original space to verify the encoding of the machine learning model 124. FIG. 6B The graph shows an example evaluation of the clustering goodness of a set of locations in the original space using a clustering evaluation metric. As shown, the optimal number of clusters obtained from the evaluation is four. When comparing... FIG. 6B and FIG. 6A At this point, it can be observed that the original space and latent space of the clustering both reach the same optimal number of clusters. Therefore, FIG. 6B Can be used in FIG. 6A Validation of the optimal number of clusters found.

[0108] FIG. 6CA graph illustrating example results of arterial pressure signal waveform clustering is provided. As shown, multiple locations are plotted within Cartesian coordinates of cardiac health parameters in the first and second latent spaces. Such graphs could be plotted in 3, 4, 5, 6, or more dimensions, but for clarity, only one is shown. FIG. 6C Two latent space cardiac health parameters are used. As shown in the figure, based on the above clustering, each location is associated with a corresponding cluster in the four clusters. The four clusters are... FIG. 6A-6B The results of the clustering goodness-of-fact assessment are shown. Once the optimal or best number of clusters has been identified, the endogenous pattern analysis system 114 can determine which locations are associated with which clusters (e.g., the relative position of each location within the latent space). As shown, the number of reference patients associated with the first cluster is 2434, with the second cluster 1851, with the third cluster 1635, and with the fourth cluster 928. Each cluster is associated with a corresponding hypotensive endogenous pattern.

[0109] FIG. 6D Multiple box-and-whisker plots show example results of a trained machine learning model configured to diagnose or determine the hypotensive subtype in a test patient. As shown, four hypotensive subtypes have been identified. These four subtypes are identified using the reference above. FIG. 6A-6C The clustering optimization described is used to determine this. FIG. 6D The names of the four identified hypotensive intraocular types are shown: vasodilation, myocardial depression, bradycardia, and hypovolemia. For each identified intraocular type, a box-and-whisker plot indicates the normalized distribution of five cardiac health parameters (in the original space) in reference patients associated with the corresponding intraocular type. These five cardiac health parameters (strike volume index (SVI), heart rate (HR), cardiac index (CI), systemic vascular resistance index (SVRI), and stroke volume variability (SVV)) are used to... FIG. 6A-6D The example analysis shown calculates five cardiac health parameters for the endotype. In some embodiments, the system does not use any other cardiac health parameters besides these five.

[0110] As shown in the figure, the 25th–75th percentile distribution of each of the five cardiac health parameters indicates that vasodilators tend to have lower-than-average systemic vascular resistance index (SVRI) and stroke volume variability (SVV). These factors suggest that, in vasodilators, vasoconstriction and / or additional hydration of the arteries and veins may benefit any diagnosis or potential diagnosis of hypotension.

[0111] As shown in the figure, the 25th–75th percentile distribution of each of the five cardiac health parameters indicates that patients with myocardial depression tend to have below-average stroke volume index (SVI) and cardiac index (CI), but normal stroke volume variability (SVV). These factors suggest that patients with myocardial depression may require vasopressors to constrict blood vessels and increase blood pressure, and / or positive inotropic agents to improve cardiac contractility.

[0112] As shown in the figure, the 25th–75th percentile distribution of each of the five cardiac health parameters indicates that patients with bradycardia tend to have below-average heart rate (HR), cardiac index (CI), and stroke volume variability (SVV). These factors suggest that patients with bradycardia may require atropine to block the effects of the vagus nerve and / or temporary or permanent pacing of the cardiac electrical conduction system. In some cases, dopamine and / or norepinephrine may be used.

[0113] As shown in the figure, the 25th–75th percentile distribution of each of the five cardiac health parameters indicates that patients with hypovolemia tend to have below-average stroke volume index (SVI) and above-average heart rate (HR) and stroke volume variability (SVV). These factors suggest that patients with hypovolemia may require intravenous fluid administration (e.g., saline), may need blood transfusions, and / or may need vasopressors to raise blood pressure.

[0114] As described in this article, FIG. 6A-6D One or more of these analysis results may be sent as alerts to the user interface (e.g., graphical user interface 132). For example, in some embodiments, identified intrinsic patterns (e.g., vasodilation, myocardial depression, bradycardia, hypovolemia, etc.) may be displayed as alerts. Additionally or alternatively, alerts may include one or more identified cardiac health parameters (e.g., SVI, HR, CI, SVRI, SVV). For example, in some embodiments, the display may also include similar... FIG. 6D The box-and-whisker plot contains one or more cardiac health parameters. Whether and / or which cardiac health parameters are displayed compared to healthy patients, patients with low blood pressure, and / or other patients with a defined hypotensive pattern may depend on the cardiac health parameters that require the most urgent attention and / or deviate the most from the normalized value. Therefore, the displayed cardiac health parameters may depend on whether the corresponding cardiac health parameter meets or exceeds a threshold.

[0115] The following text is for reference only. FIG. 6E-6K Describe the additional user interface.

[0116] Thresholds may include urgency thresholds (e.g., time thresholds). Additionally or alternatively, thresholds may include limiting thresholds (e.g., normalized value thresholds, normalized value deviation thresholds). Example limiting thresholds may include a given cardiac health parameter versus a defined reference patient group (such as...). FIG. 6D The degree of deviation from the 25th to 75th percentiles of the patients shown. In some embodiments, the alarm may include auditory indicators (e.g., one or more tones, impulse tones, increased volume tones), visual indicators (e.g., flashing text, pictures or other indicators; enlarged or highlighted text or pictures), and / or tactile indicators (e.g., vibration at a specific frequency, changes in vibration frequency or duration, vibration patterns, etc.). Other types of alarms described above are also possible.

[0117] FIG. 6E-6K This describes an example user interface that a healthcare professional can use to analyze medical information associated with a patient. In some embodiments, the user interface may represent the graphical user interface 132 described above. The user interface may be presented via a system display such as a computer, laptop, tablet, mobile device, etc. As will be described, the user interface may respond to user input, allowing a healthcare professional to interact with the user interface. For example, a healthcare professional may provide input to view detailed information associated with a specific patient. As another example, a healthcare professional may provide input to view different visualizations of a patient or a specific patient's internal profile. For this example, a healthcare professional can... FIG. 6E-6K The user interface described loops between these elements.

[0118] FIG. 6E-6K This section describes techniques for visualizing (e.g., in the context of endotype clustering previously obtained from a database of multiple patients) patient-associated, time-varying endotype trends. As described above, patient-associated measurements or input data can be obtained at different times. Example measurements or input data described herein may include hemodynamic signal measurements, cardiac health parameters, arterial pressure signal waveforms, etc. These measurements or input data may be obtained at periodic or non-periodic intervals (e.g., every 5 seconds, every 2 hours, several times a day, several times a week, etc.). Using these measurements or input data, for example... FIG. 1 As shown, the current hypotensive pattern of the patient can be determined. In this way, the current patient's hypotensive blood pressure can be monitored over time. As can be understood, the characteristics of hypotensive events can be adjusted over time, so that the determined patient pattern can be adjusted accordingly. FIG. 6E-6K This paper describes different techniques that allow such intrinsic changes (referred to as intrinsic trends in this paper) over time to be presented to healthcare professionals in a concise and essentially real-time manner.

[0119] In some embodiments, a healthcare professional can use data stored for a specific patient to select patients for detailed examination. For example, the healthcare professional can provide input indicating the patient's condition (e.g., the patient's name, ID number, etc.). The system described herein can access latent space cardiac health parameters of a patient determined at different time points. For example, latent space cardiac health parameters can be selected within a previous threshold time period. In this example, the healthcare professional can optionally indicate a threshold time period, allowing the healthcare professional to view the patient's determined endotype during a threshold time period (e.g., the previous hour, the previous day, the previous week, etc.).

[0120] Latent space cardiac health parameters can be presented in a user interface as endotype trends associated with the patient within the context of previously determined endotype clusters and their associated probabilities, such as those identified previously (e.g., from a database of multiple patients). For example, endotype trends can be graphically illustrated in real-time or in offline examinations over previously threshold time periods, illustrating the patient's determined endotype. In this way, and as an example, the user interface can reflect that the patient begins with vasodilation and ends with hypovolemia. As another example, the user interface can reflect that the patient begins with vasodilation, with a lower probability (e.g., likelihood) of bradycardia, hypovolemia, and myocardial depression. For this example, the user interface can additionally reflect that over time, the patient still has vasodilation, but a higher probability of bradycardia, hypovolemia, or myocardial depression.

[0121] In some embodiments, the techniques described herein can be used to determine intratype clusters. Thus, information associated with a large number of patients can be used to form clusters. Information from new patients can then be accessed. For example, medical professionals can obtain measurements associated with patients. These measurements can then be mapped to latent space cardiac health parameters. Therefore, in some embodiments, latent space cardiac health parameters can represent new information. The user interface described below can present (1) previously determined clusters and (2) new information about the patient. For example, newly determined latent space cardiac health parameters can be mapped to previously determined clusters (e.g., data points can be added to clusters, where each data point represents a combination of latent space cardiac health parameters determined for the patient at a specific time).

[0122] Although FIG. 6E-6K Example graphical representations illustrating the endotype trends of patients over time are provided, but in some embodiments, the user interface may include textual information. For example, the user interface may include a text summary describing the patient's endotype trends and clusters previously identified from a database. As an example, the text summary may describe and optionally include quantitative information about the patient having less vasodilation and more myocardial inhibition over time.

[0123] FIG. 6EAn example user interface 650 illustrates the endotype trend 652 associated with a patient. In the illustrated example, two latent dimensions are included as their respective axes. Individual data points and / or time points (e.g., data points 654A-654B) for each patient are plotted in the user interface 650. These data points (e.g., time points) may represent latent space cardiac health parameters corresponding to the patient, acquired at a specific time or in real time. The user interface 650 includes a large number of such data points from a large number of patients in a database. Patients may share one or more similarities or characteristics with the current patient. Data points from patients in the database are clustered into different endotypes (also known as data point clustering), such that a first portion of the data points is clustered into a vasodilatory endotype (e.g., data point 654A), a second portion into a hypovolemic endotype (e.g., 654B), a third portion into a bradycardia endotype, and a fourth portion into a myocardial depression endotype.

[0124] Regarding latent space cardiac health parameters (e.g., latent space dimensions), in the illustrated example, each parameter may range from -1 to 1. In other embodiments, the range of values ​​may differ. As described above, each data point can be mapped into this latent space based on input data associated with the patient. The data point can then be assigned an intrinsic type, for example, based on the probability that the assigned intrinsic type exceeds the probability of remaining intrinsic types determined based on the patient's input data in the database. In some embodiments, this probability can be determined based on the input data, and optionally, clustering techniques can be used to determine it.

[0125] Therefore, each of the data points depicted in the user interface 650 can be assigned a corresponding internal type. FIG. 6E Different colors are used to visually represent data points from the database that have been assigned endotypes. For example, green can be used to represent vasodilation, red to represent hypovolemia, purple to represent bradycardia, and cyan to represent myocardial depression. In some embodiments, a medical professional using the user interface 650 can customize the colors corresponding to the endotypes. Optionally, the user interface can use other visual markers (such as patterns) to represent endotypes. In this way, the user interface 650 graphically illustrates the clustering of data points in the database associated with different endotypes.

[0126] User interface 650 presents the aforementioned inner-pattern trend 652 that changes over time. In this example, the inner-pattern trend is formed by four data points (e.g., data points 652A-652D) associated with the current patient. Graphical elements (e.g., stars, circles, squares, dots, etc.) can be presented for these data points 652A-652D. In some embodiments, different graphic elements can be used. To represent the passage of time, the graphic elements have different inner colors. For example, the most recent data point can be colored opaquely (e.g., pure red). Data points that are earlier in time can be colored more transparently. For example, the innermost part of the first star (e.g., closer to the top of the graph) can be substantially transparent. Similarly, proximity can be used to adjust the color to a specific color (e.g., white) instead of transparency.

[0127] Therefore, the user interface 650 makes clustering operational, as medical professionals can determine a patient's position over time within the user interface 650 within the context provided by the clustering of data points from the database. For example, endotype trend 652 can indicate that a patient is shifting from vasodilation to a different endotype.

[0128] In some embodiments, the user interface 650 may respond to user input selecting one of the data points 652A-652D that form an intrinsic trend 652. Upon selection, the user interface 650 may be updated to include detailed measurements or medical information associated with the patient at a specific time. In this way, healthcare professionals can easily access detailed patient medical information.

[0129] FIG. 6F Another example user interface 660 illustrates the intrapattern trend 652 associated with the current patient. As illustrated, user interface 660 includes various shapes (e.g., ellipses 662). As will be described, each ellipse may include a subset of the aforementioned data points from a database, these data points having a substantially similar probability of being assigned a particular intrapattern. Although shown as ellipses in the example, it is understood that other shapes may be used. For example, the shape may follow arbitrary contour lines. In this example, arbitrary contour lines can be used to define which data points are associated with a similar probability of being assigned a particular intrapattern. Thus, contour lines may define endpoints of arbitrary shapes (e.g., data blocks). In some embodiments, user interface 660 may be accessed based on user input provided to user interface 650. For example, the aforementioned healthcare professional may toggle whether ellipses are included. User interface 660 may also be accessed directly, for example, without viewing user interface 660.

[0130] Ellipse 662 is illustrated as a first portion containing data points assigned to the low-volume endotype (e.g., from a database). Ellipse 662 may correspond to a probability region indicating the likelihood of a corresponding endotype associated with a cluster. These data points may be associated with similar assignment probabilities. For example, data points may be within each other's threshold probability percentages. As another example, the ellipse may represent a range of probabilities, and data points may extend within that range.

[0131] Ellipse 664 is illustrated as a second portion containing data points from a database. For example, the second portion may extend between 664 and 666. In the illustrated example, some of the data points in the second portion are assigned to the vasodilatory intratype. These data points may have a similar probability of being assigned to the hypovolemic intratype. However, data points may also have a higher probability of being assigned to the vasodilatory intratype. That is, the system described herein (e.g., intratype analysis system 114) can determine the probability of data points reflecting each of the intratypes. Therefore, FIG. 6F The lower part of the ellipse 664 reflects the highest probability of hypovolemia, while the upper part reflects the highest probability of vasodilation.

[0132] As an example, medical professionals can use ellipses to determine overlaps between in-type clusters. Regarding in-type trend 652, a medical professional can determine that the current (e.g., most recent) data point 652D overlaps between vasodilation and bradycardia. In this way, a medical professional can determine an increased likelihood of bradycardia.

[0133] FIG. 6G Another example user interface 670 illustrates an endogenous trend 652 associated with a patient. In the illustrated example, (e.g., from user interface 650) the color associated with data points from the database is removed. Instead, a basic solid color (e.g., gray) is included. Removing color helps healthcare professionals understand the endogenous trend 652 (e.g., the movement of a patient's map into the latent space over time).

[0134] FIG. 6H Another example user interface 680 illustrates the intramorphic trend 652 associated with a patient. In the illustrated example, (e.g., from user interface 660) the color associated with data points from a database is removed. Similar to the above, removal helps healthcare professionals understand the intramorphic trend 652. User interface 680 includes ellipses from user interface 660 and their associated intramorphic colors.

[0135] about FIG. 6E-6HIn some embodiments, the time-varying inner shape trend 692 may represent a line or vector. Optionally, a portion of the line or vector may be assigned a color based on the inner shape determined for the data points forming the trend 692. For example, the portion of the line or vector representing an earlier data point in time may be green. As the line or vector gets closer to the current (e.g., most recent) data point, the line or vector may begin to turn purple.

[0136] FIG. 6I Another example user interface 690 illustrates an endogenous trend 692 over time associated with the current patient. In the illustrated example, the endogenous trend 692 is a trend line extending between A and B. Data points A and B may represent the earliest and most recent data points associated with the patient. The data points forming the trend line may be correlated with time between the data points of A and B. The trend line can be formed using various techniques, such as mapping different linear and / or nonlinear functions to the location of the data points associated with the patient.

[0137] FIG. 6J Another example user interface 692 illustrates the intratype trend 652 associated with the current patient. In the illustrated example, similar to... FIG. 6E-6I The internal trend 652 illustrates another example group of time points. A time point can be associated with an indicator that indicates a time point relative to other indicators among a plurality of indicators. One or more of the indicators can indicate the relative time at which the time point was obtained. For example, an earlier time point could be a more transparent, different color, and / or include some other visual indicators that may differ from a later time point. Multiple degrees or levels of difference may exist between different time points.

[0138] FIG. 6K Explanation and FIG. 6J Another example of a user interface 694 is the multiple trends of various attributes associated with different time points. For example, user interface 694 may include time points associated with: the patient's MAP, HPI, SVR, SVV, harmonics associated with the location of the latent space point, SV, HR, CO, dpdt, and / or Gaussian distributions associated with the location of the latent space point. For example, harmonics may refer to the harmonic probabilities associated with multiple locations in the latent space of multiple arterial pressure signal waveforms from the patient. Additionally or alternatively, Gaussian distributions may refer to the Gaussian distributions associated with multiple locations in the latent space of multiple arterial pressure signal waveforms.

[0139] Example method of intrinsica determination The example routines or methods in this document illustrate various implementations of the system described herein. The boxes in the routines illustrate example implementations, and in various other implementations, the boxes may be rearranged and / or presented as optional. Additionally, boxes may be omitted from and / or added to the example routines below, and boxes may be moved between various example routines.

[0140] FIG. 7 A flowchart of an example method 700 for determining a patient's hypotensive endotype is shown. Method 700 may be performed by one or more systems described herein (e.g., hemodynamic sensing system 100, endotype analysis system 114, machine learning model 124, unsupervised training module 128, autoencoder model 500, etc.) and / or combinations thereof. At block 704, the system may receive simulated hemodynamic sensor signals from the patient from hemodynamic sensors. The simulated hemodynamic sensor signals may correspond to signals sensed by hemodynamic sensors (e.g., hemodynamic sensor 108, hemodynamic sensor 300, hemodynamic sensor 426).

[0141] At box 708, the system can convert the simulated hemodynamic sensor signal into an arterial pressure signal waveform. The arterial pressure signal waveform may correspond to the arterial pressure signal waveform described herein, such as arterial pressure signal waveform 200. At box 712, the system can extract multiple cardiac health parameters from the arterial pressure signal waveform. These cardiac health parameters may correspond to any cardiac health parameters described herein, such as cardiac output (CO), stroke volume (SV), stroke volume variability (SVV), diastolic blood pressure (DIA), pulse rate (PR), stroke volume index (SVI), systemic vascular resistance (SVR), mean arterial pressure (MAP), HPI, systemic vascular resistance index (SVRI), cardiac index (CI), systolic blood pressure (SYS), and / or any other parameters described herein. In some embodiments, the system can extract a combination of SVI, HR, CI, SVRI, and SVV from the arterial pressure signal waveform, but other combinations are also possible.

[0142] At box 716, the system can use a fully connected deep learning model to encode multiple cardiac health parameters into one or more latent space cardiac health parameters. The fully connected deep learning model may include an autoencoder model 500, a machine learning model 124, and / or include one or more features thereon. The number of latent space cardiac health parameters may be less than the number of cardiac health parameters extracted from the arterial pressure signal waveform. In some embodiments, the number of latent space cardiac health parameters is two.

[0143] At box 720, the system can generate the position of the arterial pressure signal waveform in the latent space. This position can be based on encoded latent space cardiac health parameters, such as their values. For example, if there are two latent space cardiac health parameters, then the position can correspond to a coordinate position in the Cartesian plane. At box 724, the system can determine the relative position of the arterial pressure signal waveform in the latent space based on the position in the latent space. The relative position can indicate the cluster associated with the arterial pressure signal waveform. The relative position and / or the associated cluster can be determined by an unsupervised training module (e.g., unsupervised training module 128). The relative position and / or the associated cluster can be determined using one or more clustering evaluation metrics or criteria described herein (e.g., contour criteria, Calinski-Harabasz index, and Davies-Bouldin index, etc.). The clustering evaluation metrics or criteria can be configured to indicate the goodness of clustering at a reference position.

[0144] At box 728, the system may determine a patient's hypotensive endotype based on the relative position of the arterial pressure signal waveform and / or associated clustering. In some embodiments, the system may generate data based on the patient's determined hypotensive endotype to display alarms indicating one or more of the patient's hypotensive endotype and / or cardiac health parameters.

[0145] FIG. 8 A flowchart of an example method 800 for training a model to determine a patient’s hypotensive endotype is shown. Method 800 may be performed by one or more systems described herein (e.g., hemodynamic sensing system 100, endotype analysis system 114, machine learning model 124, unsupervised training module 128, autoencoder model 500, etc.) and / or combinations thereof.

[0146] At box 804, the system can receive multiple reference hemodynamic sensor signals from multiple reference patients. At box 808, the system can extract a corresponding set of reference cardiac health parameters from the reference hemodynamic sensor signals. The reference cardiac health parameters may include those mentioned above. FIG. 7 One or more cardiac health parameters are described. At box 812, the system can use a fully connected deep learning model to encode multiple sets of cardiac health parameter reference values ​​into multiple corresponding reference latent space cardiac health parameters. At box 816, the reference latent space cardiac health parameters can be used by the system to generate a reference location in the latent space for each of multiple reference arterial pressure signal waveforms based on the multiple reference latent space cardiac health parameters.

[0147] At box 820, the system can determine a cluster associated with each of the reference arterial pressure signal waveforms in the latent space. Each cluster can correspond to the relative position of the arterial pressure signal waveform in the latent space. Each cluster can be determined by a clustering evaluation metric, such as those described above, that determines each of the reference positions in the latent space. The system can determine an optimal number of clusters based on the clustering evaluation metric. In some embodiments, the system can determine a set of clusters by associating each of the reference arterial pressure signal waveforms with a corresponding cluster in the latent space based on the optimal number of clusters.

[0148] At box 824, the system may determine the hypotension endotype of each of the reference patients based on clusters associated with each of the reference arterial pressure signal waveforms. Determining the hypotension endotype may include determining the clusters associated with each of the reference arterial pressure signal waveforms in the latent space based on a set of clusters and / or the reference position of each of the multiple reference arterial pressure signal waveforms in the latent space. In some embodiments, the system may use a fully connected deep learning model to decode one or more reference latent space cardiac health parameters into corresponding sets of multiple cardiac health parameter reference values. Based on the determined hypotension endotype, the system may generate a trained model for determining the hypotension endotype of patients (such as test patients in ICU, OR, and / or other patient care settings).

[0149] In some embodiments, the system includes various features presented as a single feature (as opposed to multiple features). For example, in one embodiment, the system includes a single hemodynamic sensor 108 with a single signal converter 112, a single processor 116 and a single memory 120, a single machine learning model 124 and a single unsupervised training module 128, as described herein. In alternative embodiments, multiple features or components are provided.

[0150] In some embodiments, the system (e.g., hemodynamic sensing system 100) includes one or more of the following: means for sensing cardiac pressure (e.g., hemodynamic sensor 108, hemodynamic sensor 300, hemodynamic sensor 426), means for converting the sensed data (e.g., signal converter 112), and / or means for displaying the data (e.g., graphical user interface 132).

[0151] Additional example embodiments and details related to computing systems In some embodiments, the system described herein may include or be implemented in a “virtual computing environment.” As used herein, the term “virtual computing environment” should be broadly interpreted to include, for example, computer-readable program instructions executed by one or more processors to implement one or more aspects of the modules and / or functions described herein. Additionally, in this embodiment, one or more services / modules / engines of the system may be understood to include one or more rule engines of the virtual computing environment that, in response to input received through the virtual computing environment, execute rules and / or other program instructions to modify the operation of the virtual computing environment. For example, a request received from a user computing device may be understood to modify the operation of the virtual computing environment to allow the request to access resources from the system. Such functionality may include modifying the operation of the virtual computing environment in response to input and according to various rules. Other functions implemented by the virtual computing environment (as described throughout this disclosure) may further include modifications to the operation of the virtual computing environment, for example, the operation of the virtual computing environment may be changed based on information collected through the system. Initial operation of the virtual computing environment may be understood to establish the virtual computing environment. In some embodiments, the virtual computing environment may include one or more virtual machines, containers, and / or other types of emulation of computing systems or environments. In some implementations, a virtual computing environment may include a hosted computing environment, which includes a collection of physical computing resources that can be remotely accessed and rapidly provisioned as needed (often referred to as a "cloud" computing environment).

[0152] Implementing one or more aspects of a system as a virtual computing environment advantageously enables the execution of different aspects or modules of the system on different computing devices or processors, which increases the system's scalability. Implementing one or more aspects of a system as a virtual computing environment further advantageously enables the sandboxing of various aspects, data, or services / modules of the system with each other, which increases the system's security by preventing, for example, the propagation of malicious intrusions into the system. Implementing one or more aspects of a system as a virtual computing environment further advantageously enables the parallel execution of various aspects or modules of the system, which increases the system's scalability. Implementing one or more aspects of a system as a virtual computing environment further advantageously enables the rapid provisioning (or deprovisioning) of computing resources to the system, which increases the system's scalability by, for example, expanding the computing resources available to the system or replicating the system's operations across multiple computing resources. For example, the system can be used simultaneously by thousands, hundreds of thousands, or even millions of users, and can be used by the system to transfer or process megabytes, gigabytes, or terabytes (or more) of data, and the system's scalability allows for such operations to be performed in an efficient and / or uninterrupted manner.

[0153] Various embodiments of this disclosure can be systems, methods, and / or computer program products at any possible level of integration technical detail. A computer program product may include a computer-readable storage medium (or media) having computer-readable program instructions thereon for causing a processor to execute aspects of this disclosure.

[0154] For example, the functions described herein may be executed when software instructions are executed by one or more hardware processors and / or any other suitable computing device, and / or in response to the execution of the software instructions. The software instructions and / or other executable code may be read from a computer-readable storage medium (or media). A computer-readable storage medium may also be referred to herein as a computer-readable storage device or a computer-readable storage apparatus.

[0155] Computer-readable storage media can be tangible devices that retain and store data and / or instructions for use by an instruction execution device. Computer-readable storage media can be, for example, but not limited to, electronic storage devices (including any volatile and / or non-volatile electronic storage devices), magnetic storage devices, optical storage devices, electromagnetic storage devices, semiconductor storage devices, or any suitable combination of the foregoing. A non-exhaustive list of more specific examples of computer-readable storage media includes the following: portable computer disks, hard disks, solid-state drives, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), static random access memory (SRAM), portable optical disc read-only memory (CD-ROM), digital multi-disc (DVD), memory sticks, floppy disks, mechanical encoding devices such as punch cards or raised structures in recesses on which instructions are recorded, and any suitable combination of the foregoing. As used herein, computer-readable storage media should not be construed as transient signals, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through waveguides or other transmitting media (e.g., light pulses through fiber optic cables), or electrical signals transmitted through wires.

[0156] The computer-readable program instructions described herein can be downloaded from a computer-readable storage medium to a suitable computing / processing device or to an external computer or external storage device via a network (e.g., the Internet, a local area network, a wide area network, and / or a wireless network). The network may include copper transmission cables, optical transmission fibers, wireless transmissions, routers, firewalls, switches, gateway computers, and / or edge servers. A network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards them to a computer-readable storage medium within the suitable computing / processing device.

[0157] Computer-readable program instructions (also referred to herein, for example, "code," "instructions," "module," "application," "software application," etc.) used to perform the operations of this disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, status setting data, integrated circuit configuration data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Smalltalk, C++, etc., and procedural programming languages ​​such as the "C" programming language or similar programming languages. Computer-readable program instructions may be invoked from other instructions or themselves, and / or may be invoked in response to a detected event or interrupt. Computer-readable program instructions configured for execution on a computing device may be provided on a computer-readable storage medium, and / or as a digital download (and may be initially stored in a compressed or installable format that requires installation, decompression, or decryption prior to execution), and may then be stored on a computer-readable storage medium. Such computer-readable program instructions may be stored, in part or in whole, on a memory device (e.g., a computer-readable storage medium) for execution by the computing device. Computer-readable program instructions may be executed entirely on a user's computer (e.g., a computing device in execution), partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the latter case, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet provided by an Internet service provider). In some embodiments, electronic circuitry including, for example, programmable logic circuitry, field-programmable gate arrays (FPGAs), or programmable logic arrays (PLAs) may execute computer-readable program instructions by personalizing the electronic circuitry with state information from the computer-readable program instructions in order to perform aspects of this disclosure.

[0158] This document describes aspects of the disclosure with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the disclosure. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer-readable program instructions.

[0159] These computer-readable program instructions may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions / actions specified in one or more blocks of a flowchart and / or block diagram. These computer-readable program instructions may also be stored in a computer-readable storage medium that directs a computer, programmable data processing apparatus, and / or other device to operate in a particular manner, such that the computer-readable storage medium storing the instructions comprises an article of manufacture including instructions that implement aspects of the functions / actions specified in one or more blocks of a flowchart and / or block diagram.

[0160] Computer-readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device to cause a series of operational steps to be performed on the computer, other programmable apparatus, or other device to produce a computer-implemented process, such that the instructions, which execute on the computer, other programmable apparatus, or other device, implement the functions / actions specified in one or more boxes of a flowchart and / or block diagram. For example, the instructions may initially be carried on a hard disk or solid-state drive of a remote computer. The remote computer may load the instructions and / or modules into its dynamic memory and transmit the instructions via telephone, cable, or optical line using a modem. A modem local to the server computing system may receive data on a telephone / cable / optical line and place the data on a bus using a converter device including appropriate circuitry. The bus may carry the data into memory from which the processor may retrieve and execute the instructions. Instructions received from memory may optionally be stored on a storage device (e.g., a solid-state drive) before or after execution by the computer processor.

[0161] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible embodiments of the systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a segment or portion of a service, module, or instruction, including one or more executable instructions for implementing a specified logical function. In some alternative embodiments, the functions marked in the blocks may occur in a different order than indicated in the drawings. For example, two blocks shown consecutively may actually be executed substantially simultaneously, or these blocks may sometimes be executed in reverse order, depending on the functions involved. Furthermore, in some embodiments, certain blocks may be omitted or are optional. The methods and processes described herein are not limited to any particular sequence, and the blocks or states associated with them may be performed in other suitable sequences.

[0162] It should also be noted that each box in the block diagram and / or flowchart illustrations, as well as combinations of boxes in the block diagram and / or flowchart illustrations, can be implemented by systems based on dedicated hardware that can perform specified functions or actions or combine dedicated hardware with computer instructions. For example, any of the processes, methods, algorithms, elements, boxes, applications, or other functions (or parts of functions) described in the preceding sections can be embodied in and / or fully or partially automated via electronic hardware, such as dedicated processors (e.g., application-specific integrated circuits (ASICs)), programmable processors (e.g., field-programmable gate arrays (FPGAs)), dedicated circuits, etc. (any of which can also combine custom hardwired logic, logic circuits, ASICs, FPGAs, etc. with custom programming / execution of software instructions to implement these technologies).

[0163] Any of the processors described above and / or a device containing any of the processors described above may be referred to herein as, for example, "computer," "computer device," "computing device," "hardware computing device," "hardware processor," "processing unit," etc. The computing device of the above embodiments is typically (but not necessarily) controlled and / or coordinated by operating system software such as Mac OS, iOS, Android, Chrome OS, Windows OS (e.g., Windows XP, Windows Vista, Windows 7, Windows 8, Windows 10, Windows 11, Windows Server, etc.), Windows CE, Unix, Linux, SunOS, Solaris, Blackberry OS, VxWorks, or other suitable operating systems. In other embodiments, the computing device may be controlled by a proprietary operating system. Conventional operating systems control and schedule computer processes for execution, perform memory management, provide file systems, networking, I / O services, and provide user interface functionality, such as a graphical user interface ("GUI").

[0164] For example, FIG. 9 This is a block diagram illustrating a computer system 900 that can be implemented in various ways. For example, in some implementations, the computer system 900 may be implemented as a hemodynamic sensing system 100. FIG. 1 FIG. 1 The computer system 900 includes a bus 902 or other communication mechanism for transmitting information, and a hardware processor or multiple processors 904 coupled to the bus 902 for processing information. The hardware processor 904 may be, for example, one or more general-purpose or special-purpose microprocessors.

[0165] Computer system 900 also includes main memory 906, such as random access memory (RAM), cache, and / or other dynamic storage devices, coupled to bus 902 for storing information and instructions to be executed by processor 904. Main memory 906 can also be used to store temporary variables or other intermediate information during the execution of instructions to be executed by processor 904. Such instructions, when stored in storage media accessible to processor 904, make computer system 900 a special-purpose machine customized to perform the operations specified in the instructions.

[0166] The computer system 900 further includes a read-only memory (ROM) 908 or other static storage devices coupled to the bus 902 for storing static information and instructions for the processor 904. A storage device 910, such as a disk, optical disk, or USB thumb drive (flash drive), is provided and coupled to the bus 902 for storing information and instructions.

[0167] Computer system 900 may be coupled to display 912, such as a cathode ray tube (CRT) or LCD display (or touchscreen), via bus 902 for displaying information to a computer user. Input device 914, including alphanumeric keys and other keys, is coupled to bus 902 for transmitting information and command selections to processor 904. Another type of user input device is cursor controller 916, such as a mouse, trackball, or cursor arrow keys, for transmitting directional information and command selections to processor 904 and for controlling cursor movement on display 912. This input device typically has two degrees of freedom on two axes, namely a first axis (e.g., x) and a second axis (e.g., y), allowing the device to specify a position in a plane. In some embodiments, the same directional information and command selections as cursor control may be implemented by receiving touches on a touchscreen in the absence of a cursor.

[0168] The computing system 900 may include a user interface module implementing a GUI, which may be stored in a mass storage device as computer-executable program instructions executed by a computing device. As described below, the computer system 900 may further implement the techniques described herein using custom hardwired logic, one or more ASICs or FPGAs, firmware, and / or program logic, which, combined with the computer system, enable the computer system 900 to become a dedicated machine or to be programmed thereto. According to one embodiment, the techniques described herein are executed by the computer system 900 in response to processor 904 executing one or more sequences of one or more computer-readable program instructions contained in main memory 906. Such instructions may be read into main memory 906 from another storage medium such as storage device 910. Execution of the instruction sequence contained in main memory 906 causes processor 904 to perform the processing steps described herein. In alternative embodiments, hardwired circuitry may be used in place of or in combination with software instructions.

[0169] Various forms of computer-readable storage media can participate in carrying one or more sequences of one or more computer-readable program instructions to processor 904 for execution. For example, the instructions may initially be carried on a disk or solid-state drive of a remote computer. The remote computer may load the instructions into its dynamic memory and transmit the instructions over a telephone line using a modem. A modem local to computer system 900 may receive data over a telephone line and convert the data into an infrared signal using an infrared transmitter. An infrared detector may receive the data carried in the infrared signal, and appropriate circuitry may place the data on bus 902. Bus 902 carries the data to main memory 906, from which processor 904 retrieves and executes the instructions. The instructions received by main memory 906 may optionally be stored on storage device 910 before or after execution by processor 904.

[0170] Computer system 900 also includes a communication interface 918 coupled to bus 902. Communication interface 918 provides bidirectional data communication coupled to network link 920 connected to local network 922. For example, communication interface 918 may be an Integrated Services Digital Network (ISDN) card, a cable modem, a satellite modem, or a modem providing data communication connectivity to a corresponding type of telephone line. As another example, communication interface 918 may be a local area network (LAN) card to provide data communication connectivity to a compatible LAN (or a WAN component communicating with a WAN). Wireless links may also be implemented. In any such implementation, communication interface 918 transmits and receives electrical, electromagnetic, or optical signals carrying streams of digital data representing various types of information.

[0171] Network link 920 typically provides data communication to other data devices via one or more networks. For example, network link 920 may provide a connection via local network 922 to host computer 924 or to data equipment operated by Internet Service Provider (ISP) 926. ISP 926, in turn, provides data communication services via a global packet data communication network now commonly referred to as the “Internet” 928. Both local network 922 and Internet 928 use electrical, electromagnetic, or optical signals to carry digital data streams. Signals through various networks, as well as signals on network link 920 and through communication interface 918 (which carries digital data to and from computer system 900), are example forms of transmission media.

[0172] Computer system 900 can transmit messages and receive data, including program code, via a network, network link 920, and communication interface 918. In the Internet example, server 930 can send requested code for an application via the Internet 928, ISP 926, local network 922, and communication interface 918.

[0173] The received code may be executed by processor 904 upon receipt and / or stored in storage device 910 or other non-volatile storage device for later execution.

[0174] As described above, in various embodiments, a user can access certain functions through a web-based viewer (such as a web browser) or other suitable software program. In such embodiments, the user interface may be generated by a server computing system and sent to the user's web browser (e.g., running on the user's computing system). Alternatively, the data necessary to generate the user interface (e.g., user interface data) may be provided to the browser by the server computing system, and the user interface may be generated in the browser (e.g., the user interface data may be executed by a browser accessing a web service and may be configured to render the user interface based on the user interface data). The user can then interact with the user interface through the web browser. In some embodiments, the user interface may be accessible through one or more dedicated software applications. In some embodiments, one or more of the computing devices and / or systems of this disclosure may include mobile computing devices, and the user interface may be accessible through such mobile computing devices (e.g., smartphones and / or tablets).

[0175] Various changes and modifications can be made to the above embodiments, and their elements should be understood as existing in other acceptable examples. All such modifications and changes herein are intended to be included within the scope of this disclosure. The foregoing description details certain embodiments. However, it should be understood that the system and method can be practiced in many ways, no matter how detailed the foregoing appears in the text. Also as stated above, it should be noted that the use of specific terms when describing certain features or aspects of a system and method should not be construed as implying that the term is redefined herein as limited to any specific feature or aspect of the system and method associated with that term.

[0176] Unless otherwise specifically stated or understood in the context of use, conditional languages ​​such as “can,” “could,” “might,” or “may,” among others, are generally intended to convey that certain implementations include certain features, elements, and / or steps, while other implementations do not. Therefore, such conditional languages ​​are generally not intended to imply that one or more implementations require features, elements, and / or steps in any way, or to imply that one or more implementations must include logic for determining whether to include or perform such features, elements, and / or steps in any particular implementation, with or without user input or prompting.

[0177] The term “substantially” when used in conjunction with the term “real-time” forms a phrase readily understood by one of ordinary skill in the art. For example, it is readily understood that such language would include speeds with little or no perceptible delay or waiting, or delays so short as to not disturb, irritate, or otherwise annoy the user.

[0178] Unless otherwise specified, connective language such as the phrase "at least one of X, Y, and Z" or "at least one of X, Y, or Z" is understood, along with the context, to generally convey that an item, term, etc., may be X, Y, or Z, or a combination thereof. For example, the term "or" is used in its inclusive sense (rather than its exclusive sense), so when used, for example, to connect a list of elements, the term "or" means one, some, or all of the elements in the list. Therefore, such connective language is generally not intended to imply that some implementation requires at least one of X, at least one of Y, and at least one of Z to each be present.

[0179] As used herein, the term “a” should be interpreted inclusively rather than exclusively. For example, unless specifically indicated, the term “a” should not be construed as meaning “exactly one” or “one and only one”; rather, the term “a” means “one or more” or “at least one”, whether used in the claims or elsewhere in the specification, regardless of whether quantifiers such as “at least one,” “one or more,” or “multiple” are used elsewhere in the claims or specification.

[0180] As used herein, the term "comprising" should be interpreted in an inclusive rather than exclusive sense. For example, a computer that includes one or more processors should not be interpreted as excluding other computer components, but may include components such as memory, input / output devices, and / or network interfaces.

[0181] While the above detailed description illustrates, describes, and points out novel features applicable to various embodiments, it should be understood that various omissions, substitutions, and changes may be made to the form and details of the illustrated apparatus or process without departing from the spirit of this disclosure. It will be appreciated that certain embodiments of the invention described herein may not provide a formal embodiment of all the features and benefits set forth herein, as some features may be used or practiced separately from other features. Each of the disclosed aspects and examples of this disclosure may be considered individually or in combination with other aspects, examples, and variations of this disclosure. The headings provided herein are for readability purposes only and are not intended to limit the scope of embodiments disclosed in a particular section to the features or elements disclosed in that section. A feature or element from one embodiment of this disclosure may be adopted by other embodiments of this disclosure. For example, a feature described in one figure may be used in combination with embodiments illustrated in other figures. The foregoing description and examples are merely illustrative of this disclosure and are not intended to be restrictive. The scope of certain inventions disclosed herein is indicated by the appended claims rather than by the foregoing description. All changes within the meaning and equivalence of the claims are included within their scope.

[0182] For illustrative purposes, the following includes some example embodiments. These examples should not be considered limiting.

[0183] In the first example, a hemodynamic sensor system configured to determine and display a patient's hypotension endotype includes: a hemodynamic sensor that generates a hemodynamic sensor signal representing an arterial pressure signal waveform of the patient; a user interface; a non-transitory memory storing executable instructions; and an electronic hardware processor communicating with the non-transitory memory and configured to execute the instructions such that the system at least: receives the hemodynamic sensor signal from the patient from the hemodynamic sensor; determines the patient's hypotension endotype based on the arterial pressure signal waveform; and generates data based on the determined hypotension endotype of the patient for displaying an alarm indicating the patient's hypotension endotype on the user interface.

[0184] In the second example, according to the hemodynamic sensor system of Example 1, determining the hypotension pattern of the patient includes: extracting a plurality of cardiac health parameters from the arterial pressure signal waveform; encoding the plurality of cardiac health parameters into one or more latent space cardiac health parameters using a deep learning model; and generating the position of the arterial pressure signal waveform in the latent space using the one or more latent space cardiac health parameters.

[0185] In the third example, according to the hemodynamic sensor system of Example 2, determining the patient's hypotension endotype further includes: determining the relative position of the arterial pressure signal waveform in the latent space based on a set of clusters of reference arterial pressure signal waveforms from within the latent space, based on the position of the arterial pressure signal waveform in the latent space; determining a cluster associated with the arterial pressure signal waveform in the latent space based on the set of clusters; and determining the patient's hypotension endotype based on the clusters of the arterial pressure signal waveform.

[0186] In the fourth example, the hemodynamic sensor system according to any one of Examples 2-3, wherein the alarm includes a graphical user interface feature indicating at least one of the plurality of cardiac health parameters characterizing the patient's hypotensive endotype.

[0187] In the fifth example, according to the hemodynamic sensor system described in Example 4, at least one of the plurality of cardiac health parameters includes stroke volume index (SVI), heart rate (HR), cardiac index (CI), systemic vascular resistance index (SVRI), and stroke volume variation (SVV).

[0188] In the 6th example, the hemodynamic sensor system according to any one of Examples 1-5, wherein the hypotensive intratype is selected from the group consisting of: vasodilatory intratype, myocardial depression intratype, bradycardia intratype, and hypovolemic intratype.

[0189] In the 7th example, according to the hemodynamic sensor system of Example 1, a plurality of cardiac health parameters are extracted from the arterial pressure signal waveform, wherein the cardiac health parameters are encoded into a latent space to form latent space cardiac health parameters, wherein generating data for display includes causing the presentation of the user interface, and wherein the user interface: presents a plurality of clusters, wherein each cluster is associated with a respective endotype, and wherein each cluster is formed by latent space cardiac health parameters associated with a respective patient, and is updated based on information indicating a particular patient and presents an endotype trend associated with the particular patient.

[0190] In the 8th example, according to the hemodynamic sensor system of Example 7, each cluster includes one or more probability regions, and each probability region indicates the probability of a corresponding endotype associated with the cluster.

[0191] In example 9, according to the hemodynamic sensor system described in example 7, the clusters are presented as different colors.

[0192] In the 10th example, according to the hemodynamic sensor system described in Example 7, the internal shape trend is presented as multiple graphical elements.

[0193] In example 11, according to the hemodynamic sensor system described in example 10, the graphical elements are adjusted based on the temporal relevance of the latent space cardiac health parameters.

[0194] In the 12th example, according to the hemodynamic sensor system described in Example 11, the internal color of each graphic element is adjusted in terms of transparency based on its time proximity.

[0195] In the 13th example, according to the hemodynamic sensor system of Example 7, the intrinsic trend of the particular patient is a trend line.

[0196] In Example 14, the hemodynamic sensor system according to any one of Examples 1-13, wherein: the hemodynamic sensor signal includes an analog signal and the system includes an analog-to-digital converter for converting the hemodynamic sensor signal into the arterial pressure signal waveform; and determining the hypotensive pattern of the patient includes using the analog-to-digital converter to convert the hemodynamic sensor signal into the arterial pressure signal waveform.

[0197] In Example 15, a hemodynamic sensor system configured to determine and display a patient's hypotensive endotype includes: a hemodynamic sensor that generates a hemodynamic sensor signal representing an arterial pressure signal waveform of the patient; an infusion pump; a non-transitory memory storing executable instructions; and an electronic hardware processor communicating with the non-transitory memory and configured to execute the instructions such that the system at least: receives the hemodynamic sensor signal from the patient from the hemodynamic sensor; determines the patient's hypotensive endotype based on the arterial pressure signal waveform; and generates a control signal for the infusion pump to deliver intravenous therapy to the patient based on the determined hypotensive endotype and the patient's determined treatment regimen.

[0198] In the 16th example, according to the hemodynamic sensor system of Example 15, determining the hypotension pattern of the patient includes: extracting a plurality of cardiac health parameters from the arterial pressure signal waveform; encoding the plurality of cardiac health parameters into one or more latent space cardiac health parameters using a deep learning model; and generating the position of the arterial pressure signal waveform in the latent space using the one or more latent space cardiac health parameters.

[0199] In the 17th example, according to the hemodynamic sensor system of Example 16, determining the patient's hypotension endotype further includes: determining the relative position of the arterial pressure signal waveform in the latent space based on the position of the arterial pressure signal waveform in the latent space based on a set of clusters of reference arterial pressure signal waveforms from within the latent space; determining a cluster associated with the arterial pressure signal waveform in the latent space based on the set of clusters; and determining the patient's hypotension endotype based on the clusters of the arterial pressure signal waveform.

[0200] In example 18, according to the hemodynamic sensor system of example 17, the execution of the instructions further causes the system to generate an alarm based on the patient's determined hypotensive pattern.

[0201] In example 19, the hemodynamic sensor system according to example 18 further includes a graphical user interface feature that indicates the alarm and at least one of the plurality of cardiac health parameters characterizing the patient's hypotensive endotype.

[0202] In example 20, according to the hemodynamic sensor system described in example 18 or 19, the use of the infusion pump to deliver the intravenous treatment to the patient requires the approval of the user of the system.

[0203] In Example 21, the hemodynamic sensor system according to any one of Examples 16-20, wherein at least one of the plurality of cardiac health parameters includes stroke volume index (SVI), heart rate (HR), cardiac index (CI), systemic vascular resistance index (SVRI), and stroke volume variation (SVV).

[0204] In Example 22, the hemodynamic sensor system according to any one of Examples 15-21, wherein the hypotensive intratype is selected from the group consisting of: vasodilatory intratype, myocardial depression intratype, bradycardia intratype, and hypovolemic intratype.

[0205] In example 23, according to the hemodynamic sensor system of example 19, a plurality of cardiac health parameters are extracted from the arterial pressure signal waveform, wherein the cardiac health parameters are encoded into a latent space to form latent space cardiac health parameters, wherein generating data for display includes causing the presentation of the user interface, and wherein the user interface: presents a plurality of clusters, wherein each cluster is associated with a respective endotype, and wherein each cluster is formed by latent space cardiac health parameters associated with a respective patient; and updates based on information indicating a particular patient and presents an endotype trend associated with the particular patient.

[0206] In example 24, according to the hemodynamic sensor system described in example 23, each cluster includes one or more probability regions, and each probability region indicates the probability of a corresponding intrinsic type associated with the cluster.

[0207] In example 25, according to the hemodynamic sensor system described in example 23, the internal shape trend is presented as multiple graphical elements.

[0208] In example 26, according to the hemodynamic sensor system described in example 23, the intrinsic trend of the particular patient is a trend line.

[0209] In Example 27, the hemodynamic sensor system according to any one of Examples 15-26, wherein: the hemodynamic sensor signal includes an analog signal and the system includes an analog-to-digital converter for converting the hemodynamic sensor signal into the arterial pressure signal waveform; and determining the hypotensive pattern of the patient includes using the analog-to-digital converter to convert the hemodynamic sensor signal into the arterial pressure signal waveform.

[0210] In Example 28, according to the hemodynamic sensor system of Example 15, determining the patient's hypotension endotype includes: extracting a plurality of cardiac health parameters from the arterial pressure signal waveform; comparing the plurality of cardiac health parameters with reference values ​​that describe the profile of each of the plurality of cardiac health parameters in an endotype cluster; and determining the patient's hypotension endotype based on the best-fit model of the plurality of cardiac health parameters and the clusters.

[0211] In example 29, according to the system described in example 28, the clusters associated with the arterial pressure signal are formed in the latent space and then used to generate physical parameters.

[0212] In Example 30, a hemodynamic sensor system configured to determine a patient's hypotension pattern includes: a hemodynamic sensor that generates an analog hemodynamic sensor signal representing an arterial pressure signal waveform of the patient; an analog-to-digital converter that converts the analog hemodynamic sensor signal into the arterial pressure signal waveform; a non-transitory memory storing executable instructions and a fully connected deep learning model; and an electronic hardware processor communicating with the non-transitory memory and configured to execute the instructions such that the system at least: receives the analog hemodynamic sensor signal from the patient using the hemodynamic sensor; converts the analog hemodynamic sensor signal into the arterial pressure signal waveform using the analog-to-digital converter; extracts a plurality of cardiac health parameters from the arterial pressure signal waveform; encodes the plurality of cardiac health parameters into one or more latent space cardiac health parameters using the fully connected deep learning model; generates the position of the arterial pressure signal waveform in a latent space using the one or more latent space cardiac health parameters; and determines the relative position of the arterial pressure signal waveform in the latent space based on the position of the arterial pressure signal waveform in the latent space, wherein the arterial pressure signal is determined... The relative position of the waveform in the latent space includes: obtaining multiple reference arterial pressure signal waveforms from multiple patients; extracting multiple sets of cardiac health parameter reference values ​​from the multiple reference arterial pressure signal waveforms; combining each set of the multiple cardiac health parameter reference values ​​into multiple corresponding one or more reference latent space cardiac health parameters; generating a reference position in the latent space for each of the multiple reference arterial pressure signal waveforms based on the multiple one or more reference latent space cardiac health parameters of each set of the multiple cardiac health reference values; determining a clustering evaluation metric for the reference position in the latent space, the clustering evaluation metric being configured to indicate the clustering goodness of the reference position; determining an optimal number of clusters associated with the reference position based on the clustering evaluation metric; associating each of the reference arterial pressure signal waveforms with a corresponding cluster in the latent space based on the optimal number of clusters to determine a set of clusters; and determining a cluster associated with the arterial pressure signal waveform in the latent space based on the set of clusters; determining the hypotension endotype of the patient based on the clusters of the arterial pressure signal waveforms; and generating data for displaying an alarm indicating the hypotension endotype of the patient based on the determined hypotension endotype of the patient.

[0213] In Example 31, according to the hemodynamic sensor system of Example 30, the electronic hardware processor is further configured to execute the instructions to cause the system to at least: decode the one or more latent space cardiac health parameters into the plurality of cardiac health parameters in the raw space using the fully connected deep learning model.

[0214] In example 32, according to the hemodynamic sensor system of example 31, the electronic hardware processor is further configured to execute the instructions to cause the system to at least: determine a reference position in the original space for each of the plurality of reference arterial pressure signal waveforms based on a plurality of decoded cardiac health parameters in the original space; and verify a determined optimal cluster number associated with the reference position by comparing the determined reference position in the latent space with the determined reference position in the original space.

[0215] In Example 33, the hemodynamic sensor system according to any one of Examples 30-32, wherein the plurality of cardiac health parameters includes one or more of the following: a first cardiac health parameter including stroke volume index (SVI); a second cardiac health parameter including heart rate (HR); a third cardiac health parameter including cardiac index (CI); a fourth cardiac health parameter including systemic vascular resistance index (SVRI); and a fifth cardiac health parameter including stroke volume variability (SVV).

[0216] In example 34, according to the hemodynamic sensor system of example 33, the electronic hardware processor is further configured to execute the instructions to cause the system to at least: generate data for displaying at least one of the first, second, third, fourth, or fifth cardiac health parameters based on the patient's determined hypotensive endotype.

[0217] In Example 35, the hemodynamic sensor system according to any one of Examples 30-34, wherein the clustering evaluation metric includes at least one of the following: a contour metric, the Calinski-Harabasz index, the Davies-Bouldin index, or a Gaussian mixture model.

[0218] In Example 36, the hemodynamic sensor system according to any one of Examples 30-35, wherein the fully connected deep learning model includes an autoencoder configured to automatically convert the plurality of cardiac health parameters into the one or more latent space cardiac health parameters.

[0219] In example 37, the hemodynamic sensor system according to any one of Examples 30-36, wherein the relative position of the arterial pressure signal waveform in the latent space includes the clustering associated with the arterial pressure signal waveform in the latent space.

[0220] In Example 38, the hemodynamic sensor system according to any one of Examples 30-37, wherein the hypotensive intratype includes at least one of the following: vasodilatory intratype, myocardial depression intratype, bradycardia intratype, or hypovolemic intratype.

[0221] In example 39, the hemodynamic sensor system according to any one of examples 30-38, wherein the optimal number of clusters associated with the reference location is exactly four clusters.

[0222] In Example 40, a hemodynamic sensor system configured to determine the hypotensive pattern of a patient includes: a hemodynamic sensor that generates an analog hemodynamic sensor signal representing an arterial pressure signal waveform of the patient; an analog-to-digital converter that converts the analog hemodynamic sensor signal into the arterial pressure signal waveform; a non-transitory memory storing executable instructions and a fully connected deep learning model; and an electronic hardware processor communicating with the non-transitory memory and configured to execute the instructions such that the system at least: receives the analog hemodynamic sensor signal from the patient using the hemodynamic sensor; converts the analog hemodynamic sensor signal into the arterial pressure signal waveform using the analog-to-digital converter; and extracts multiple cardiac health parameters from the arterial pressure signal waveform. The process involves: encoding the plurality of cardiac health parameters into one or more latent space cardiac health parameters using the fully connected deep learning model; generating the position of the arterial pressure signal waveform in the latent space using the one or more latent space cardiac health parameters; determining the relative position of the arterial pressure signal waveform in the latent space based on the position of the arterial pressure signal waveform in the latent space, wherein determining the relative position of the arterial pressure signal waveform in the latent space includes: determining a set of clusters based on a plurality of reference arterial pressure signal waveforms; determining clusters associated with the arterial pressure signal waveform in the latent space based on the set of clusters; determining the patient's hypotension endotype based on the clusters of the arterial pressure signal waveform; and generating data for displaying an alert indicating the patient's hypotension endotype based on the determined hypotension endotype of the patient.

[0223] In Example 41, according to the hemodynamic sensor system of Example 40, the electronic hardware processor is further configured to execute the instructions to cause the system to at least: decode the one or more latent space cardiac health parameters into the plurality of cardiac health parameters in the raw space using the fully connected deep learning model.

[0224] In Example 42, according to the hemodynamic sensor system of Example 41, the electronic hardware processor is further configured to execute the instructions to cause the system to at least: determine a reference position in the original space for each of the plurality of reference arterial pressure signal waveforms based on a plurality of decoded cardiac health parameters in the original space; and verify a determined cluster associated with the arterial pressure signal waveforms in the latent space by comparing the determined reference position in the latent space with the determined reference position in the original space.

[0225] In Example 43, the hemodynamic sensor system according to any one of Examples 40 to 42, wherein the plurality of cardiac health parameters includes one or more of the following: a first cardiac health parameter including stroke volume index (SVI); a second cardiac health parameter including heart rate (HR); a third cardiac health parameter including cardiac index (CI); a fourth cardiac health parameter including systemic vascular resistance index (SVRI); and a fifth cardiac health parameter including stroke volume variability (SVV).

[0226] In example 44, according to the hemodynamic sensor system of example 43, the electronic hardware processor is further configured to execute the instructions to cause the system to at least: generate data for displaying at least one of the first, second, third, fourth, or fifth cardiac health parameters based on the patient's determined hypotensive endotype.

[0227] In Example 45, the hemodynamic sensor system according to any one of Examples 40 to 44, wherein determining the set of clusters based on the reference arterial pressure signal waveform comprises: determining a clustering evaluation metric of the reference arterial pressure signal waveform, the clustering evaluation metric being configured to indicate the clustering goodness of the reference location; and determining the optimal number of clusters associated with the reference location based on the clustering evaluation metric.

[0228] In example 46, according to the hemodynamic sensor system described in example 45, the clustering evaluation metric includes at least one of the following: contour metric, Calinski-Harabasz index, Davies-Bouldin index, or Gaussian mixture model.

[0229] In example 47, according to the hemodynamic sensor system described in example 45, the optimal number of clusters associated with the reference location is exactly four clusters.

[0230] In Example 48, the hemodynamic sensor system according to any one of Examples 40 to 47, wherein the fully connected deep learning model includes an autoencoder configured to automatically convert the plurality of cardiac health parameters into the one or more latent space cardiac health parameters.

[0231] In example 49, the hemodynamic sensor system according to any one of examples 40 to 48, wherein the relative position of the arterial pressure signal waveform in the latent space includes the clustering associated with the arterial pressure signal waveform in the latent space.

[0232] In Example 50, the hemodynamic sensor system according to any one of Examples 40 to 49, wherein the hypotensive intratype includes at least one of the following: vasodilatory intratype, myocardial depression intratype, bradycardia intratype, or hypovolemic intratype.

[0233] In Example 51, a hemodynamic sensor system configured to generate a training model for determining a patient's hypotension endotype is provided. The system includes: a hemodynamic sensor that generates an analog hemodynamic sensor signal representing a patient's arterial pressure signal waveform; an analog-to-digital converter that converts the analog hemodynamic sensor signal into the arterial pressure signal waveform; a non-transitory memory storing executable instructions and a fully connected deep learning model; and an electronic hardware processor communicating with the non-transitory memory and configured to execute the instructions such that the system at least: receives multiple analog hemodynamic sensor signals from multiple reference patients from the hemodynamic sensor; converts the multiple reference analog hemodynamic sensor signals into corresponding multiple reference arterial pressure signal waveforms using the analog-to-digital converter; extracts corresponding reference cardiac health parameter sets from the multiple reference arterial pressure signal waveforms; encodes the multiple cardiac health parameter reference value sets into multiple corresponding one or more reference latent space cardiac health parameters using the fully connected deep learning model; and, based on each of the multiple cardiac health reference value sets… The method involves: using one or more reference latent space cardiac health parameters from a set of multiple reference arterial pressure signal waveforms to generate a reference location in the latent space for each of the multiple reference arterial pressure signal waveforms; determining a clustering evaluation metric for the reference location in the latent space, the clustering evaluation metric being configured to indicate the clustering goodness of the reference location; determining an optimal number of clusters associated with the reference location based on the clustering evaluation metric; associating each of the reference arterial pressure signal waveforms with a corresponding cluster in the latent space based on the optimal number of clusters to determine a set of clusters; determining a cluster associated with each of the reference arterial pressure signal waveforms in the latent space based on the set of clusters and the reference location of each of the multiple reference arterial pressure signal waveforms in the latent space; determining the hypotension endotype of each of the reference patients based on the clusters associated with each of the reference arterial pressure signal waveforms; decoding the one or more reference latent space cardiac health parameters into a corresponding set of multiple cardiac health parameter reference values ​​using the fully connected deep learning model; and generating a training model for determining the hypotension endotype of the patient.

[0234] In example 52, according to the hemodynamic sensor system described in example 51, each of the multiple groups of reference cardiac health parameters includes one or more of the following: a first cardiac health parameter including stroke volume index (SVI); a second cardiac health parameter including heart rate (HR); a third cardiac health parameter including cardiac index (CI); a fourth cardiac health parameter including systemic vascular resistance index (SVRI); and a fifth cardiac health parameter including stroke volume variability (SVV).

[0235] In example 53, according to the hemodynamic sensor system of example 52, wherein the electronic hardware processor is further configured to execute the instructions to cause the system to at least: generate data for displaying at least one of the first, second, third, fourth, or fifth cardiac health parameters based on the patient's determined hypotensive endotype.

[0236] In Example 54, the hemodynamic sensor system according to any one of Examples 36 to 38, wherein the clustering evaluation metric includes at least one of the following: a contour metric, the Calinski-Harabasz index, the Davies-Bouldin index, or a Gaussian mixture model.

[0237] In example 55, the hemodynamic sensor system according to any one of examples 51 to 54, wherein the optimal number of clusters associated with the reference location is exactly four clusters.

[0238] In Example 56, the hemodynamic sensor system according to any one of Examples 51 to 55, wherein the fully connected deep learning model includes an autoencoder configured to automatically convert the plurality of cardiac health parameters into the one or more latent space cardiac health parameters.

[0239] In Example 57, the hemodynamic sensor system according to any one of Examples 51 to 56, wherein the hypotensive intratype includes at least one of the following: vasodilatory intratype, myocardial depression intratype, bradycardia intratype, or hypovolemic intratype.

[0240] In example 58, a hemodynamic sensor system configured to determine the hypotensive pattern of a patient includes: a hemodynamic sensor that generates an analog hemodynamic sensor signal representing an arterial pressure signal waveform of the patient; an analog-to-digital converter that converts the analog hemodynamic sensor signal into the arterial pressure signal waveform; a non-transitory memory storing executable instructions and a fully connected deep learning model; and an electronic hardware processor communicating with the non-transitory memory and configured to execute the instructions such that the system at least: receives the analog hemodynamic sensor signal from the patient from the hemodynamic sensor; and converts the analog hemodynamic sensor signal into the arterial pressure signal waveform using the analog-to-digital converter. The simulated hemodynamic sensor signal is converted into the arterial pressure signal waveform; multiple cardiac health parameters, including the following, are extracted from the arterial pressure signal waveform: a first cardiac health parameter including stroke volume index (SVI); a second cardiac health parameter including heart rate (HR); a third cardiac health parameter including cardiac index (CI); a fourth cardiac health parameter including systemic vascular resistance index (SVRI); and a fifth cardiac health parameter including stroke volume variability (SVV); the multiple cardiac health parameters are encoded into first and second latent space cardiac health parameters using the fully connected deep learning model; and the arterial pressure signal waveform is generated using the first and second latent space cardiac health parameters. The position of the arterial pressure signal waveform in the latent space; determining the relative position of the arterial pressure signal waveform in the latent space based on the position of the arterial pressure signal waveform in the latent space, wherein determining the relative position of the arterial pressure signal waveform in the latent space includes: obtaining multiple reference arterial pressure signal waveforms from multiple patients; extracting multiple sets of cardiac health parameter reference values ​​from the multiple reference arterial pressure signal waveforms, each set of the cardiac health parameter reference values ​​including corresponding SVI, HR, CI, SVRI, and SVV; combining each set of the multiple cardiac health parameter reference values ​​into multiple corresponding first and second reference latent space cardiac health parameters; based on each set of the multiple cardiac health reference values... The plurality of first and second reference latent space cardiac health parameters are used to generate a reference location in the latent space for each of the plurality of reference arterial pressure signal waveforms; a contour criterion for the reference location in the latent space is determined, the contour criterion being configured to indicate the clustering goodness of the reference location; an optimal number of clusters associated with the reference location is determined based on the contour criterion; each of the reference arterial pressure signal waveforms is associated with a corresponding cluster in the latent space based on the optimal number of clusters to determine a set of clusters; and a cluster associated with the arterial pressure signal waveform in the latent space is determined based on the set of clusters; and the hypotensive endotype of the patient is determined based on the clusters of the arterial pressure signal waveforms.And data is generated based on the patient's determined hypotension pattern to display an alarm indicating the patient's hypotension pattern and at least one of the first, second, third, fourth, or fifth cardiac health parameters.

[0241] In example 59, according to the hemodynamic sensor system of example 58, the electronic hardware processor is further configured to execute the instructions to cause the system to at least: decode the first and second latent space cardiac health parameters into the plurality of cardiac health parameters using the fully connected deep learning model.

[0242] In example 60, a hemodynamic sensor system configured to determine the hypotensive pattern of a patient includes: a hemodynamic sensor that generates an analog hemodynamic sensor signal representing an arterial pressure signal waveform of the patient; an analog-to-digital converter that converts the analog hemodynamic sensor signal into the arterial pressure signal waveform; a non-transitory memory storing executable instructions and a fully connected deep learning model; and an electronic hardware processor communicating with the non-transitory memory and configured to execute the instructions such that the system at least: receives the analog hemodynamic sensor signal from the patient using the hemodynamic sensor; converts the analog hemodynamic sensor signal into the arterial pressure signal waveform using the analog-to-digital converter; and extracts from the arterial pressure signal waveform a plurality of cardiac health parameters including: a first cardiac health parameter including stroke volume index (SVI); a second cardiac health parameter including heart rate (HR); a third cardiac health parameter including cardiac index (CI); and a third cardiac health parameter including systemic vascular... The system comprises: a fourth cardiac health parameter including the resistance index (SVRI); and a fifth cardiac health parameter including stroke volume variation (SVV); encoding the plurality of cardiac health parameters into first and second latent space cardiac health parameters using the fully connected deep learning model; generating the position of the arterial pressure signal waveform in the latent space using the first and second latent space cardiac health parameters; determining the relative position of the arterial pressure signal waveform in the latent space based on the position of the arterial pressure signal waveform in the latent space, wherein determining the relative position of the arterial pressure signal waveform in the latent space includes: determining a set of clusters based on a reference arterial pressure signal waveform; and determining a cluster associated with the arterial pressure signal waveform in the latent space based on the set of clusters; determining the patient's hypotension endotype based on the clusters of the arterial pressure signal waveform; and generating data based on the determined hypotension endotype of the patient for displaying an alarm indicating the patient's hypotension endotype and at least one of the first, second, third, fourth, or fifth cardiac health parameters.

[0243] In example 61, a hemodynamic sensor system configured to generate a training model for determining a patient's hypotension pattern includes: a hemodynamic sensor that generates an analog hemodynamic sensor signal representing a patient's arterial pressure signal waveform; an analog-to-digital converter that converts the analog hemodynamic sensor signal into the arterial pressure signal waveform; a non-transitory memory storing executable instructions and a fully connected deep learning model; and an electronic hardware processor communicating with the non-transitory memory and configured to execute the instructions such that the system at least: receives multiple data from the hemodynamic sensor from multiple reference patients. The system uses multiple analog hemodynamic sensor signals; it uses the analog-to-digital converter to convert the multiple reference analog hemodynamic sensor signals into multiple corresponding reference arterial pressure signal waveforms; it extracts corresponding reference cardiac health parameter sets from the multiple reference arterial pressure signal waveforms, each of the reference cardiac health parameter sets including: a first cardiac health parameter including stroke volume index (SVI); a second cardiac health parameter including heart rate (HR); a third cardiac health parameter including cardiac index (CI); a fourth cardiac health parameter including systemic vascular resistance index (SVRI); and a fifth cardiac health parameter including stroke volume variability (SVV). The fully connected deep learning model is used to encode the plurality of cardiac health parameter reference value groups into a plurality of corresponding first and second reference latent space cardiac health parameters; a reference position in the latent space is generated for each of the plurality of reference arterial pressure signal waveforms based on the plurality of first and second reference latent space cardiac health parameters of each of the plurality of cardiac health reference value groups; a contour criterion for the reference position in the latent space is determined, the contour criterion being configured to indicate the clustering goodness of the reference position; an optimal number of clusters is determined based on the contour criterion; and each of the reference arterial pressure signal waveforms is clustered based on the optimal number of clusters. Associating with corresponding clusters in the latent space to determine a set of clusters; determining a cluster associated with each of the plurality of reference arterial pressure signal waveforms in the latent space based on the set of clusters and the reference position of each of the plurality of reference arterial pressure signal waveforms in the latent space; determining the hypotension endotype of each of the reference patients based on the clusters associated with each of the reference arterial pressure signal waveforms; decoding the plurality of first and second reference latent space cardiac health parameters into corresponding plurality of cardiac health parameter reference value sets using the fully connected deep learning model; and generating a training model for determining the hypotension endotype of the patients.

[0244] In example 62, a method implemented by a system of one or more processors, wherein the system generates a user interface for presentation, and wherein the user interface: presents multiple data point clusters, wherein each data point cluster is associated with a respective intrinsic type, and wherein each data point cluster is formed by data points associated with multiple patients, and is updated based on information indicating a particular patient and presents intrinsic type trends reflecting the multiple data points associated with the particular patient.

[0245] In example 63, according to the method described in example 62, the specific patient is not included among the plurality of patients.

[0246] In example 64, the method described in example 62 is used, wherein the data points are clustered based on latent space cardiac health parameters determined for each patient at different time points.

[0247] In example 65, according to the method of example 62, the specific patient is not included in the plurality of patients, wherein latent space cardiac health parameters are determined for the specific patient at different times, and wherein the latent space cardiac health parameters are mapped to the data point clusters.

[0248] In example 66, the data point clusters are presented as different colors, according to the method described in example 62.

[0249] In example 67, according to the method described in example 62, the inner-type trend is presented as a plurality of graphical elements associated with the plurality of data points related to the particular patient.

[0250] In example 68, the graphic element is a star, according to the method described in example 67.

[0251] In example 69, the method described in example 67 is used, wherein the graphical elements are adjusted based on the proximity of the data points.

[0252] In example 70, the method described in example 69 is followed, wherein the internal color of each graphic element is adjusted in terms of transparency based on its proximity.

[0253] In example 71, according to the method of example 70, the user interface presents a plurality of ellipses, and each ellipse includes a subset of data points with similar probabilities assigned a particular inner shape.

[0254] In example 72, the data point clusters are assigned the same color, according to the method described in example 70.

[0255] In example 73, according to the method described in example 70, the in-type trend is a trend line connecting the data points associated with the particular patient.

[0256] In example 74, according to the method described in example 70, the text summary of the internal trend is included in the user interface.

[0257] In Example 75, a system includes one or more processors and a computational storage medium storing instructions that, when executed by the one or more processors, cause the processors to perform the method described in Examples 62-74.

[0258] In Example 76, a non-transitory computer storage medium stores instructions that, when executed by a system of one or more processors, cause the one or more processors to perform the method described in Examples 62-74.

[0259] In example 77, a computer-implemented method includes: using input data from a database reflecting latent feature space cardiac health parameters previously calculated for multiple patients, the input data including probabilities associated with the patient's hypotensive endotype, and the input data being associated with multiple time periods; and generating a user interface for presentation via a user device, wherein the user interface: presents multiple data point clusters previously calculated from the database, wherein each data point cluster is associated with a respective endotype, and wherein each data point cluster is formed by data points included in the input data; and presents multiple data points obtained in real time for new and currently monitored patients who are not part of the database containing data points for multiple patients; and presents an endotype trend reflecting the changes over time of the multiple data points associated with the new and currently monitored patients, the data points being associated with at least two or more time periods.

[0260] In example 78, according to the method of example 77, the user interface presents multiple ellipses for each data point cluster from the database, and each ellipse includes a subset of the data points that are assigned a similar probability of a specific internal type within each data point cluster from the database.

[0261] In example 79, the data point clusters are presented as different colors, according to the method described in examples 77-78.

[0262] In Example 80, according to the method described in Examples 77-79, the intrinsic trend is presented as multiple graphical elements associated with the multiple data points related to the new and currently monitored patients.

[0263] In example 81, the graphic element is a star, according to the method described in example 80.

[0264] In example 82, the method described in example 81 is followed, wherein the graphical elements are adjusted based on the temporal relevance of the data points.

[0265] In example 83, the method according to any one of examples 77-82 is used, wherein the internal color of each graphic element is adjusted in terms of transparency based on its time relevance.

[0266] In example 84, the method according to any one of examples 77-83 is used, wherein the clusters of data points from the database are assigned the same color.

[0267] In example 85, according to the method of any one of examples 77-84, the intrinsic trend of the new and currently monitored patients is a trend line connecting the data points associated with the new and currently monitored patients.

[0268] In example 86, according to the method of any one of examples 77-85, a text summary of the intratype clusters of the database and the intratype trends of the new and currently monitored patients are included in the user interface.

[0269] In Example 87, a system includes one or more processors and a computational storage medium storing instructions that, when executed by the one or more processors, cause the processors to perform a method according to any one of Examples 77-86.

[0270] In example 88, a non-transitory computer storage medium stores instructions that, when executed by a system of one or more processors, cause the one or more processors to perform the method according to any one of examples 77-87.

[0271] In Example 89, according to the hemodynamic sensor system of Example 3, generating the position of the arterial pressure signal waveform in the latent space using the one or more latent space cardiac health parameters includes: determining multiple positions of multiple arterial pressure signal waveforms in the latent space using corresponding latent space cardiac health parameters at multiple time points.

[0272] In example 90, according to the hemodynamic sensor system of example 89, determining the plurality of positions of the plurality of arterial pressure signal waveforms in the latent space using the corresponding latent space cardiac health parameters includes: determining the harmonic probabilities associated with the plurality of positions of the plurality of arterial pressure signal waveforms in the latent space.

[0273] In example 91, according to the hemodynamic sensor system of example 89, determining the plurality of locations of the plurality of arterial pressure signal waveforms in the latent space using the corresponding latent space cardiac health parameters includes: determining a Gaussian distribution associated with the plurality of locations of the plurality of arterial pressure signal waveforms in the latent space.

[0274] In example 92, according to the hemodynamic sensor system described in example 89, the alarm includes multiple indicators corresponding to each of the plurality of time points.

[0275] In example 93, according to the hemodynamic sensor system of example 92, at least one of the plurality of indicators indicates a time point relative to at least one of the other plurality of indicators.

[0276] In example 94, according to the hemodynamic sensor system of example 1, determining the hypotension endotype of the patient includes: determining a hypotension probability index (HPI) corresponding to the probability that the patient will have the hypotension endotype.

[0277] In example 95, according to the hemodynamic sensor system described in example 94, the data used to generate an alarm indicating the patient's hypotensive pattern is further based on the HPI being higher than a threshold HPI.

[0278] In example 96, according to the hemodynamic sensor system of example 1, wherein the processor is configured to execute the instructions to further enable the system to at least: determine a treatment plan for the patient based on the determined hypotensive pattern.

[0279] In example 97, according to the hemodynamic sensor system of example 96, wherein the processor is configured to execute the instructions to further cause the system to at least: generate a command based on a determined treatment regimen, the command being configured to cause an infusion pump to deliver an intravenous therapeutic agent to the patient.

[0280] In example 98, the hemodynamic sensor system according to example 97 further includes the infusion pump.

[0281] In example 99, according to the hemodynamic sensor system of example 1, determining the patient's hypotension endotype includes: extracting a plurality of cardiac health parameters from the arterial pressure signal waveform; comparing the plurality of cardiac health parameters with reference values ​​that describe the profile of each of the plurality of cardiac health parameters in an endotype cluster; and determining the patient's hypotension endotype based on the best-fit model of the plurality of cardiac health parameters and the clusters.

[0282] In example 100, according to the hemodynamic sensor system of example 99, the reference values ​​include a table.

Claims

1. A hemodynamic sensor system configured to determine and display the hypotensive pattern of a patient, the system comprising: A hemodynamic sensor that generates a hemodynamic sensor signal representing the waveform of the patient's arterial pressure signal; Infusion pump; Non-transitory memory, on which executable instructions are stored; and An electronic hardware processor, which communicates with the non-transitory memory and is configured to execute the instructions to cause the system to at least: Receive the hemodynamic sensor signal from the patient from the hemodynamic sensor; The patient's hypotension type is determined based on the arterial pressure signal waveform; as well as Based on the patient's identified hypotensive pattern and the patient's identified treatment regimen, control signals are generated for the infusion pump to deliver intravenous therapy to the patient.

2. The hemodynamic sensor system of claim 1, wherein determining the hypotensive pattern of the patient comprises: Multiple cardiac health parameters are extracted from the arterial pressure signal waveform; The multiple cardiac health parameters are encoded into one or more latent space cardiac health parameters using a deep learning model. as well as The location of the arterial pressure signal waveform in the latent space is generated using one or more latent space cardiac health parameters.

3. The hemodynamic sensor system of claim 2, wherein determining the hypotensive type of the patient further comprises: Based on the position of the arterial pressure signal waveform in the latent space, the relative position of the arterial pressure signal waveform in the latent space is determined based on a set of clusters from reference arterial pressure signal waveforms within the latent space; Based on the aforementioned set of clusters, clusters associated with the arterial pressure signal waveform in the latent space are determined; as well as The patient's hypotension pattern is determined based on the clustering of the arterial pressure signal waveform.

4. The hemodynamic sensor system of claim 3, wherein executing the instructions further causes the system to: An alarm is generated based on the patient's identified hypotensive pattern.

5. The hemodynamic sensor system of claim 4, further comprising a graphical user interface feature that indicates the alarm and at least one of the plurality of cardiac health parameters characterizing the patient's hypotensive pattern.

6. The hemodynamic sensor system of claim 4, wherein the use of the infusion pump to deliver the intravenous treatment to the patient requires the approval of the user of the system.

7. The hemodynamic sensor system according to claim 2, wherein at least one of the plurality of cardiac health parameters includes stroke volume index (SVI), heart rate (HR), cardiac index (CI), systemic vascular resistance index (SVRI), and stroke volume variation (SVV).

8. The hemodynamic sensor system according to claim 1, wherein the hypotension type is selected from the group consisting of: vasodilatory type, myocardial inhibition type, bradycardia type and hypovolemic type.

9. The hemodynamic sensor system of claim 5, wherein a plurality of cardiac health parameters are extracted from the arterial pressure signal waveform, wherein the cardiac health parameters are encoded into a latent space to form latent space cardiac health parameters, wherein generating data for display includes causing the presentation of the user interface, and wherein the user interface: Multiple clusters are presented, each associated with a specific endotype, and each cluster is formed by latent space cardiac health parameters associated with its respective patient; and Updates are made based on information indicating a specific patient and endotype trends associated with that specific patient are presented.

10. The hemodynamic sensor system of claim 9, wherein each cluster comprises one or more probability regions, and wherein each probability region indicates the probability of a corresponding intrinsic type associated with the cluster.

11. The hemodynamic sensor system of claim 10, wherein the internal shape trend is presented as a plurality of graphic elements.

12. The hemodynamic sensor system according to any one of claims 1-11, wherein: The hemodynamic sensor signal includes an analog signal and the system includes an analog-to-digital converter that converts the hemodynamic sensor signal into the arterial pressure signal waveform; as well as Determining the patient's hypotension pattern includes using the analog-to-digital converter to convert the hemodynamic sensor signal into the arterial pressure signal waveform.

13. The hemodynamic sensor system of claim 1, wherein determining the hypotensive pattern of the patient comprises: Multiple cardiac health parameters are extracted from the arterial pressure signal waveform; The plurality of cardiac health parameters are compared with reference values, which describe the profile of each of the plurality of cardiac health parameters in an intratype cluster; as well as The patient's hypotensive endotype is determined based on the best-fit model of the multiple cardiac health parameters and the clustering.

14. The system of claim 13, wherein the clusters associated with the arterial pressure signal are formed in a latent space and then used to generate physical parameters.

15. A hemodynamic sensor system configured to determine and display the hypotensive pattern of a patient, the system comprising: A hemodynamic sensor that generates a hemodynamic sensor signal representing the waveform of the patient's arterial pressure signal; user interface; Non-transitory memory, on which executable instructions are stored; and An electronic hardware processor, which communicates with the non-transitory memory and is configured to execute the instructions to cause the system to at least: Receive the hemodynamic sensor signal from the patient from the hemodynamic sensor; The patient's hypotension type is determined based on the arterial pressure signal waveform; as well as Data is generated based on the patient's determined hypotension pattern to display an alert indicating the patient's hypotension pattern on the user interface.

16. The hemodynamic sensor system of claim 15, wherein determining the hypotensive pattern of the patient comprises: Multiple cardiac health parameters are extracted from the arterial pressure signal waveform; The multiple cardiac health parameters are encoded into one or more latent space cardiac health parameters using a deep learning model. as well as The location of the arterial pressure signal waveform in the latent space is generated using one or more latent space cardiac health parameters.

17. The hemodynamic sensor system of claim 16, wherein determining the hypotensive type of the patient further comprises: Based on the position of the arterial pressure signal waveform in the latent space, the relative position of the arterial pressure signal waveform in the latent space is determined based on a set of clusters from reference arterial pressure signal waveforms within the latent space; Based on the aforementioned set of clusters, clusters associated with the arterial pressure signal waveform in the latent space are determined; as well as The patient's hypotension pattern is determined based on the clustering of the arterial pressure signal waveform.

18. The hemodynamic sensor system of claim 16, wherein the alarm includes a graphical user interface feature indicating at least one of the plurality of cardiac health parameters characterizing the patient's hypotensive endotype.

19. The hemodynamic sensor system of claim 18, wherein at least one of the plurality of cardiac health parameters includes stroke volume index (SVI), heart rate (HR), cardiac index (CI), systemic vascular resistance index (SVRI), and stroke volume variation (SVV).

20. The hemodynamic sensor system according to claim 15, wherein the hypotension type is selected from the group consisting of: vasodilatory type, myocardial inhibition type, bradycardia type and hypovolemic type.

21. The hemodynamic sensor system of claim 15, wherein a plurality of cardiac health parameters are extracted from the arterial pressure signal waveform, wherein the cardiac health parameters are encoded into a latent space to form latent space cardiac health parameters, wherein generating data for display includes causing the presentation of the user interface, and wherein the user interface: Multiple clusters are presented, each associated with a specific endotype, and each cluster is formed by latent space cardiac health parameters associated with its respective patient. Updates are made based on information indicating a specific patient and endotype trends associated with that specific patient are presented.

22. The hemodynamic sensor system of claim 21, wherein each cluster comprises one or more probability regions, and wherein each probability region indicates the probability of a corresponding intrinsic type associated with the cluster.

23. The hemodynamic sensor system of claim 15, wherein determining the hypotensive pattern of the patient comprises: Determine the hypotension probability index (HPI) corresponding to the probability of the patient developing the hypotension type.

24. The hemodynamic sensor system of claim 23, wherein the data for generating an alarm indicating the patient’s hypotensive pattern is further based on the HPI being higher than a threshold HPI.

25. The hemodynamic sensor system of claim 15, wherein the processor is configured to execute the instructions to further cause the system to at least: Treatment plans are determined for the patient based on the identified hypotension pattern.

26. The hemodynamic sensor system of claim 25, wherein the processor is configured to execute the instructions to further cause the system to at least: Commands are generated based on a defined treatment plan and are configured to cause an infusion pump to deliver an intravenous therapeutic agent to the patient.

27. The hemodynamic sensor system of claim 26, further comprising the infusion pump.

28. The hemodynamic sensor system of claim 15, wherein determining the hypotensive pattern of the patient comprises: Multiple cardiac health parameters are extracted from the arterial pressure signal waveform; The plurality of cardiac health parameters are compared with reference values, which describe the profile of each of the plurality of cardiac health parameters in an intratype cluster; as well as The patient's hypotensive endotype is determined based on the best-fit model of the multiple cardiac health parameters and the clustering.

29. The hemodynamic sensor system according to any one of claims 15-28, wherein: The hemodynamic sensor signal includes an analog signal and the system includes an analog-to-digital converter that converts the hemodynamic sensor signal into the arterial pressure signal waveform; as well as Determining the patient's hypotension pattern includes using the analog-to-digital converter to convert the hemodynamic sensor signal into the arterial pressure signal waveform.

30. A hemodynamic sensor system configured to determine the hypotensive pattern of a patient, the system comprising: A hemodynamic sensor that generates an analog hemodynamic sensor signal representing the patient's arterial pressure signal waveform; An analog-to-digital converter that converts the analog hemodynamic sensor signal into the arterial pressure signal waveform; Non-transitory memory, which stores executable instructions and fully connected deep learning models; as well as An electronic hardware processor, which communicates with the non-transitory memory and is configured to execute the instructions to cause the system to at least: Receive the simulated hemodynamic sensor signal from the patient from the hemodynamic sensor; The analog-to-digital converter is used to convert the analog hemodynamic sensor signal into the arterial pressure signal waveform. Multiple cardiac health parameters are extracted from the arterial pressure signal waveform; The fully connected deep learning model is used to encode the multiple cardiac health parameters into one or more latent space cardiac health parameters. The position of the arterial pressure signal waveform in the latent space is generated using one or more latent space cardiac health parameters; Determining the relative position of the arterial pressure signal waveform in the latent space based on its position in the latent space includes: Multiple reference arterial pressure signal waveforms were obtained from multiple patients; Multiple sets of reference values ​​for cardiac health parameters are extracted from the multiple reference arterial pressure signal waveforms; Each of the multiple sets of reference values ​​for cardiac health parameters is combined into one or more corresponding reference latent space cardiac health parameters. Based on one or more reference latent space cardiac health parameters of each of the plurality of cardiac health reference value groups, a reference position in the latent space is generated for each of the plurality of reference arterial pressure signal waveforms; Determine a clustering evaluation metric for the reference location in the latent space, the clustering evaluation metric being configured to indicate the clustering goodness of the reference location; The optimal number of clusters associated with the reference location is determined based on the clustering evaluation metric. Based on the optimal number of clusters, each of the reference arterial pressure signal waveforms is associated with a corresponding cluster in the latent space to determine a set of clusters; and Based on the aforementioned set of clusters, clusters associated with the arterial pressure signal waveform in the latent space are determined; The patient's hypotension pattern is determined based on the clustering of the arterial pressure signal waveform; and Data is generated based on the patient's identified hypotension pattern to display an alarm indicating the patient's hypotension pattern.

Images