WO2025199425A1

WO2025199425A1 - Computational-based predictions of post-induction hypotension

Info

Publication number: WO2025199425A1
Application number: PCT/US2025/020894
Authority: WO
Inventors: Harikesh SUBRAMANIAN; Shyam VISWESWARAN
Original assignee: University of Pittsburgh
Current assignee: University of Pittsburgh
Priority date: 2024-03-22
Filing date: 2025-03-21
Publication date: 2025-09-25
Anticipated expiration: 2026-09-22

Abstract

A method for predicting a mean blood pressure (MBP) post induction and associated with post-induction hypotension (PIH) in a patient is presented. The method includes accessing a data set including demographics and clinical data. The demographics and clinical data include a set of preoperative input variables and anesthetic induction input variables. The method includes refining a set of hyperparameters associated with a machine¬ learning model trained to generate a prediction of MBP post induction. The process includes preprocessing the data set by executing encoding processes on each of the set of preoperative input variables and anesthetic induction input variables, calculating cross- validation losses based on the preprocessed set of preoperative input variables and anesthetic induction input variables and a predetermined set of target MBP values, updating the set of hyperparameters based on the cross-validation losses, and outputting the prediction of MBP post induction based on the updated set of hyperparameters.

Description

COMPUTATIONAL-BASED PREDICTIONS OF POST-INDUCTION HYPOTENSION

TECHNICAL FIELD

This application relates generally to post-induction hypotension, and, more particularly, to computational-based predictions of post-induction hypotension.

CROSS-REFERENCE TO RELATED APPLICATIONS

This application claims the benefit of priority of U.S. Provisional Patent Application No. 63/568570, filed March 22, 2024, the content of which is incorporated herein by reference in its entirety, and to which priority is claimed.

BACKGROUND

Generally, an anesthetic episode may include three key chronological phases: induction, maintenance, and emergence. Induction occurs at the start of the anesthetic episode in which a generally rapid and smooth loss of consciousness, analgesia, and amnesia of the patient are achieved. Arterial hypotension during anesthesia may occur frequently and be generally associated with adverse patient outcomes. For example, hypotension during the early phase of anesthesia, which may be referred to as postinduction hypotension (PIH), may be related to multiple causative mechanisms. Such causative mechanisms may include an age of the patient, a pre-induction systolic blood pressure (SBP) of the patient, a weight of the patient, a sex of the patient, and so forth. Additionally, comorbidity, preoperative use of medications, and anesthesia techniques, including the type and dose of the anesthetic agent administered, may also contribute to the development of PUT in anesthesia patients.

In some instances, an appropriate combination and dosage of medications, as well as pre-treatment with certain vasopressors, may minimize PUT in anesthesia patients. However, identifying anesthesia patients at risk of PIH remains a complex and elusive task. For example, induction may generally include a complex, high-stakes procedure that includes taking into account the patient’s entire medical history, current illness, previous medication administration, and any number of other factors. It may be thus useful to provide computational -based techniques to predict PIH in anesthesia patients.

SUMMARY

Embodiments of the present disclosure are directed toward one or more computing devices, methods, and non-transitory computer-readable media for predicting a mean blood pressure (MBP) post induction and associated with post-induction hypotension (PIH) in a patient and recommending candidate therapies for intraoperative hypotension throughout each phase of anesthesia. In certain embodiments, the one or more computing devices may access a data set including demographics and clinical data associated with the patient. For example, in one embodiment, the demographics and clinical data may include a set of preoperative input variables and anesthetic induction input variables associated with the patient. In certain embodiments, the set of preoperative input variables and anesthetic induction input variables may include one or more of a set of demographics input variables, a set of preoperative comorbidity input variables, a set of preoperative laboratory testing input variables, a set of medical procedure input variables, a set of peri-induction medication input variables, or a set of pre-induction MBP measurement input variables.

In certain embodiments, the set of demographics input variables may include one or more of an age of the patient, a sex of the patient, a race or ethnicity of the patient, or a weight of the patient. In certain embodiments, the set of preoperative laboratory testing input variables may include serum concentrations of one or more of sodium, creatinine, blood urea nitrogen (BUN), hemoglobin, white blood cell (WBC) counts, albumin, bilirubin, aspartate transferase (AST), estimated glomerular filtration rate (EGFR), platelet concentration, international normalized ratio (INR), or partial thromboplastin time (PTT). In certain embodiments, the set of preoperative input variables and anesthetic induction input variables may include a set of peri-induction medication input variables. In certain embodiments, the set of peri-induction medication input variables may include a predetermined set of anesthetic agents and their respective dosages. For example, in one embodiment, the predetermined set of anesthetic agents may be selected from the group consisting of propofol, etomidate, ketamine, dexmedetomidine, nitrous oxide (NO), halothane, sevoflurane, desflurane, isoflurane, midazolam, diazepam, lorazepam, alfentanil, sufentanil, remifentanil, fentanyl, atracurium, cisatracurium, pancuronium, vecuronium, rocuronium, succinylcholine, hydromorphone, hydrocodone, oxycodone, phenylephrine, and lidocaine.

In certain embodiments, the one or more computing devices may refine a set of hyperparameters associated with a machine-learning model trained to generate a prediction of a mean blood pressure (MBP) post induction and associated with post-induction hypotension (PIH) in the patient. For example, in certain embodiments, refining the set of hyperparameters may include iteratively executing a process until a desired precision is reached. In certain embodiments, the process may include the one or more computing devices may preprocess the data set by executing one or more encoding processes on one or more input variables of the set of preoperative input variables and anesthetic induction input variables. For example, in one embodiment, the one or more computing devices may execute the one or more encoding processes on the one or more input variables by executing an imputation of one or more missing values associated with the one or more input variables. In another embodiment, the one or more computing devices may execute the one or more encoding processes on the one or more input variables by executing a one-hot encoding of the one or more input variables.

In certain embodiments, the process may further include the one or more computing devices calculating one or more cross-validation losses based at least in part on the preprocessed set of preoperative input variables and anesthetic induction input variables and a predetermined set of target MBP values, and further updating the set of hyperparameters based on the one or more cross-validation losses. In certain embodiments, the process may further include the one or more computing devices outputting, by the machine learning model, the prediction of the MBP associated with PIH in the patient based at least in part on the updated set of hyperparameters. For example, in some embodiments, the one or more computing devices may output the prediction of the MBP by outputting, by the machinelearning model, a prediction of MBP post induction over a predetermined time interval corresponding to a post-induction period associated with the patient.

In certain embodiments, the one or more computing devices may calculate the one or more cross-validation losses by evaluating a cross-validation loss function based on the preprocessed set of preoperative input variables and anesthetic induction input variables, the predetermined set of target MBP values, the set of hyperparameters, and a set of learnable parameters associated with the machine-learning model. In certain embodiments, the one or more computing devices may further minimize the cross-validation loss function by varying the set of learnable parameters while the preprocessed set of preoperative input variables and anesthetic induction input variables, the predetermined set of target MBP values, the set of hyperparameters remain constant. For example, in one embodiment, the one or more computing devices may minimize the cross-validation loss function by optimizing the set of hyperparameters. In one embodiment, the set of hyperparameters may include one or more of a set of general parameters, a set of booster parameters, or a set of learning-task parameters.

In certain embodiments, the one or more computing devices may minimize the cross-validation loss function by minimizing a loss between a prediction of one or more MBP values post induction and associated with PIH in the patient and an experimentally- determined one or more MBP values post induction and associated with PIH in the patient. In certain embodiments, the set of learnable parameters may include one or more weights or decision variables determined by the machine learning model based at least in part on the preprocessed set of preoperative input variables and anesthetic induction input variables and the predetermined set of target MBP values. In certain embodiments, the updated set of hyperparameters may include one or more of an updated set of general parameters, an updated set of booster parameters, or an updated set of learning-task parameters.

In one embodiment, the machine learning model may include one or more of a gradient boosting model, an adaptive boosting (AdaBoost) model, an extreme gradient boosting (XGBoost) model, a light gradient boosted machine (LightGBM) model, or a categorical boosting (CatBoost) model. In another embodiment, the machine learning model may include one or more of a linear regression model, a ridge regression model, or a Bayesian ridge regression model. In certain embodiments, prior to accessing the data set including the demographics and clinical data, the one or more computing devices may split the data set into a training data set, a test data set, and a validation data set for training and executing the machine-learning model. In certain embodiments, the one or more computing devices may generate a report based on the prediction of the MBP associated with PIH in the patient, and further transmit the report to a computing device associated with a clinician. In certain embodiments, the one or more computing devices may cause, based on the prediction of the MBP post induction and associated with PIH in the patient, a user interface (UI) executing on the computing device to display a visual representation of the prediction of the MBP post induction and associated with PIH in the patient. In certain embodiments, the one or more computing devices may receive a set of post-induction MBP measurement input variables. For example, the set of post-induction MBP measurement input variables may include MBP measurements at a set of time intervals in a post-induction period.

The present embodiments described herein may further provide a number of technical improvements to the functioning of computing systems. For example, in some embodiments, the training and execution of the one or more machine-learning models may be memory-efficient and compute-efficient in that categorical data in the input data set, such as preoperative input variables and anesthetic induction input variables, may be individually preprocessed utilizing one or more encoding processes (e.g., ^-nearest neighbor imputation, one-hot encoding) before the categorical data in the input data set is fed to the one or more machine-learning models. Thus, the one or more machine-learning models may be trained and executed in a memory-efficient and compute-efficient manner. In this way, overall processing device (e.g., CPU, GPU, or Al accelerator) performance in terms of execution time, latency, power consumption, and clock speed may all be markedly improved.

BRIEF DESCRIPTION OF THE DRAWINGS

One or more drawings included herein are in color in accordance with 37 CFR §1.84. The color drawings are necessary to illustrate the invention.

FIG. 1 illustrates a clinical and computing environment in accordance with some embodiments disclosed herein.

FIG. 2 illustrates a workflow diagram for predicting a mean blood pressure (MBP) post induction and associated with post-induction hypotension (PUT) in a patient in accordance with some embodiments disclosed herein.

FIG. 3 illustrates an example embodiment of an anesthesia clinical support user interface (UI) for displaying a prediction of a mean blood pressure (MBP) post induction and associated with post-induction hypotension (PUT) in a patient in accordance with some embodiments disclosed herein.

FIG. 4 illustrates a flow diagram of a method for predicting a mean blood pressure (MBP) post induction and associated with post-induction hypotension (PUT) in a patient in accordance with some embodiments disclosed herein. FIG. 5 illustrates an example computing system in accordance with some embodiments disclosed herein.

FIG. 6 illustrates a diagram of an example artificial intelligence (Al) architecture included as part of the example computing system of FIG. 5 in accordance with some embodiments disclosed herein.

DESCRIPTION OF EXAMPLE EMBODIMENTS

Embodiments of the present disclosure are directed toward one or more computing devices, methods, and non-transitory computer-readable media for predicting a mean blood pressure (MBP) post induction and associated with post-induction hypotension (PIH) in a patient and recommending candidate therapies for intraoperative hypotension throughout each phase of anesthesia including induction, maintenance, and emergence. In certain embodiments, the one or more computing devices may access a data set including demographics and clinical data associated with the patient. For example, in one embodiment, the demographics and clinical data may include a set of preoperative input variables and anesthetic induction input variables associated with the patient. In certain embodiments, the set of preoperative input variables and anesthetic induction input variables may include one or more of a set of demographics input variables, a set of preoperative comorbidity input variables, a set of preoperative laboratory testing input variables, a set of medical procedure input variables, a set of peri-induction medication input variables, or a set of pre-induction MBP measurement input variables.

In certain embodiments, the set of demographics input variables may include one or more of an age of the patient, a sex of the patient, a race or ethnicity of the patient, or a weight of the patient. In certain embodiments, the set of preoperative laboratory testing input variables may include serum concentrations of one or more of sodium, creatinine, blood urea nitrogen (BUN), hemoglobin, white blood cell (WBC) counts, albumin, bilirubin, aspartate transferase (AST), estimated glomerular filtration rate (EGFR), platelet concentration, international normalized ratio (INR), or partial thromboplastin time (PTT). In certain embodiments, the set of preoperative input variables and anesthetic induction input variables may include a set of peri-induction medication input variables. In certain embodiments, the set of peri-induction medication input variables may include a predetermined set of anesthetic agents. For example, in one embodiment, the predetermined set of anesthetic agents may be selected from the group consisting of propofol, etomidate, ketamine, dexmedetomidine, nitrous oxide (NO), halothane, sevoflurane, desflurane, isoflurane, midazolam, diazepam, lorazepam, alfentanil, sufentanil, remifentanil, fentanyl, atracurium, cisatracurium, pancuronium, vecuronium, rocuronium, succinylcholine, hydromorphone, hydrocodone, oxycodone, phenylephrine, and lidocaine.

In certain embodiments, the one or more computing devices may minimize the cross-validation loss function by minimizing a loss between a prediction of one or more MBP values associated with PIH in the patient and an experimentally-determined one or more MBP values associated with PIH in the patient. In certain embodiments, the set of learnable parameters may include one or more weights or decision variables determined by the machine learning model based at least in part on the preprocessed set of preoperative input variables and anesthetic induction input variables and the predetermined set of target MBP values. In certain embodiments, the updated set of hyperparameters may include one or more of an updated set of general parameters, an updated set of booster parameters, or an updated set of learning-task parameters.

In one embodiment, the machine learning model may include one or more of a gradient boosting model, an adaptive boosting (AdaBoost) model, an extreme gradient boosting (XGBoost) model, a light gradient boosted machine (LightGBM) model, or a categorical boosting (CatBoost) model. In another embodiment, the machine learning model may include one or more of a linear regression model, a ridge regression model, a Bayesian ridge regression model, a tree-based model, or a neural network. In certain embodiments, prior to accessing the data set including the demographics and clinical data, the one or more computing devices may split the data set into a training data set, a test data set, and a validation data set for training and executing the machine-learning model.

In certain embodiments, the one or more computing devices may execute a reinforcement-learning model on the prediction of the MBP post induction to generate one or more candidate therapies. In one embodiment, the one or more candidate therapies may comprise one or more of adjusting a dosage of an anesthetic agent, adding an anesthetic agent, ceasing the administration of an anesthetic agent, administrating a medication, or administrating a fluid.

In certain embodiments, the one or more computing devices may automatically apply one or more of the candidate therapies to the patient.

In certain embodiments, the one or more computing devices may receive a MBP measurement of the patient at a first timepoint post induction. Outputting the prediction of the MBP post induction may further comprise inputting the MBP measurement of the patient at the first timepoint post induction to the machine-learning model and outputting the prediction of the MBP at a second timepoint subsequent to the first timepoint. The prediction of the MBP at the second timepoint may be generated based on the MBP measurement of the patient at the first timepoint post induction.

In certain embodiments, the one or more computing devices may generate a report based on the prediction of the MBP post induction and associated with PIH in the patient and the one or more candidate therapies. The one or more computing devices may further transmit the report to a computing device associated with a clinician. In certain embodiments, the one or more computing devices may cause, based on the prediction of the MBP post induction and associated with PIH in the patient and the one or more candidate therapies, a user interface (UI) executing on the computing device to display a visual representation of the prediction of the MBP post induction and associated with PIH in the patient and the one or more candidate therapies.

FIG. 1 illustrates an example embodiment of a clinical and computing environment 100 that may be utilized to predict a mean blood pressure (MBP) post induction and associated with post-induction hypotension (PIH) in a patient, in accordance with the presently disclosed embodiments. As depicted, the clinical and computing environment 100 may include a number of patients 102A (e.g., “Patient 1”), 102B (e.g., “Patient 2”), 102C (e.g., “Patient 3”), and 102D (e.g., “Patient TV”) each associated with a number of respective anesthesia systems 104A (e.g., “System 1”), 104B (e.g., “ System 2”), 104C (e.g., “System 3”), and 104D (e.g., “System TV”) that may be suitable for providing, for example, oxygenation, ventilation, and administration of anesthetic agents to one or more of the number of patients 102A, 102B, 102C, and 102D. For example, in one embodiment, the number of patients 102 A, 102B, 102C, and 102D may each include a patient undergoing anesthesia, such as a patient within an operating room (OR) or emergency room (ER) scheduled to undergo one or more invasive surgical procedures.

As further depicted, in certain embodiments, the number of anesthesia systems 104A, 104B, 104C, and 104D may each include a respective infusion pump 106A, 106B, 106C, and 106D and a respective control display 108 A, 108B, 108C, and 108D. For example, the number of infusion pumps 106 A, 106B, 106C, and 106D may each include any medical device suitable for delivering oxygenation, ventilation, and administration of anesthetic agents to one or more of the number of patients 102 A, 102B, 102C, and 102D while being controlled and monitored by one or more bedside clinicians (e.g., anesthesiologist, anesthesia nurse, or other physician) utilizing the control displays 108 A, 108B, 108C, and 108D.

In certain embodiments, the anesthetic agents that may be delivered to one or more of the number of patients 102 A, 102B, 102C, and 102D may include a predetermined set of anesthetic agents may be selected from the group consisting of propofol, etomidate, ketamine, dexmedetomidine, nitrous oxide (NO), halothane, sevoflurane, desflurane, isoflurane, midazolam, diazepam, lorazepam, alfentanil, sufentanil, remifentanil, fentanyl, atracurium, cisatracurium, pancuronium, vecuronium, rocuronium, succinylcholine, hydromorphone, hydrocodone, oxycodone, phenylephrine, and lidocaine.

In certain embodiments, as further depicted by FIG. 1, the anesthesia systems 104 A, 104B, 104C, and 104D may be coupled to a computing platform 112 via one or more communication network(s) 110. In certain embodiments, the computing platform 112 may include, for example, a cloud-based computing architecture suitable for hosting and executing one or more machine-learning models 118 that may be trained to predict a MBP post induction and associated with PIH in one or more of the number of patients 102 A, 102B, 102C, and 102D in accordance with the presently disclosed embodiments. For example, in one embodiment, the computing platform 112 may include a Platform as a Service (PaaS) architecture, a Software as a Service (SaaS) architecture, an Infrastructure as a Service (laaS) architecture, a Compute as a Service (CaaS) architecture, a Data as a Service (DaaS) architecture, a Database as a Service (DBaaS) architecture, or other similar cloud-based computing architecture (e.g., “X” as a Service (XaaS)).

In certain embodiments, as further depicted by FIG. 1, the computing platform 112 may include one or more processing devices 114 (e.g., servers) and one or more data stores 116. For example, in some embodiments, the one or more processing devices 114 (e.g., servers) may include one or more general purpose processors, graphic processing units (GPUs), application-specific integrated circuits (ASICs), systems-on-chip (SoCs), microcontrollers, field-programmable gate arrays (FPGAs), central processing units (CPUs), application processors (APs), visual processing units (VPUs), neural processing units (NPUs), neural decision processors (NDPs), deep learning processors (DLPs), tensor processing units (TPUs), neuromorphic processing units (NPUs), or any of various other processing device(s) or artificial intelligence (Al) accelerators that may be suitable for inputting patient demographics and clinical data 119 into one or more machine-learning models 118 and executing the one or more machine-learning models 118 to generate one or more predictions based thereon. Similarly, the data stores 116 may include, for example, one or more internal databases that may be utilized to store the patient demographics and clinical data 119 and the one or more machine-learning models 118.

In certain embodiments, as will be discussed in greater detail below with respect to FIG. 2, upon the data stores 116 receiving and storing the patient demographics and clinical data 119, the one or more processing devices 114 (e.g., servers) may then access the patient demographics and clinical data 119 and execute the one or more machine-learning models 118 to generate one or more predictions of a mean blood pressure (MBP) post induction over a predetermined time interval 120 in one or more of the number of patients 102 A, 102B, 102C, and 102D based on the patient demographics and clinical data 119. For example, in accordance with the presently disclosed embodiments, the one or more processing devices 114 (e.g., servers) may load and execute the one or more machinelearning models 118 to predict a mean blood pressure (MBP) post induction over a predetermined time interval (e.g., to - tis minutes) corresponding to a post-induction period based on the patient demographics and clinical data 119 and generate and output the prediction of MBP post induction over the predetermined time interval 120. Specifically, in one embodiment, the one or more machine-learning models 118 may generate and output a prediction of a MBP value in terms of millimeters of mercury (mmHg) at each minute over a 15-minute time interval (e.g., to - tis minutes), which corresponds to a post-induction period for one or more of the number of patients 102A, 102B, 102C, and 102D.

In certain embodiments, the one or more processing devices 114 (e.g., servers) may execute the one or more machine-learning models 118 to recommend candidate therapies during or before a hypotension event. For example, using reinforcement learning techniques, the optimal therapy from a set of potential therapies for hypotension can be determined during or just before an episode of hypotension to reduce the burden of hypotension.

In certain embodiments, the one or more machine-learning models 118 may include one or more of a gradient boosting model, an adaptive boosting (AdaBoost) model, an extreme gradient boosting (XGBoost) model, a light gradient boosted machine (LightGBM) model, a categorical boosting (CatBoost) model, a linear regression model, a ridge regression model, a Bayesian ridge regression model, a tree-based model, or a neural network trained to generate and output the prediction of MBP post induction over the predetermined time interval 120, or a reinforcement-learning (RL) model trained to recommend candidate therapies during or before a hypotension event. In certain embodiments, as further illustrated by FIG. 1, the one or more processing devices 114 (e.g., servers) may then transmit the prediction of MBP post induction and the recommended candidate therapies over the predetermined time interval 120 to a computing device 122 and present a report 124 to a clinician 126 (e.g., a cardiologist, an anesthesiologist, a surgeon) that may be associated with one or more of the number of patients 102 A, 102B, 102C, and 102D.

In one embodiment, the report 124 may include a clinical report that may be associated with one or more of the number of patients 102 A, 102B, 102C, and 102D to be provided and displayed, for example, to the clinician 126 (e.g., a cardiologist, an anesthesiologist, a surgeon) for purposes of the real-time or near real-time diagnosis, prognosis, and treatment of PIH in one or more of the number of patients 102A, 102B, 102C, and 102D. In certain embodiments, based on the prediction of MBP post induction over the predetermined time interval 120 and recommended candidate therapies as presented in the report 124, the clinician 126 (e.g., a cardiologist, an anesthesiologist, a surgeon) may utilize the computing device 122 to transmit instructions to one or more of the anesthesia systems 104A, 104B, 104C, and 104D to adjust a dosage of an administration of an anesthetic agent to one or more of the number of patients 102A, 102B, 102C, and 102D and/or to cease the administration of the anesthetic agent to one or more of the number of patients 102A, 102B, 102C, and 102D in real-time or near real-time.

In certain embodiments, based on the prediction of MBP post induction over the predetermined time interval 120 and recommended candidate therapies, the one or more processing devices 114 (e.g., servers) may automatically transmit instructions to one or more of the anesthesia systems 104A, 104B, 104C, and 104D to apply one or more of the candidate therapies to one or more of the number of patients 102A, 102B, 102C, and 102D in real-time or near real-time. For example, the candidate therapies may comprise one or more of adjusting a dosage of an anesthetic agent, adding an anesthetic agent, ceasing the administration of an anesthetic agent, administrating a medication, or administrating a fluid.

FIG. 2 illustrates a workflow diagram 200 for predicting a mean blood pressure (MBP) post induction and associated with post-induction hypotension (PIH) in a patient and recommending candidate therapies, in accordance with the presently disclosed embodiments. In one embodiment, the workflow diagram 200 may include a workflow process that may be implemented and executed by the one or more processing devices 114 (e.g., servers) of computing platform 112 as discussed above with respect to FIG. 1. As depicted, the workflow diagram 200 may begin at functional block 202 with the one or more processing devices 114 accessing a data set. For example, the one or more processing devices 114 may access a data set including demographics and clinical data associated with one or more of the number of patients 102A, 102B, 102C, and 102D.

In certain embodiments, the demographics and clinical data may include a set of preoperative input variables and anesthetic induction input variables associated with the patient, such as one or more of a set of demographics input variables, a set of preoperative comorbidity input variables, a set of preoperative laboratory testing input variables, a set of medical procedure input variables, a set of peri-induction medication input variables, or a set of pre-induction MBP measurement input variables. In certain embodiments, the set of demographics input variables may include one or more of an age of the patient, a sex of the patient, a race or ethnicity of the patient, or a weight of the patient. In certain embodiments, the set of preoperative laboratory testing input variables may include serum concentrations of one or more of sodium, creatinine, blood urea nitrogen (BUN), hemoglobin, white blood cell (WBC) counts, albumin, bilirubin, aspartate transferase (AST), estimated glomerular filtration rate (EGFR), platelet concentration, international normalized ratio (INR), or partial thromboplastin time (PTT).

In certain embodiments, the set of preoperative input variables and anesthetic induction input variables may include a set of peri-induction medication input variables. In certain embodiments, the set of peri-induction medication input variables may include a predetermined set of anesthetic agents. For example, in one embodiment, the predetermined set of anesthetic agents may be selected from the group consisting of propofol, etomidate, ketamine, dexmedetomidine, nitrous oxide (NO), halothane, sevoflurane, desflurane, isoflurane, midazolam, diazepam, lorazepam, alfentanil, sufentanil, remifentanil, fentanyl, atracurium, cisatracurium, pancuronium, vecuronium, rocuronium, succinylcholine, hydromorphone, hydrocodone, oxycodone, phenylephrine, and lidocaine. The set of periinduction medication input variables may include the anesthetic agents and their respective dosages.

In certain embodiments, the data set may include Electronic Health Record (EHR) data. In an embodiment, the EHR data includes over 93,000 anesthetic records of elective non-cardiac surgeries. The variables included in the HER data include demographic variables (age, race, sex, weight, co-morbidities including left ventricular function), surgical variables (type of procedure, ASA status) and pre-induction variables (medication administration and doses, pre-induction MBP). The variables in the HER data additionally include the MBP in the first 15 minutes after induction of anesthesia.

Additionally, the data set may include intraoperative data in a structured format. The intraoperative data may be generated by Anesthesia Information Management Systems (AIMS) from the monitor that is captured automatically, providing an un-adulterated stream. The data set may further include medication, fluid, and blood administration data that are manually entered into AIMS systems on the same intraoperative timeline.

In certain embodiments, the workflow diagram 200 may then continue at functional block 204 with the one or more processing devices 114 preprocessing the accessed data set and at functional block 206 with the one or more processing devices 114 performing a feature generation and extraction. For example, the one or more processing devices 114 may preprocess the accessed data set by executing one or more encoding processes on one or more input variables of the set of preoperative input variables and anesthetic induction input variables. For example, in one embodiment, the one or more processing devices 114 may execute the one or more encoding processes on the one or more input variables by executing an imputation (e.g., ^-nearest neighbor imputation) of one or more missing values associated with the one or more input variables. In another embodiment, the one or more processing devices 114 may execute the one or more encoding processes on the one or more input variables by executing a one-hot encoding of the one or more input variables. In another embodiment, the one or more processing devices 114 may execute a preprocessing process, which includes a normalization and scaling of one or more input variables when the one or more input variables includes continuous or fixed values, such as sex, age, and weight.

In another embodiment, the one or more processing devices 114 may preprocess the MBP measurements in the HER data. For example, arterial lines may lead to noises in the MBP measurements and the one or more processing devices 114 may remove these noises.

In another embodiment, the one or more processing devices 114 may preprocess the medication input variables. For example, medications can be bolus or infusion. The one or more processing devices 114 may convert these two types of medication to a uniformed timeline, which represents a medication being administered into the body as a timesequence signal.

In certain embodiments, the workflow diagram 200 may then continue at functional block 208 with the one or more processing devices 114 training the one or more machinelearning models 118, at functional block 210 with the one or more processing devices 114 evaluating the one or more machine-learning models 118, and at functional block 212 with the one or more processing devices 114 executing the one or more machine-learning models 118. For example, as previously noted above with respect to FIG. 1, the one or more machine-learning models 118 may include one or more of a gradient boosting model, an adaptive boosting (AdaBoost) model, an extreme gradient boosting (XGBoost) model, a light gradient boosted machine (LightGBM) model, a categorical boosting (CatBoost) model, a linear regression model, a ridge regression model, a Bayesian ridge regression model, a tree-based model, or a neural network. The one or more machine-learning models 118 may also include a reinforcement-learning (RL) model trained to recommend candidate therapies during or before a hypotension event.

Accordingly, in certain embodiments, the one or more processing devices 114 may first split the accessed data set into a training data set, a test data set, and a validation data set for training and executing the one or more machine-learning models 118. For example, each intraoperative record may be split, e.g., into 20-minute intervals, using a moving window, where the first 10 minutes will be included into the training data and the validation data will be the MBP in the second 10 minutes of the window.

In certain embodiments, the one or more processing devices 114 may then train and execute the one or more machine-learning models 118 by refining a set of hyperparameters associated with the one or more machine-learning models 118 iteratively until a desired precision is reached. For example, the one or more processing devices 114 may validate and evaluate the one or more machine-learning models 118 by calculating one or more cross- validation losses based on the preprocessed set of preoperative input variables and anesthetic induction input variables and a predetermined set of target MBP values, and further updating the set of hyperparameters based on the one or more cross-validation losses. As another example, the one or more machine-learning models 118 may be evaluated using R-squared value, mean absolute error, mean squared error, and mean absolute percentage error. The R-squared may be used for refining the set of hyperparameters. In certain embodiments, the one or more machine-learning models 118 may thus output the prediction of MBP post induction over the predetermined time interval 120 based on the updated set of hyperparameters.

In certain embodiments, the one or more processing devices 114 may calculate the one or more cross-validation losses by evaluating a cross-validation loss function based on the preprocessed set of preoperative input variables and anesthetic induction input variables, the predetermined set of target MBP values, the set of hyperparameters, and a set of learnable parameters associated with the one or more machine-learning models 118. In certain embodiments, the one or more processing devices 114 may further minimize the cross-validation loss function by varying the set of learnable parameters while the preprocessed set of preoperative input variables and anesthetic induction input variables, the predetermined set of target MBP values, the set of hyperparameters remain constant. For example, in one embodiment, the one or more processing devices 114 may minimize the cross-validation loss function by optimizing the set of hyperparameters. In one embodiment, the set of hyperparameters may include one or more of a set of general parameters, a set of booster parameters, or a set of learning-task parameters. In certain embodiments, the one or more computing devices may minimize the cross-validation loss function by minimizing a loss between a prediction of one or more MBP values and an experimentally-determined or measured one or more MBP values. In certain embodiments, the one or more processing devices 114 may execute the machine-learning models to predict MBP based on the preprocessed data sets and measured MBP post induction associated with a patient. For example, the one or more processing devices 114 may receive a MBP measurement of the patient at a first timepoint post induction. The one or more processing devices 114 may input the MBP measurement of the patient at the first timepoint post induction together with the preprocessed data sets to the machine-learning models, which may output the prediction of the MBP at a second timepoint subsequent to the first timepoint.

In certain embodiments, the one or more machine-learning models 118 may include a reinforcement-learning (RL) model. The reinforcement-learning model may be trained to utilize intraoperative anesthesia record data to evaluate and recommend potential candidate therapies for intraoperative hypotension throughout each unique phase of anesthesia (induction, maintenance, and emergence phases). Managing intraoperative hypotension can be categorized as a task in the sequential decision-making domain, and the reinforcementlearning model provides a formal framework for making such decisions. In the reinforcement-learning model, an agent or algorithm, interprets the clinical condition and takes an action, and based on the impact of the action the agent/algorithm receives a positive reward (if the clinical condition improves) or a negative reward (if the condition deteriorates).

In certain embodiments, the reinforcement-learning model may be based on Q- leaming. Q-leaming is a machine learning algorithm that teaches an agent to take actions that maximize rewards over time. For Q-learning, the one or more processing devices 114 may use a simple representation for predicted patient’s blood pressure status. For example, the blood pressure status may include categories like normal, mild hypotension (e.g., drop in blood pressure of 10 mmHg or less compared to preinduction blood pressure), moderate hypotension (e.g., drop in blood pressure of 10 mmHg to 20 mmHg compared to preinduction BP), severe hypotension (e.g., drop in blood pressureof greater than 20 mmHg compared to preinduction BP), a simple set of actions (e.g., do nothing, give medications to raise blood pressure, give fluids to raise BP), and a simple reward system (e.g., +1 reward if the patient’s low blood pressure improves (e.g., from mild to normal, or moderate to normal or severe to normal), -1 reward if the patient’s blood pressure worsens or if side effects appear, 0 reward if there’s no change). In certain embodiments, the workflow diagram 200 may then continue at functional block 214 and functional block 220 with the one or more processing devices 114 generating and executing a clinical support user interface (UI) based on, for example, the prediction of MBP post induction over the predetermined time interval 120 and the recommended candidate therapies. In some embodiments, the clinical support UI may also source and include one or more user experience (UX) studies inputs 216 and heuristics models inputs 218. Specifically, as previously discussed above with respect to FIG. 1, the one or more processing devices 114 may generate a report 124 based on the prediction of MBP post induction over the predetermined time interval 120 and the recommended candidate therapies, and further transmit the report 124 to the computing device 122 associated with the clinician 126 (e.g., a cardiologist, an anesthesiologist, a surgeon).

In certain embodiments, the one or more processing devices 114 may then cause the clinical support UI executing on the computing device 122 to display a visual representation of the prediction of MBP post induction over the predetermined time interval 120 and the recommended candidate therapies. In certain embodiments, based on the prediction of MBP post induction over the predetermined time interval 120 and the recommended candidate therapies as presented in the report 124, the clinician 126 (e.g., a cardiologist, an anesthesiologist, a surgeon) may utilize the computing device 122 to transmit instructions to one or more of the anesthesia systems 104A, 104B, 104C, and 104D to adjust a dosage of an administration of an anesthetic agent to one or more of the number of patients 102A, 102B, 102C, and 102D and/or to cease the administration of the anesthetic agent to one or more of the number of patients 102A, 102B, 102C, and 102D in real-time or near real-time. In certain embodiments, the workflow diagram 200 may then conclude at functional block 222 with the one or more processing devices 114 evaluating the clinical support UI by capturing various inputs performed by the clinician 126 (e.g., a cardiologist, an anesthesiologist, a surgeon) and updating (e.g., fine-tuning, retraining) the one or more machine-learning models 118 based thereon.

In certain embodiments, a software clinical decision support (CDS) tool may be provided for acquiring and processing data and applying the one or more machine-learning models 118. The CDS tool may take as input anesthetic data streams from past patient records, apply the blood pressure (BP) forecasting and therapy recommendation models, and display the model outputs alongside the anesthetic data. In certain embodiments, the CDS tool may include a data parser module capable of parsing anesthesia data streams and performing the necessary preprocessing for the application of models. The CDS tool may include a blood pressure forecast module that applies the machine-learning models 118 trained for blood pressure forecasting to the data and outputs blood pressure forecasts for the upcoming timeframe (e.g., 5 minutes). The CDS tool may include a therapy recommendation module that consumes the data and blood pressure forecasts and outputs therapeutic recommendations. The CDS tool may include a display module that provides a running display of the data, forecasted blood pressure values, and recommendations.

FIG. 3 illustrates an example embodiment of an anesthesia clinical support user interface (UI) 300 of the CDS tool for displaying a prediction of a mean blood pressure (MBP) post induction and associated with post-induction hypotension (PIH) in a patient, in accordance with the presently disclosed embodiments. As depicted, the anesthesia clinical support UI 300 may include a graphical display of a post-induction of a listing of induction medication 320 (i.e., anesthetic agents) being delivered or having been delivered to a patient “John Doe.” The anesthesia clinical support UI 300 may also include the biological information 310 of the patient. As further depicted, the anesthesia clinical support UI 300 may include a graphical display of a post-induction MBP 330 of the patient “John Doe.” In accordance with the presently disclosed embodiments, the post-induction MBP of the patient “John Doe” may include a generated prediction of a MBP value in terms of mmHg at each minute over a 15-minute time interval (e.g., to - tis minutes), which corresponds to a post-induction period for the patient “John Doe.”

FIG. 4 illustrates a flow diagram of a method 400 for predicting a mean blood pressure (MBP) post induction and associated with post-induction hypotension (PIH) in a patient, in accordance with the disclosed embodiments. The method 400 may be performed utilizing one or more processing devices 114 that may include hardware (e.g., a general purpose processor, a graphic processing unit (GPU), an application-specific integrated circuit (ASIC), a system-on-chip (SoC), a microcontroller, a field-programmable gate array (FPGA), a central processing unit (CPU), an application processor (AP), a visual processing unit (VPU), a neural processing unit (NPU), a neural decision processor (NDP), a deep learning processor (DLP), a tensor processing unit (TPU), a neuromorphic processing unit (NPU), or any other artificial intelligence (Al) / machine-learning (ML) accelerators device(s) that may be suitable for processing medical data and making one or more predictions or decisions based thereon), firmware (e.g., microcode), or some combination thereof. The method 400 may begin at block 402 with the one or more processing devices 114 accessing a data set including demographics and clinical data associated with a patient. For example, in one embodiment, the demographics and clinical data may include a set of preoperative input variables and anesthetic induction input variables associated with the patient. The method 400 may then continue at block 404 with the one or more processing devices 114 refining a set of hyperparameters associated with a machine-learning model trained to generate a prediction of a mean blood pressure (MBP) post induction and associated with post-induction hypotension (PUT) in the patient. For example, in certain embodiments, refining the set of hyperparameters may include iteratively executing a process until a desired precision is reached.

The method 400 may continue at block 406 with the one or more processing devices 114 preprocessing the data set by executing one or more encoding processes on one or more input variables of the set of preoperative input variables and anesthetic induction input variables. For example, in some embodiments, the one or more processing devices 114 may execute the one or more encoding processes on the one or more input variables by executing an imputation of one or more missing values associated with the one or more input variables. In some embodiments, the one or more processing devices 114 may further execute the one or more encoding processes on the one or more input variables by executing a one-hot encoding of the one or more input variables. The method 400 may then continue at block 408 with the one or more processing devices 114 calculating one or more cross- validation losses based at least in part on the preprocessed set of preoperative input variables and anesthetic induction input variables and a predetermined set of target MBP values.

The method 400 may then continue at block 410 with the one or more processing devices 114 updating the set of hyperparameters based on the one or more cross-validation losses. For example, in certain embodiments, calculating the one or more cross-validation losses may include evaluating a cross-validation loss function based on the preprocessed set of preoperative input variables and anesthetic induction input variables, the predetermined set of target MBP values, the set of hyperparameters, and a set of learnable parameters associated with the machine-learning model, and further minimizing the cross-validation loss function by varying the set of learnable parameters while the preprocessed set of preoperative input variables and anesthetic induction input variables, the predetermined set of target MBP values, the set of hyperparameters remain constant. The method 400 may then conclude at block 412 with the one or more processing devices 114 outputting, by the machine learning model, the prediction of the MBP post induction and associated with PUT in the patient based at least in part on the updated set of hyperparameters. For example, in certain embodiments, the one or more processing devices 114 may generate a report based on the prediction of the MBP post induction and associated with PUT in the patient, and further transmit the report to a computing device associated with a clinician. In certain embodiments, the one or more processing devices 114 may further cause, based on the prediction of the MBP associated with PUT in the patient, a user interface (UI) executing on the computing device to display a visual representation of the prediction of the MBP post induction and associated with PUT in the patient.

FIG. 5 illustrates an example of one or more computing device(s) 500 that may be utilized for predicting a mean blood pressure (MBP) post induction and associated with post-induction hypotension (PIH) in a patient and recommending candidate therapies, in accordance with the presently disclosed embodiments. In certain embodiments, the one or more computing device(s) 500 may perform one or more steps of one or more methods described or illustrated herein. In certain embodiments, the one or more computing device(s) 500 provide functionality described or illustrated herein. In certain embodiments, software running on the one or more computing device(s) 500 performs one or more steps of one or more methods described or illustrated herein, or provides functionality described or illustrated herein. Certain embodiments include one or more portions of the one or more computing device(s) 500.

This disclosure contemplates any suitable number of computing device(s) 500. This disclosure contemplates one or more computing device(s) 500 taking any suitable physical form. As example and not by way of limitation, one or more computing device(s) 500 may be an embedded computer system, a system-on-chip (SOC), a single-board computer system (SBC) (e.g., a computer-on-module (COM) or system-on-module (SOM)), a desktop computer system, a laptop or notebook computer system, an interactive kiosk, a mainframe, a mesh of computer systems, a mobile telephone, a personal digital assistant (PDA), a server, a tablet computer system, an augmented/virtual reality device, or a combination of two or more of these. Where appropriate, the one or more computing device(s) 500 may be unitary or distributed; span multiple locations; span multiple machines; span multiple data centers; or reside in a cloud, which may include one or more cloud components in one or more networks. Where appropriate, the one or more computing device(s) 500 may perform without substantial spatial or temporal limitation one or more steps of one or more methods described or illustrated herein. As an example, and not by way of limitation, the one or more computing device(s) 500 may perform, in real-time or in batch mode, one or more steps of one or more methods described or illustrated herein. The one or more computing device(s) 500 may perform, at different times or at different locations, one or more steps of one or more methods described or illustrated herein, where appropriate.

In certain embodiments, the one or more computing device(s) 500 includes a processor 502, memory 504, database 506, an input/output (I/O) interface 508, a communication interface 510, and a bus 512. Although this disclosure describes and illustrates a particular computer system having a particular number of particular components in a particular arrangement, this disclosure contemplates any suitable computer system having any suitable number of any suitable components in any suitable arrangement. In certain embodiments, processor 502 includes hardware for executing instructions, such as those making up a computer program. As an example, and not by way of limitation, to execute instructions, processor 502 may retrieve (or fetch) the instructions from an internal register, an internal cache, memory 504, or database 506; decode and execute them; and then write one or more results to an internal register, an internal cache, memory 504, or database 506. In certain embodiments, processor 502 may include one or more internal caches for data, instructions, or addresses. This disclosure contemplates processor 502 including any suitable number of any suitable internal caches, where appropriate. As an example, and not by way of limitation, processor 502 may include one or more instruction caches, one or more data caches, and one or more translation lookaside buffers (TLBs). Instructions in the instruction caches may be copies of instructions in memory 504 or database 506, and the instruction caches may speed up retrieval of those instructions by processor 502.

Data in the data caches may be copies of data in memory 504 or database 506 for instructions executing at processor 502 to operate on; the results of previous instructions executed at processor 502 for access by subsequent instructions executing at processor 502 or for writing to memory 504 or database 506; or other suitable data. The data caches may speed up read or write operations by processor 502. The TLBs may speed up virtual-address translation for processor 502. In certain embodiments, processor 502 may include one or more internal registers for data, instructions, or addresses. This disclosure contemplates processor 502 including any suitable number of any suitable internal registers, where appropriate. Where appropriate, processor 502 may include one or more arithmetic logic units (ALUs); be a multi-core processor; or include one or more processors 502. Although this disclosure describes and illustrates a particular processor, this disclosure contemplates any suitable processor.

In certain embodiments, memory 504 includes main memory for storing instructions for processor 502 to execute or data for processor 502 to operate on. As an example, and not by way of limitation, the one or more computing device(s) 500 may load instructions from database 506 or another source (such as, for example, another one or more computing device(s) 500) to memory 504. Processor 502 may then load the instructions from memory 504 to an internal register or internal cache. To execute the instructions, processor 502 may retrieve the instructions from the internal register or internal cache and decode them. During or after execution of the instructions, processor 502 may write one or more results (which may be intermediate or final results) to the internal register or internal cache. Processor 502 may then write one or more of those results to memory 504.

In certain embodiments, processor 502 executes only instructions in one or more internal registers, internal caches, or memory 504 (as opposed to database 506 or elsewhere) and operates only on data in one or more internal registers, internal caches, or memory 504 (as opposed to database 506 or elsewhere). One or more memory buses (which may each include an address bus and a data bus) may couple processor 502 to memory 504. Bus 512 may include one or more memory buses, as described below. In certain embodiments, one or more memory management units (MMUs) reside between processor 502 and memory 504 and facilitate accesses to memory 504 requested by processor 502. In certain embodiments, memory 504 includes random access memory (RAM). This RAM may be volatile memory, where appropriate. Where appropriate, this RAM may be dynamic RAM (DRAM) or static RAM (SRAM). Moreover, where appropriate, this RAM may be singleported or multi-ported RAM. This disclosure contemplates any suitable RAM. Memory 504 may include one or more memory 504, where appropriate. Although this disclosure describes and illustrates particular memory, this disclosure contemplates any suitable memory.

In certain embodiments, database 506 includes mass storage for data or instructions. As an example, and not by way of limitation, database 506 may include a hard disk drive (HDD), a floppy disk drive, flash memory, an optical disc, a magneto-optical disc, magnetic tape, or a Universal Serial Bus (USB) drive, or a combination of two or more of these. Database 506 may include removable or non-removable (or fixed) media, where appropriate. Database 506 may be internal or external to the one or more computing device(s) 500, where appropriate. In certain embodiments, database 506 is non-volatile, solid-state memory. In certain embodiments, database 506 includes read-only memory (ROM). Where appropriate, this ROM may be mask-programmed ROM, programmable ROM (PROM), erasable PROM (EPROM), electrically erasable PROM (EEPROM), electrically alterable ROM (EAROM), flash memory, or a combination of two or more of these. This disclosure contemplates mass database 506 taking any suitable physical form. Database 506 may include one or more storage control units facilitating communication between processor 502 and database 506, where appropriate. Where appropriate, database 506 may include one or more databases 506. Although this disclosure describes and illustrates particular storage, this disclosure contemplates any suitable storage.

In certain embodiments, VO interface 508 includes hardware, software, or both, providing one or more interfaces for communication between the one or more computing device(s) 500 and one or more VO devices. The one or more computing device(s) 500 may include one or more of these VO devices, where appropriate. One or more of these VO devices may enable communication between a person and the one or more computing device(s) 500. As an example, and not by way of limitation, an VO device may include a keyboard, keypad, microphone, monitor, mouse, printer, scanner, speaker, still camera, stylus, tablet, touch screen, trackball, video camera, another suitable VO device, or a combination of two or more of these. An VO device may include one or more sensors. This disclosure contemplates any suitable VO devices and any suitable VO interfaces 508 for them. Where appropriate, VO interface 508 may include one or more device or software drivers enabling processor 502 to drive one or more of these VO devices. VO interface 508 may include one or more VO interfaces 508, where appropriate. Although this disclosure describes and illustrates a particular VO interface, this disclosure contemplates any suitable VO interface.

In certain embodiments, communication interface 510 includes hardware, software, or both providing one or more interfaces for communication (such as, for example, packetbased communication) between the one or more computing device(s) 500 and one or more other computing device(s) 500 or one or more networks. As an example, and not by way of limitation, communication interface 510 may include a network interface controller (NIC) or network adapter for communicating with an Ethernet or other wire-based network or a wireless NIC (WNIC) or wireless adapter for communicating with a wireless network, such as a WI-FI network. This disclosure contemplates any suitable network and any suitable communication interface 510 for it.

As an example, and not by way of limitation, the one or more computing device(s) 500 may communicate with an ad hoc network, a personal area network (PAN), a local area network (LAN), a wide area network (WAN), a metropolitan area network (MAN), one or more portions of the Internet, or a combination of two or more of these. One or more portions of one or more of these networks may be wired or wireless. As an example, the one or more computing device(s) 500 may communicate with a wireless PAN (WPAN) (such as, for example, a BLUETOOTH WPAN), a WI-FI network, a WI-MAX network, a cellular telephone network (such as, for example, a Global System for Mobile Communications (GSM) network), other suitable wireless network, or a combination of two or more of these. The one or more computing device(s) 500 may include any suitable communication interface 510 for any of these networks, where appropriate. Communication interface 510 may include one or more communication interfaces 510, where appropriate. Although this disclosure describes and illustrates a particular communication interface, this disclosure contemplates any suitable communication interface.

In certain embodiments, bus 512 includes hardware, software, or both coupling components of the one or more computing device(s) 500 to each other. As an example, and not by way of limitation, bus 512 may include an Accelerated Graphics Port (AGP) or other graphics bus, an Enhanced Industry Standard Architecture (EISA) bus, a front-side bus (FSB), a HYPERTRANSPORT (HT) interconnect, an Industry Standard Architecture (ISA) bus, an INFINIBAND interconnect, a low-pin-count (LPC) bus, a memory bus, a Micro Channel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI- Express (PCIe) bus, a serial advanced technology attachment (SATA) bus, a Video Electronics Standards Association local (VLB) bus, another suitable bus, or a combination of two or more of these. Bus 512 may include one or more buses 512, where appropriate. Although this disclosure describes and illustrates a particular bus, this disclosure contemplates any suitable bus or interconnect.

Herein, a computer-readable non-transitory storage medium or media may include one or more semiconductor-based or other integrated circuits (ICs) (such, as for example, field-programmable gate arrays (FPGAs) or application-specific ICs (ASICs)), hard disk drives (HDDs), hybrid hard drives (HHDs), optical discs, optical disc drives (ODDs), magneto-optical discs, magneto-optical drives, floppy diskettes, floppy disk drives (FDDs), magnetic tapes, solid-state drives (SSDs), RAM-drives, SECURE DIGITAL cards or drives, any other suitable computer-readable non-transitory storage media, or any suitable combination of two or more of these, where appropriate. A computer-readable non- transitory storage medium may be volatile, non-volatile, or a combination of volatile and non-volatile, where appropriate.

FIG. 6 illustrates a diagram 600 of an example artificial intelligence (Al) architecture 602 (which may be included as part of the one or more computing device(s) 500 as discussed above with respect to FIG. 5) that may be utilized for predicting a mean blood pressure (MBP) post induction and associated with post-induction hypotension (PIH) in a patient, in accordance with the disclosed embodiments. In certain embodiments, the Al architecture 602 may be implemented utilizing, for example, one or more processing devices that may include hardware (e.g., a general purpose processor, a graphic processing unit (GPU), an application-specific integrated circuit (ASIC), a system-on-chip (SoC), a microcontroller, a field-programmable gate array (FPGA), a central processing unit (CPU), an application processor (AP), a visual processing unit (VPU), a neural processing unit (NPU), a neural decision processor (NDP), a deep learning processor (DLP), a tensor processing unit (TPU), a neuromorphic processing unit (NPU), and/or other artificial intelligence (Al) / machine-learning (ML) accelerator device(s) that may be suitable for processing various data and making one or more predictions or decisions based thereon), software (e.g., instructions running/executing on one or more processing devices), firmware (e.g., microcode), or some combination thereof.

In certain embodiments, as depicted by FIG. 6, the Al architecture 602 may include machine learning (ML) models 604, natural language processing (NLP) models 606, expert systems 608, computer-based vision models 610, speech recognition models 612, planning models 614, and robotics models 616. In certain embodiments, the ML models 604 may include any statistics-based models that may be suitable for finding patterns across large amounts of data (e.g., “Big Data” such as genomics data, proteomics data, metabolomics data, metagenomics data, transcriptomics data, or other omics data). For example, in certain embodiments, the ML models 604 may include deep learning models 618, supervised learning models 620, and unsupervised learning models 622. In certain embodiments, the deep learning models 618 may include any artificial neural networks (ANNs) that may be utilized to learn deep levels of representations and abstractions from large amounts of data. For example, the deep learning models 618 may include ANNs, such as a perceptron, a multilayer perceptron (MLP), an autoencoder (AE), a convolution neural network (CNN), a recurrent neural network (RNN), long short term memory (LSTM), a gated recurrent unit (GRU), a restricted Boltzmann Machine (RBM), a deep belief network (DBN), a bidirectional recurrent deep neural network (BRDNN), a generative adversarial network (GAN), and deep Q-networks, a neural autoregressive distribution estimation (NADE), an adversarial network (AN), attentional models (AM), a spiking neural network (SNN), deep reinforcement learning, and so forth.

In certain embodiments, the supervised learning models 620 may include any algorithms that may be utilized to apply, for example, what has been learned in the past to new data using labeled examples for predicting future events. For example, starting from the analysis of a known training data set, the supervised learning models 620 may produce an inferred function to make predictions about the output values. The supervised learning models 620 may also compare its output with the correct and intended output and find errors in order to modify the supervised learning models 620 accordingly. On the other hand, the unsupervised learning models 622 may include any algorithms that may be applied, for example, when the data used to train the unsupervised learning models 622 are neither classified nor labeled. For example, the unsupervised learning models 622 may study and analyze how systems may infer a function to describe a hidden structure from unlabeled data.

In certain embodiments, the NLP models 606 may include any algorithms or functions that may be suitable for automatically manipulating natural language, such as speech and/or text. For example, in some embodiments, the NLP models 606 may include content extraction models 624, classification models 626, machine translation models 628, question answering (QA) models 630, and text generation models 632. In certain embodiments, the content extraction models 624 may include a means for extracting text or images from electronic documents (e.g., webpages, text editor documents, and so forth) to be utilized, for example, in other applications.

In certain embodiments, the classification models 626 may include any algorithms that may utilize a supervised learning model (e.g., logistic regression, naive Bayes, stochastic gradient descent (SGD), ^-nearest neighbors, decision trees, random forests, support vector machine (SVM), and so forth) to learn from the data input to the supervised learning model and to make new observations or classifications based thereon. The machine translation models 628 may include any algorithms or functions that may be suitable for automatically converting source text in one language, for example, into text in another language. The QA models 630 may include any algorithms or functions that may be suitable for automatically answering questions posed by humans in, for example, a natural language, such as that performed by voice-controlled personal assistant devices. The text generation models 632 may include any algorithms or functions that may be suitable for automatically generating natural language texts.

In certain embodiments, the expert systems 608 may include any algorithms or functions that may be suitable for simulating the judgment and behavior of a human or an organization that has expert knowledge and experience in a particular field (e.g., stock trading, medicine, sports statistics, and so forth). The computer-based vision models 610 may include any algorithms or functions that may be suitable for automatically extracting information from images (e.g., photo images, video images). For example, the computer- based vision models 610 may include image recognition algorithms 634 and machine vision algorithms 636. The image recognition algorithms 634 may include any algorithms that may be suitable for automatically identifying and/or classifying objects, places, people, and so forth that may be included in, for example, one or more image frames or other displayed data. The machine vision algorithms 636 may include any algorithms that may be suitable for allowing computers to “see”, or, for example, to rely on image sensors or cameras with specialized optics to acquire images for processing, analyzing, and/or measuring various data characteristics for decision making purposes.

In certain embodiments, the speech recognition models 612 may include any algorithms or functions that may be suitable for recognizing and translating spoken language into text, such as through automatic speech recognition (ASR), computer speech recognition, speech-to-text (STT) 638, or text-to-speech (TTS) 640 in order for the computing to communicate via speech with one or more users, for example. In certain embodiments, the planning models 614 may include any algorithms or functions that may be suitable for generating a sequence of actions, in which each action may include its own set of preconditions to be satisfied before performing the action. Examples of Al planning may include classical planning, reduction to other problems, temporal planning, probabilistic planning, preference-based planning, conditional planning, and so forth. Lastly, the robotics models 616 may include any algorithms, functions, or systems that may enable one or more devices to replicate human behavior through, for example, motions, gestures, performance tasks, decision-making, emotions, and so forth.

Herein, “or” is inclusive and not exclusive, unless expressly indicated otherwise or indicated otherwise by context. Therefore, herein, “A or B” means “A, B, or both,” unless expressly indicated otherwise or indicated otherwise by context. Moreover, “and” is both joint and several, unless expressly indicated otherwise or indicated otherwise by context. Therefore, herein, “A and B” means “A and B, jointly or severally,” unless expressly indicated otherwise or indicated otherwise by context.

Herein, “automatically” and its derivatives means “without human intervention,” unless expressly indicated otherwise or indicated otherwise by context.

The embodiments disclosed herein are only examples, and the scope of this disclosure is not limited to them. Embodiments according to this disclosure are in particular disclosed in the attached claims directed to a method, a storage medium, a system and a computer program product, wherein any feature mentioned in one claim category, e.g. method, may be claimed in another claim category, e.g. system, as well. The dependencies or references back in the attached claims are chosen for formal reasons only. However, any subject matter resulting from a deliberate reference back to any previous claims (in particular multiple dependencies) may be claimed as well, so that any combination of claims and the features thereof are disclosed and may be claimed regardless of the dependencies chosen in the attached claims. The subject-matter which may be claimed comprises not only the combinations of features as set out in the attached claims but also any other combination of features in the claims, wherein each feature mentioned in the claims may be combined with any other feature or combination of other features in the claims. Furthermore, any of the embodiments and features described or depicted herein may be claimed in a separate claim and/or in any combination with any embodiment or feature described or depicted herein or with any of the features of the attached claims.

The scope of this disclosure encompasses all changes, substitutions, variations, alterations, and modifications to the example embodiments described or illustrated herein that a person having ordinary skill in the art would comprehend. The scope of this disclosure is not limited to the example embodiments described or illustrated herein. Moreover, although this disclosure describes and illustrates respective embodiments herein as including particular components, elements, feature, functions, operations, or steps, any of these embodiments may include any combination or permutation of any of the components, elements, features, functions, operations, or steps described or illustrated anywhere herein that a person having ordinary skill in the art would comprehend. Furthermore, reference in the appended claims to an apparatus or system or a component of an apparatus or system being adapted to, arranged to, capable of, configured to, enabled to, operable to, or operative to perform a particular function encompasses that apparatus, system, component, whether or not it or that particular function is activated, turned on, or unlocked, as long as that apparatus, system, or component is so adapted, arranged, capable, configured, enabled, operable, or operative. Additionally, although this disclosure describes or illustrates certain embodiments as providing particular advantages, certain embodiments may provide none, some, or all of these advantages.

Claims

CLAIMS What is claimed is:

1. A method for predicting a mean blood pressure (MBP) post induction and associated with post-induction hypotension (PIH) in a patient, comprising, by one or more computing devices: accessing a data set comprising demographics and clinical data associated with the patient, wherein the demographics and clinical data comprise a set of preoperative input variables and anesthetic induction input variables associated with the patient; and refining a set of hyperparameters associated with a machine-learning model trained to generate a prediction of a mean blood pressure (MBP) post induction and associated with post-induction hypotension (PIH) in the patient, wherein refining the set of hyperparameters comprises iteratively executing a process until a desired precision is reached, the process, comprising: preprocessing the data set by executing one or more encoding processes on one or more input variables of the set of preoperative input variables and anesthetic induction input variables; calculating one or more cross-validation losses based at least in part on the preprocessed set of preoperative input variables and anesthetic induction input variables and a predetermined set of target MBP values; updating the set of hyperparameters based on the one or more cross- validation losses; and outputting, by the machine-learning model, the prediction of the MBP post induction and associated with PIH in the patient based at least in part on the updated set of hyperparameters.

2. The method of Claim 1, wherein outputting the prediction of the MBP post induction comprises outputting, by the machine-learning model, a prediction of MBP post induction over a predetermined time interval corresponding to a post-induction period associated with the patient.

3. The method of any of Claims 1-2, wherein the set of preoperative input variables and anesthetic induction input variables comprises one or more of a set of demographics input variables, a set of preoperative comorbidity input variables, a set of preoperative laboratory testing input variables, a set of medical procedure input variables, a set of periinduction medication input variables, or a set of pre-induction MBP measurement input variables.

4. The method of any of Claims 1-3, wherein the set of demographics input variables comprises one or more of an age of the patient, a sex of the patient, a race or ethnicity of the patient, or a weight of the patient.

5. The method of any of Claims 1-4, wherein the set of preoperative laboratory testing input variables comprises serum concentrations of one or more of sodium, creatinine, blood urea nitrogen (BUN), hemoglobin, white blood cell (WBC) counts, albumin, bilirubin, aspartate transferase (AST), estimated glomerular filtration rate (EGFR), platelet concentration, international normalized ratio (INR), or partial thromboplastin time (PTT).

6. The method of any of Claims 1-5, wherein the set of preoperative input variables and anesthetic induction input variables comprises a set of peri-induction medication input variables.

7. The method of any of Claims 1-6, wherein the set of peri -induction medication input variables comprises a predetermined set of anesthetic agents.

8. The method of any of Claims 1-7, wherein the predetermined set of anesthetic agents is selected from the group consisting of propofol, etomidate, ketamine, dexmedetomidine, nitrous oxide (NO), halothane, sevoflurane, desflurane, isoflurane, midazolam, diazepam, lorazepam, alfentanil, sufentanil, remifentanil, fentanyl, atracurium, cisatracurium, pancuronium, vecuronium, rocuronium, succinylcholine, hydromorphone, hydrocodone, oxycodone, phenylephrine, and lidocaine.

9. The method of any of Claims 1-8, wherein executing the one or more encoding processes on the one or more input variables comprises executing an imputation of one or more missing values associated with the one or more input variables.

10. The method of any of Claims 1-9, wherein executing the one or more encoding processes on the one or more input variables comprises executing a one-hot encoding of the one or more input variables.

11. The method of any of Claims 1-10, wherein calculating the one or more cross- validation losses further comprises: evaluating a cross-validation loss function based on the preprocessed set of preoperative input variables and anesthetic induction input variables, the predetermined set of target MBP values, the set of hyperparameters, and a set of learnable parameters associated with the machine-learning model; and minimizing the cross-validation loss function by varying the set of learnable parameters while the preprocessed set of preoperative input variables and anesthetic induction input variables, the predetermined set of target MBP values, the set of hyperparameters remain constant.

12. The method of any of Claims 1-11, wherein minimizing the cross-validation loss function comprises optimizing the set of hyperparameters, and wherein the set of hyperparameters comprises one or more of a set of general parameters, a set of booster parameters, or a set of learning-task parameters.

13. The method of any of Claims 1-12, wherein minimizing the cross-validation loss function comprises minimizing a loss between a prediction of one or more MBP values post induction and associated with PIH in the patient and an experimentally-determined one or more MBP values post induction and associated with PIH in the patient.

14. The method of any of Claims 1-13, wherein the set of learnable parameters comprises one or more weights or decision variables determined by the machine learning model based at least in part on the preprocessed set of preoperative input variables and anesthetic induction input variables and the predetermined set of target MBP values.

15. The method of any of Claims 1-14, wherein the updated set of hyperparameters comprises one or more of an updated set of general parameters, an updated set of booster parameters, or an updated set of learning-task parameters.

16. The method of any of Claims 1-15, wherein the machine-learning model comprises one or more of a gradient boosting model, an adaptive boosting (AdaBoost) model, an extreme gradient boosting (XGBoost) model, a light gradient boosted machine (LightGBM) model, or a categorical boosting (CatBoost) model.

17. The method of any of Claims 1-16, wherein the machine learning model comprises one or more of a linear regression model, a ridge regression model, a Bayesian ridge regression model, a tree-based model, or a neural network.

18. The method of any Claims 1-17, further comprising: prior to accessing the data set comprising demographics and clinical data, splitting the data set into a training data set, a test data set, and a validation data set for training and executing the machine-learning model.

19. The method of any of Claims 1-18, further comprising: executing a reinforcement-learning model on the prediction of the MBP post induction to generate one or more candidate therapies, wherein the one or more candidate therapies comprise one or more of adjusting a dosage of an anesthetic agent, adding an anesthetic agent, ceasing the administration of an anesthetic agent, administrating a medication, or administrating a fluid.

20. The method of any of Claims 1-19, further comprising: automatically applying one or more of the candidate therapies to the patient.

21. The method of any of Claims 1-20, further comprising: receiving a MBP measurement of the patient at a first timepoint post induction, wherein outputting the prediction of the MBP post induction further comprises inputting the MBP measurement of the patient at the first timepoint post induction to the machinelearning model and outputting the prediction of the MBP at a second timepoint subsequent to the first timepoint, and wherein the prediction of the MBP at the second timepoint is generated based on the MBP measurement of the patient at the first timepoint post induction.

22. The method of any of Claims 1-21, further comprising: generating a report based on the prediction of the MBP post induction and associated with PIH in the patient and the one or more candidate therapies; and transmitting the report to a computing device associated with a clinician.

23. The method of any of Claims 1-22, further comprising causing, based on the prediction of the MBP post induction and associated with PIH in the patient and the one or more candidate therapies, a user interface (UI) executing on the computing device to display a visual representation of the prediction of the MBP post induction and associated with PIH in the patient and the one or more candidate therapies.

24. A system including one or more computing devices for predicting a mean blood pressure (MBP) post induction and associated with post-induction hypotension (PIH) in a patient, comprising: one or more non-transitory computer-readable storage media including instructions; and one or more processors coupled to the one or more storage media, the one or more processors configured to execute the instructions to: access a data set comprising demographics and clinical data associated with the patient, wherein the demographics and clinical data comprise a set of preoperative input variables and anesthetic induction input variables associated with the patient; and refine a set of hyperparameters associated with a machine-learning model trained to generate a prediction of a mean blood pressure (MBP) post induction and associated with post-induction hypotension (PIH) in the patient, wherein refining the set of hyperparameters comprises iteratively executing a process until a desired precision is reached, the process, comprising: preprocessing the data set by executing one or more encoding processes on each input variable of the set of preoperative input variables and anesthetic induction input variables; calculating one or more cross-validation losses based at least in part on the preprocessed set of preoperative input variables and anesthetic induction input variables and a predetermined set of target MBP values; updating the set of hyperparameters based on the one or more cross- validation losses; and outputting, by the machine-learning model, the prediction of the MBP post induction and associated with PIH in the patient based at least in part on the updated set of hyperparameters.

25. A non-transitory computer-readable medium comprising instructions that, when executed by one or more processors of one or more computing devices, cause the one or more processors to: access a data set comprising demographics and clinical data associated with the patient, wherein the demographics and clinical data comprise a set of preoperative input variables and anesthetic induction input variables associated with the patient; and refine a set of hyperparameters associated with a machine-learning model trained to generate a prediction of a mean blood pressure (MBP) post induction and associated with post-induction hypotension (PIH) in the patient, wherein refining the set of hyperparameters comprises iteratively executing a process until a desired precision is reached, the process, comprising: preprocessing the data set by executing one or more encoding processes on each input variable of the set of preoperative input variables and anesthetic induction input variables; calculating one or more cross-validation losses based at least in part on the preprocessed set of preoperative input variables and anesthetic induction input variables and a predetermined set of target MBP values; updating the set of hyperparameters based on the one or more cross- validation losses; and outputting, by the machine-learning model, the prediction of the MBP post induction and associated with PIH in the patient based at least in part on the updated set of hyperparameters.