Artificial intelligence system that includes expertise related to hypertension treatment

Abstract

A process is provided that includes obtaining test data quantifying a patient's cardiovascular condition and outputting a recommended sequence for administering medications.

Description

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This PCT application claims the benefit of U.S. Provisional Patent Application No. 62 / 888,928, filed on August 19, 2019, entitled “ARTIFICIAL INTELLIGENCE SYSTEMS THAT INCORPORATE EXPERT KNOWLEDGE RELATED TO HYPERTENSION TREATMENTS,” and is a continuation of U.S. Patent Application No. 16 / 546,156, filed on August 20, 2019, entitled “ARTIFICIAL INTELLIGENCE SYSTEMS THAT INCORPORATE EXPERT KNOWLEDGE RELATED TO HYPERTENSION TREATMENTS.” The entire contents of each of the foregoing patent applications are incorporated herein by reference for all purposes. Technical Field

[0003] The present disclosure relates generally to artificial intelligence (AI), and more particularly to AI systems that incorporate expert knowledge related to the management and care of hypertensive patients while retaining the ability to generalize to provide appropriate responses to novel inputs. Background Art

[0004] Physicians and other healthcare practitioners face tremendous complexity when treating patients. A large part of this complexity involves choosing medications that will best help a particular patient and a large number of variables specific to that patient. Often, doctors must make decisions under uncertainty based on relatively noisy, high-dimensional data about the patient, and these signals may evolve over time and in some cases be difficult to predict.

[0005] Medical practitioners are trained and taught to understand and apply the vast amount of medical research to select appropriate interventions for their patients. Currently, approximately 800,000 new medical journal articles are published each year. To obtain approval from the Food and Drug Administration (FDA), each drug must be tested in a carefully constructed and executed study under FDA oversight and approval. Subsequent studies of the same drug may be conducted under independent auspices or sponsored by the drug manufacturer. Each study has a specific set of included and excluded variables. Different studies of the same drug can, and often do, have different included and excluded variables and values. When a physician prescribes a drug, they are unlikely to recall from memory which studies investigated that condition or drug, and are extremely unlikely to recall any of the included and excluded variables. Therefore, determining which drug to use is extremely difficult.

[0006] To assist those working in the field, panels of experts are sometimes convened under the authority of various national or international authorities, such as the American College of Cardiology, the American Heart Association, the Institute of Medicine, etc., to comb through the various studies related to a particular disease condition and update the available science about that disease condition to produce high-level findings. After publishing hundreds of thousands of pages of scientific papers, for example, an agency may establish a standard of care, which is sometimes referred to as an "evidence-based standard of care" (EBS). In some areas, the EBS or other standards will at a minimum define a criterion, such as a treatment goal and a general review of the various tools that can be used to achieve that goal, by which successful treatment or therapy can be measured. However, these standards may be difficult to implement, outdated, or apply less granular heuristics that favor manageability over precision.

[0007] Given these challenges, only recently have developers and computer science researchers begun trying to help doctors choose the right treatment for their patients, and so far many of these attempts have been limited to diseases such as cancer. However, these attempts have generally not been successful. Some approaches have attempted to apply expert systems that encode the entire medical knowledge into a set of rules. The problem with rules is that the number of variables combined with the ways in which the variables can interact and the number of solution options can approach tens of millions of permutations, which is beyond the power of rule formulation. In general, expert systems have proven to be too brittle to generalize outside of scenarios explicitly considered by the system architects. Such systems often struggle with patients presenting novel scenarios. On the other hand, some researchers have attempted to train models with a relatively large number of degrees of freedom, such as deep neural networks, based on historical records of treatments and patient responses. However, these systems generally fail to benefit from the knowledge generated by medical research and in many cases have relatively low training efficiency. These models struggle with small training sets and training sets where the samples are sparse in areas where the model may later be asked to perform. Another problem is that medical data is often very noisy and full of errors. Summary of the invention

[0008] The following is a non-exhaustive list of some aspects of the present technology. These and other aspects are described in the following disclosure.

[0009] Some aspects include a process comprising: obtaining, by a computer system, data quantifying, for example, a patient's cardiovascular function status based on one or more tests, the test data specifying patient attributes in multiple cardiac, vascular, and cardiovascular dimensions of the patient, a first subset of the multiple cardiovascular dimensions being independent cardiovascular dimensions, and a second subset of the multiple cardiovascular dimensions being non-independent cardiovascular dimensions; determining, by the computer system, multiple normalized differences between the test data and a target standard in each of the cardiovascular dimensions, the multiple normalized differences quantifying different aspects of the patient's hypertension; determining, by the computer system, a predicted impact vector for each of a plurality of different classes of drugs having hypertension as an indication, wherein: the predicted impact vectors each correspond to a different corresponding class of drugs in the different classes of drugs, and the predicted The impact vectors each have multiple values, which quantify the corresponding impact of the corresponding category of drugs on the corresponding dimensions in the third set of the patient's cardiovascular dimensions, each of which is based on the corresponding normalized difference on the corresponding dimension in the third set and the corresponding strength of the impact of the corresponding category of drugs on the corresponding dimension, the corresponding strength impact is determined based on the respective multidimensional models of the corresponding category of drugs, and the third set overlaps with at least the first subset and the second subset; through the computer system, based on the values ​​of the corresponding predicted impact vectors of the corresponding category of drugs, the total score of each corresponding category of drugs in the different categories of drugs is determined; through the computer system, the different categories of drugs are ranked based on the total scores to form a ranked list of different categories of drugs; and through the computer system, based on the ranking, a recommended sequence for administering the category of drugs to the patient is output.

[0010] Some aspects include a tangible, non-transitory machine-readable medium storing instructions that, when executed by a data processing apparatus, cause the data processing apparatus to perform operations including those described according to the above-described procedures.

[0011] Some aspects include a system comprising: one or more processors; and a memory storing instructions that, when executed by the processors, cause the processors to implement operations described according to the above process. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] The above and other aspects of the present technology will be better understood when reading this application with reference to the following drawings, in which like reference numerals indicate similar or identical elements:

[0013] Figure 1 is a logical architecture block diagram of a medical recommendation AI application based on knowledge representation according to some embodiments of the present technology;

[0014] Figure 2 According to some embodiments of the present technology, Figure 1A flowchart of the process of recommending medical interventions for hypertension implemented by the logical architecture of

[0015] Figure 3 Some embodiments of the present technology may be performed on Figure 1 Application and Figure 2 A block diagram of an example of a computing device that performs a process.

[0016] Although the present technology is susceptible to various modifications and alternative forms, specific embodiments thereof are shown by way of example in the drawings and will be described in detail herein. The drawings may not be drawn to scale. However, it should be understood that the drawings and their detailed description are not intended to limit the present technology to the specific forms disclosed, but on the contrary, are intended to cover all modifications, equivalent forms and alternative forms falling within the spirit and scope of the present technology as defined by the appended claims. DETAILED DESCRIPTION

[0017] In order to alleviate the problems described herein, the inventors must both invent solutions and, in some cases, equally importantly, recognize problems in the field of artificial intelligence that have been overlooked (or not yet foreseen) by others. Indeed, the inventors wish to emphasize that if industry trends continue as the inventors anticipate, the difficulty of those problems that are new and will become more apparent in the future should be recognized. In addition, because multiple problems are solved, it should be understood that some embodiments are problem-specific and that not all embodiments solve every problem of the conventional systems described herein or provide every benefit described herein. That is, improvements that solve various arrangements of these problems are described below.

[0018] Some embodiments mitigate some of the above-described problems through the application of artificial intelligence that encodes knowledge from medical research in a collection of models responsive to patient conditions in a relatively large envelope of a relatively high-dimensional patient condition space. As a result, these models are expected to be less brittle than rules in expert systems, while still capturing medical knowledge from relevant literature in the field. In addition, the efficiency of training data for these models is expected to be greater than the naive application of machine learning models that do not benefit from expert knowledge. None of this is a waiver of the right to use the present technology in conjunction with expert systems or such machine learning models, or of any other subject matter discussed herein.

[0019] The present technology is described with reference to recommended interventions for hypertension, but it should be emphasized that the present computational technology is expected to be applicable to a wide range of areas with similar characteristics to the use case discussed. Just 30 years ago, there were few medications to treat hypertension, and many patients died or suffered serious complications (e.g., stroke, heart attack, transient ischemic attack (TIA), kidney failure, etc.) due to uncontrolled blood pressure. In those days, physicians would try several available medications to lower blood pressure. However, blood pressure was often uncontrollable.

[0020] In the years that followed, a great deal of research has been done on hypertension, as well as most other common chronic diseases, leading to a deeper and more sophisticated understanding of these diseases, their causes, and the various factors or variables associated with them. The additional knowledge, in turn, has led to the development of more therapies. In 2019, there are more than a dozen drug classes that can be used to treat hypertension, and within these classes, there are often multiple different drugs in each class, each of which is slightly different from the other drugs in the same class.

[0021] As described above, the EBS or other criteria may define a criterion, such as a treatment goal and general comments about the various tools that may be used to achieve that goal, by which successful treatment or therapy may be measured. For example, in hypertension, a blood pressure of ≤139 / 89 (also expressed as <140 / 90) is considered controlled, while a blood pressure of ≥140 / 90 is considered "uncontrolled" and therefore considered more dangerous to the patient by several authorities. Other authorities have persuasively advocated blood pressure as low as <130 / 80, so there is no universal consensus in some cases.

[0022] The Centers for Disease Control (CDC) states that a goal of <140 / 90 is the minimum standard for controlling high blood pressure. The Centers for Medicare and Medicaid (CMS) requires providers to report the percentage of all their adult patients younger than 85 years of age who are diagnosed with high blood pressure less than 140 / 90, which is a measure of the physician's success as a hypertension therapist. The CDC also reports that only about half of American hypertensive patients achieve a safe blood pressure level of <140 / 90.

[0023] Blood pressure readings typically include both systolic and diastolic blood pressure (i.e., SBP / DBP). These are high-order measures. Systolic and diastolic blood pressure are the result of a series of "cause and effect" relationships. Not all of these relationships are fully understood or known by medicine, but many are. For example, blood pressure depends in part on various performance characteristics of the heart, blood vessels, and fluid status. Heart function, blood vessel function, or the patient's fluid status are in turn influenced or determined by many neurohormonal, biochemical, mechanical, age, ethnic, genetic, and other variables.

[0024] Treating a patient in a way that achieves a desired endpoint blood pressure is acceptable, but it is unacceptable for other cardiac variables, vascular variables, or humoral variables to be out of range. In this case, although the patient’s blood pressure is “good” (e.g., less than 140 / 90), their care is suboptimal because, while the overall blood pressure meets the EBS of care, the underlying parameters are not ideal. This occurs, for example, when one variable is overtreated and another is undertreated resulting in an “average” blood pressure at an acceptable level. In these cases, despite the blood pressure being within the desired range, there may be ongoing harm to the patient’s overall health status, either now or in the future.

[0025] Another set of variables is that a particular hypertensive patient may have any one of a large number of comorbidities that may, alone or in combination, determine which therapies may or may not be used effectively or safely. Comorbidities have a variety of potential effects on treatment strategies, including:

[0026] a. Comorbidities may facilitate the use of certain medications or therapies;

[0027] b. Comorbidities can alter the therapeutic properties of an agent or therapy, making it more or less effective;

[0028] c. Comorbidities may eliminate the possibility of using certain therapies;

[0029] d. Comorbidities may change the priority or order of treatment;

[0030] e. Comorbidities may lead to potential problems that must then be assessed when adding therapy; and

[0031] f. Comorbidities can alter or completely change the treatment endpoint of the disease.

[0032] In some cases, the effect of one comorbidity points treatment in one direction, while the effect of a second comorbidity points treatment in the opposite direction, and this conflict must be resolved based on some assessment of the weight of the various factors.

[0033] The development of scientific knowledge has produced an appreciation of all of these patient attributes that drive blood pressure, which has led to a proliferation of therapies that include hundreds of medications that exist today but did not exist 30 or fewer years ago.

[0034] Thus, treating hypertension involves the aforementioned complexities. It is worth remembering that as of 2019, over 800,000 scientific medical articles are published each year. This is a lot of information for a human actor to absorb. The role of the physician or other licensed care provider in blood pressure treatment is to calculate all or as many of the above variables as possible in order to determine the effective treatment as well as the best treatment for a particular patient. When all cardiovascular, demographic, pharmacological, and comorbid conditions are considered, there are literally tens of millions of permutations involved. This complexity is more than even a very smart person can manage in her mind. It is believed that this is the underlying reason why the success rate for hypertension treatment is as low as 50% on average. What usually happens is that patients with the most common forms of hypertension without many complications or comorbidities are most likely to be successfully treated, while more complex cases are often not treated successfully.

[0035] To manage this level of extreme complexity, some embodiments are described below with reference to Figure 3 The description is implemented on one or more computing devices as follows Figure 1 and Figure 2 AI systems described. Some embodiments obtain a "starting point" for each specific hypertensive patient - e.g., all medically available variables that influence the patient's current blood pressure condition - and then determine a (typically multi-step) route to a desired outcome (e.g., endpoint blood pressure, adjusting for other cardiac, vascular, and fluid parameters, and accounting for all comorbidities) via optimal drug therapy.

[0036] Embodiments determine this route in a manner different from previous applications of AI to recommend medical interventions. A rules engine alone is considered unsuitable for this task. The variety of permutations of patient conditions is simply too great, making expert systems brittle to new types of inputs that are expected to be commonly encountered. Furthermore, a commercially viable rules engine would be unable to account for the various combinations or interactive effects that would also occur.

[0037] The embodiments determine this route through what is believed to be a fundamentally different approach from how a human would perform the analysis. General AI is not available, and there are no written algorithms that are universally implemented by physicians. It is believed that the way humans make these decisions is based on some combination of Bayesian analysis and crude heuristics as a shortcut to interpret the available knowledge in the field. As a matter of experience, the results are suboptimal and inferior to those obtained using the present technology. Therefore, the present technology is necessarily different, in part because its performance is higher, in part because it is more comprehensive, and in part because the knowledge in the field is systematized by the model to make that knowledge available to the AI ​​system across a wide range of high-dimensional novel (in the sense that conditions have not been previously input to the system or are not part of the training set) patient condition inputs.

[0038] In some embodiments, physiological conditions including cardiac function, vascular function, vascular compliance, and fluid levels can be determined. Many of these variables can be evaluated and measured as functions of each other. Therefore, some variables are not absolute measurements that are universal to all patients, and when all medically useful variables are evaluated, there is a combinatorial explosion of possible patient conditions.

[0039] Some embodiments implement a cardiovascular model to define cardiac, vascular, and fluid forces that relate to one another, and as a result, through such a model, measured patient attributes can be converted to other estimated or otherwise inferred patient attributes based on the interaction of measured patient attributes, patient demographic attributes, and comorbidities in some cases.

[0040] Some embodiments implement a drug efficacy model that quantifies the effectiveness and other performance characteristics of each drug or drug combination according to each of the various patterns generated by the cardiovascular model generated in the first step.

[0041] Some embodiments implement a comorbidity model to quantify the impact of over 25 comorbidities, as well as other factors that may alter recommendations in various ways.

[0042] Some of these models may implement one or more performance curves to quantify the impact of a particular given variable, as well as interactive performance curves that quantify the impact of one or more variables collectively or one-to-one.

[0043] In some embodiments, the computer system uses test data, such as diagnostic test data, to quantify the patient's cardiovascular condition. The data can be plotted according to multiple cardiovascular measurement scales. In many cases, these measurement scales can be mathematically related to each other. The subsets can be independent. In some embodiments, four independent variables are the only inputs to the drug class score: systolic blood pressure (S), pulse pressure (p), cardiac index (x), and heart rate (z).

[0044] The relative distance between the patient data (or other attribute sets that collectively represent the patient's condition) and multiple target standards can be calculated to quantify different aspects of hypertension. The various standards can be heterogeneous units of measurement and have significantly different scales. That is, one metric value is usually between 0.25 and 0.95, while another metric value is between 1500 and 6000. Scale rationalization (or other forms of normalization) techniques can be used to quantify the relative comparative magnitude of the distance, resulting in all metrics being on a shared scale (e.g., -3.0 to +3.0). In some embodiments, the shared scale can be the number of standard deviations relative to the target, such as the population mean based on which the standard deviation is determined, such as a z score. A negative distance represents being lower than or less than the target, while a positive distance represents being higher than or greater than the target.

[0045] The test data can be processed to provide a patient's condition that indicates the most important aspects of their hypertension and which variables most need to be moved and in which direction (increase / decrease).

[0046] In some embodiments, a drug efficacy model can be constructed to accommodate the following situation: different agents or classes affect different cardiovascular parameters. Literature analysis and consultation with pharmacology experts on the best way to explain the differences can be used to construct a weighted scoring system that can be applied to reflect the efficacy of drug classes on specific parameters. Example weights can be stored in a matrix where columns correspond to drug classes and rows correspond to various cardiovascular parameters such as V, C, z, and Si. Different drug classes can have different weights for different cardiovascular parameters. An example of such a matrix is ​​presented in the following table:

[0047]

[0048] The resulting score for a particular hypertension drug class can be a quantification of the cumulative net beneficial effect, which can be calculated in a variety of ways, such as a weighted sum of the magnitude of the effect multiplied by the distance of each cardiovascular parameter. This approach can be used to score various hypertension drug classes in a nearly orthogonal array of possible patient data. With this approach, in some embodiments, a model can be implemented for each drug class. In some cases, the model can correspond to a non-planar surface in four or five or more dimensions, with one output dimension and several input dimensions corresponding to patient attributes.

[0049] Through these models, some embodiments determine an overall score for each drug category for a given patient given a unique set of independent inputs. These scores can then be compared to each other to determine the best hypertension drug treatment regimen.

[0050] Some embodiments may apply two additional processes to generate a "ladder" of drug classes. The concept of the ladder is based on the evidence that more than one drug class is often needed to effectively control hypertension. In addition, often the patient is already taking the recommended drug class. The provider will then decide to increase the dose of the highest recommended drug class or add the next highest recommended drug class. In some cases, the ladder can be represented as a sequence that can specify one or more stages, where, in each stage, one or more drug classes will be administered to the patient.

[0051] Most hypertension drug classes can be safely taken together. However, some combinations of drug classes should not be taken together. For example, drug classes Cd and Cn should not be taken together; similarly, drug classes B1 and B3 should not be taken together. The result list can be filtered to skip drug classes that should not be taken in combination. The higher-scoring drug class can be the only one of a pair of conflicting drug classes allowed to remain on the ladder.

[0052] In some embodiments, the cardiovascular score can be modified to take into account the presence of certain comorbidities or demographic information of the hypertensive patient. Certain hypertension drug classes are contraindicated for specific medical conditions. For example, a given drug class A may contraindicate pregnancy. In other cases, the drug class may have a very beneficial effect on a given comorbidity. For example, the A drug class may have a positive impact when renal function or diabetes is present. In response to input indicating potential comorbidities or related attributes, the drug class score of the corresponding drug class can be adjusted accordingly. In some cases, there may be multiple clinical situations, and all situations make additional adjustments to the final drug class score and its associated ranking in the ladder.

[0053] In some embodiments, the above-described techniques and related technologies may be used in Figure 1 The knowledge representation medical recommendation AI application 10 can be implemented in the following reference Figure 3 The application 10 may be executed on one of the computing devices described herein, such as in its operating system, or on a collection of such computing devices, such as where the various components are implemented as services in a service-oriented architecture. Alternatively, in some cases, the application 10 may be executed as a monolithic application on a single computing device, such as running as a single process on a single processor core.

[0054] In some embodiments, the application 10 may be configured to receive input from the physician 12 regarding the patient 14 via the user input 16. In some embodiments, the application 10 may be configured to present information in response to the input on a display 18, such as a graphical user interface on a monitor, head mounted display, etc., or via an audio display, via speech to text conversion output as audio transmitted via a speaker. Or in some embodiments, the display 18 is implemented with a printer that prints a printed page with the output.

[0055] Input 16 may include a keyboard, a microphone coupled to a speech-to-text converter, a touch screen, a mouse, or an application program interface of another application that provides a collection of inputs to application 10, such as a patient records database that is configured to export patient medical records to application 10 in response to a query from application 10 or other instructions from a medical practitioner.

[0056] In some embodiments, application 10 may be executed on a computing device in a doctor's office or hospital and may receive and provide input and output without transmitting the information over a network for security purposes, or in some cases the application may be hosted remotely and the information may be transmitted encrypted during transmission.

[0057] In some embodiments, the application includes a test data ingestion module 20 configured to receive a patient record indicating a medical condition of the patient at a given point in time. In some embodiments, the ingestion module 20 may be configured to receive a plurality of such records corresponding to a patient condition history, each record having a timestamp indicating a time when the patient exhibited the condition. In some embodiments, the records may include a plurality of attributes of the patient related to hypertension in various cardiovascular dimensions. Some of the dimensions may be independent dimensions, such as those listed above, and some of the dimensions may be dependent on dimensions that are at least partially caused by the independent dimensions and, in some cases, by their interactions. Examples of such dimensions include:

[0058] S = Systolic blood pressure (SBP)

[0059] D = Diastolic blood pressure (DBP)

[0060] p = pulse pressure (SBP-DBP)

[0061] y = mean arterial pressure (MAP = DBP + ?p)

[0062] x = cardiac index (CI)

[0063] z=heart rate (HR / pulse{bpm})

[0064] V = Total Peripheral Resistance Index (TPRI = 80*MAP / CI)

[0065] C = cardiac power index (CPI = 0.0022*MAP*CI)

[0066] Si = stroke index (SI = 1000*CI / HR)

[0067] In some embodiments, the patient's attributes and each of these dimensions may be entered into a user interface presented on display 18, such as by a medical technician into fields of a form after the various attributes of the patient have been measured. Alternatively, in some cases, the attributes may be received from another application via an application program interface, such as in a hierarchical data serialization format.

[0068] In some cases, based on a patient model such as the patient model discussed above, the set of received attributes can be transformed into a larger set of inferred patient attributes. For example, some such models can model patients with a causal graph, where input nodes correspond to observable attributes and edges and downstream causal nodes specify transformations by which, in some cases, other patient attributes can be inferred based on the interaction of observed or calculated patient attributes.

[0069] Some embodiments may validate the patient record via a validator 22. In some embodiments, this may include determining that all required attributes (i.e., values ​​on a dimension) are present, that a received attribute is greater than a minimum possible value, and that a received attribute is less than a maximum possible value on the corresponding dimension. Upon determining that a given patient attribute is outside of one of these ranges, some embodiments may issue an error message and present an input via display 18, inviting a medical professional to confirm that the value is correct or to change the input via input 16. Some embodiments may determine whether an attribute in a patient record differs by more than a threshold amount from a previous attribute on the same dimension in a previous record, where an outlier difference potentially indicates an erroneous input. Likewise, some embodiments may signal and issue an alarm and invite correction in response to detecting such an event.

[0070] In some embodiments, after verifying all patient record attributes, the results can be advanced to the normalizer 24 in the pipeline shown in the application 10. In some embodiments, the normalizer 24 can calculate the normalized difference between each patient attribute and the target value of the attribute on the attribute dimension. In some embodiments, these differences can be quantified relative to the demographics of the corresponding population attributes reflected in the population model 26. For example, the normalized attribute can be expressed as a z score, which indicates the standard deviation amount above or below the mean value of the attribute of the population encoded in the population model 26. Some embodiments can calculate similar statistics for other types of non-Gaussian distributions. Some embodiments can calculate the percentage of the population above the patient attribute measured in a given dimension and subtract 50 percentage points to center the result around zero. In some embodiments, the normalized value of each attribute can have a negative value indicating that the patient attribute is lower than the expected value and a positive value indicating that the patient attribute is higher than the expected value. In some embodiments, the normalized values ​​can all be on the same scale, although the attributes in the patient record have been measured on different scales. In some cases, these values ​​may be normalized after subsequent transformations described below, for example, the category model 28 described below may be calibrated to non-normalized values ​​and its output may be normalized to facilitate the combination of values ​​across various attributes with heterogeneous unit measurements and ranges.

[0071] In some embodiments, the population model 26 can characterize the demographics of the attributes in each of the dimensions of the patient record. The characterization can take a variety of forms, including histograms, measures of central tendency (such as mean, median, and mode), and measures of variation (such as standard deviation or variance). In some embodiments, the measure of central tendency can be a target value for the corresponding dimension of the patient's cardiovascular health. In some embodiments, the population can be a known healthy population or a reference population. In some cases, the target value can be determined based on the EBS described above.

[0072] The output of the normalizer 24 may be a normalized patient record having a plurality of normalized attributes in a plurality of different cardiovascular dimensions. For example, the normalizer 24 may output more than four, more than six, or more than eight normalized values ​​ranging from -3 to +3 and each corresponding to a different cardiovascular dimension of the patient.

[0073] The output of the normalizer 24 may be provided to a predictor 26, which may predict the magnitude and direction of the effects of various classes of drugs, which indicate the effects of hypertension on corresponding attributes in corresponding dimensions of the patient's cardiovascular health. For example, the predictor may predict that a first class of drugs decreases a first normalized attribute by a first amount on a normalized scale and increases a second normalized attribute by a second amount on a normalized scale, while a second class of drugs may be predicted to decrease the first normalized attribute by a third amount on a normalized scale, have no effect on the second normalized attribute, and decrease the third normalized attribute by a fourth amount on a normalized scale.

[0074] In some embodiments, the predictor 26 can determine a set of predictions indicative of hypertension for each of a plurality of different classes of medications. For example, nine different predictions for nine different cardiovascular dimensions can be determined for a first class of medications, and nine different predictions for the same nine different cardiovascular dimensions can be determined for a second class of medications. In some embodiments, each set of predictions can be characterized as a prediction influence vector that can specify a location in a continuous vector space or a discrete vector space corresponding to a patient's cardiovascular dimension. Program states do not need to be marked as vectors in the program code to constitute a vector, as long as the information of the vector is encoded. In some embodiments, the dimensions for which predictions are made can be the same as the dimensions of the data obtained by the ingestion module 20, or in some cases, predictions can be made only for a subset of these dimensions, such as only for independent cardiovascular dimensions.

[0075] In some embodiments, each predicted impact vector may have a plurality of values ​​corresponding to the predicted impact on each of the different cardiovascular dimensions of the patient. In some embodiments, these values ​​may indicate whether the predicted impact tends to bring the corresponding attribute closer to the target on the cardiovascular dimension or further away from the corresponding target. For example, an attribute that is below the target but the predicted impact tends to increase the attribute to be closer to the target may have a positive value, while an attribute that is above the target and the predicted impact tends to increase the attribute may have a negative value. In some embodiments, the resulting value may be the difference between the absolute value of the attribute in the normalized record and the absolute value of the attribute predicted to occur when the class of drugs in question is applied. An attribute that is farther away from his target becomes closer to his target may tend to correspond to a larger value, and vice versa.

[0076] In some embodiments, the predictor 26 calculates the predicted impact vector based on a corresponding one of the class models 28. In some embodiments, each class model 28 corresponds to a different class of drugs, or in some cases, the interactions between class drugs may also have their own class models. In some embodiments, each class model encodes a three-dimensional, four-dimensional, five-dimensional, six-dimensional, seven-dimensional or higher dimensional surface, thereby indicating the impact of the corresponding class of drugs on each dimension in the cardiac dimension, for which the prediction is made based on the normalized condition of the patient received from the normalizer 24. In some cases, these high-dimensional models can be decomposed into low-dimensional slices, such as multiple different three-dimensional or four-dimensional representations, which together still constitute a higher-dimensional model, regardless of whether a lower-dimensional encoding is performed. In some cases, the model has an output dimension for each of the above-mentioned cardiovascular dimensions and an input dimension for each of the above-mentioned cardiovascular dimensions. The surface can indicate how the input is mapped to the output for each drug class. Models can be formed by synthesizing medical studies, for example, by averaging empirical data across studies, training Markov chain Monte Carlo models with empirical results from these studies, interpolating between areas examined in different studies based on applying results from one study to a model proven by another study, etc. Models 28 can take many forms. Examples include polynomial functions or collections thereof. Other examples include lookup tables, support vector machines, neural networks, and other lossy or lossless encodings of the relationship between inputs and outputs.

[0077] In some cases, a predictive impact matrix may be calculated for each class of drugs, with a specific drug predictive impact vector for the drug within that class. Model 28 may be developed for different members of a class. In some cases, the highest scoring drug in a class (based on the following analysis) may be selected as the predictive impact vector for that class, and that specific member of the class may be recommended.

[0078] In some embodiments, the predicted impact vector in each different drug category can be output to a category scorer 30, which can calculate a total score (e.g., a single ordinal value or cardinality value) for each drug category based on some or all values ​​in the corresponding predicted impact vector for that drug category. Or in some embodiments, scores can be calculated for multiple different metrics, or some embodiments can calculate a single total score for each category, which in some cases can be the sum of the calculated values ​​of the scalar used as the predicted impact vector, or in some cases, the total score can be a weighted sum of these values. The weights can be determined by various techniques, including hand-coded weights based on the knowledge of experts in the field or empirical literature in the field. The category scorer 30 can output a set of total scores, for example, one total score for each of the different categories of drugs. The total score can indicate the preliminary suitability of the corresponding category of drugs for the patient, where a larger score indicates a higher suitability, and a lower score indicates a lower suitability. (References to the signs of fractions and other values ​​should be understood to refer also to the semantic equivalent of inversion of the signs, e.g., discussion of a fraction in which a higher value indicates a desired outcome should be understood to be equivalent to discussion of the negated version of that fraction, or the reciprocal of that fraction in which a lower value indicates a desired outcome).

[0079] The collection of scores for classes of drugs can be received by a comorbidity filter 32. In some embodiments, the scores can be adjusted based on the patient's various comorbidities. For example, certain classes of drugs may be documented in the literature as having negative effects or being excluded for use in patients with particular comorbidities, or certain classes of drugs may be more appropriate for certain comorbidities. In some embodiments, the relationship can be encoded as an adjustment factor, such as a weight added to the sum of the scores or multiplied by the scores. For example, a pregnant patient may have a particular class of drug score multiplied by 0.8 to indicate that the class of drugs may have negative effects, while another class of drug score may be multiplied by 0 to indicate that pregnant women are prohibited from taking the class of drugs. The term filter is used to refer to this adjustment process, which in some cases may take the form of removing candidate classes of drugs, or in other cases may take the form of converting the scores based on which these classes of drugs are selected.

[0080] Some embodiments may similarly apply a demographic filter 34 which may apply similar adjustments based on demographic attributes of the patient, such as race, gender, age, etc.

[0081] Some embodiments may apply compatibility filters 36 to class drugs. Some embodiments may maintain a square matrix or other data structure that encodes pairwise comparisons between candidate class drugs, where the matrix values ​​indicate whether different pairs of class drugs indexed to these entries are compatible with each other. Alternatively, when pairs of class drugs are used together (e.g., simultaneously, sequentially, or as part of the same sequence even if not sequentially, and some embodiments may apply different compatibility matrices to each of these different types of relationships between drug classes during treatment, for example, to identify those that can be grouped for simultaneous use or should be taken sequentially), some embodiments may encode weights or sums applied as adjustments to the above scores. Some embodiments may detect incompatible pairs of drugs and remove lower-ranked members of the pair from the candidate set.

[0082] Some embodiments may implement a ranker 38 that can rank the adjusted results or otherwise filtered category drugs based on the scores converted by components 30 to 36. In some embodiments, the operations implemented by these components may be performed in an order different from the order depicted in the illustrated pipeline, which does not mean that any other features are limited to the illustrated arrangement. In some cases, the ranking may include grouping category drugs to be taken at the same time, and in some cases, it may include one or more groups to be taken consecutively according to the ranking instructions, the ranking starting, for example, from the highest ranked category and proceeding downward.

[0083] Some embodiments may then transmit the resulting ordered list from the ranker 38 to a user interface generator 40, which may generate a user interface that displays the results via any of the types of displays described above on the display 18. In some embodiments, in the user interface, various classes of drugs may be further annotated with data regarding a particular vulnerability or patient's suitability for that class of drugs. For example, where a comorbidity causes the total score for a class of drugs to be adjusted, a prose description of the adjustment and rationale may be retrieved from memory in response and presented in association with that class of drugs in the user interface. Similarly, where a class of drugs is prohibited or incompatible, a prose description of a warning may be retrieved from memory in response to detecting that such an event has occurred and presented in the user interface by the user interface generator 40.

[0084] In some embodiments, Figure 1 The application can execute Figure 2The process 50 shown in the figure is used to recommend an intervention process for hypertension, and although the embodiment of the process 50 is not limited to this embodiment, this does not mean that any other features described herein are limiting in all embodiments. In some embodiments, in addition to managing treatment, the functions of the process 50 and other functions described herein can be implemented as computer code stored in a tangible, non-transitory machine-readable medium and executable on one or more processors, so that when the computer code is executed, the corresponding functions are implemented. In some embodiments, the operations described can be performed in a different order, additional steps can be inserted, certain steps can be performed simultaneously, certain steps can be omitted, certain steps can be repeated multiple times before proceeding to other steps, and the process can be changed in other ways, which also does not mean that any other description in this article is limiting.

[0085] Some embodiments of process 50 include obtaining test data that quantifies the patient's cardiovascular condition, as indicated by block 52. In some embodiments, this may include obtaining the patient record described above. In some embodiments, the patient record may be calculated in part by applying the measured attributes to a patient model to calculate a dependent variable of the patient's cardiovascular health, for example, where the patient model is configured to map those inputs into corresponding outputs.

[0086] Some embodiments may determine if there are additional dimensions in the test data, as indicated by box 54. Upon determining that there are additional dimensions, in some embodiments, process 50 may select the next dimension, as indicated by box 56. Some embodiments may then normalize the attributes on the current cardiovascular dimension, such as by the normalization technique described above, as indicated by box 58. Embodiments may then return to box 54 and determine if there are more dimensions in the test data process. Upon determining that there are no more dimensions, some embodiments may proceed to box 60 and determine if there are more categories of drugs to process. Upon determining that there are, some embodiments may select the next category, as indicated by box 62.

[0087] Some embodiments may then be used, for example, in Figure 1 , as indicated by box 64. Some embodiments may then determine whether there are more dimensions of predicted influence vectors to be processed, as indicated by box 66. Upon determining that there are more dimensions, some embodiments may select the predicted influence vector for the next dimension being calculated, as indicated by box 68. Some embodiments may then determine the strength of the influence of the current category on the current dimension through the current category model, as indicated by box 70. The embodiment may then return to box 66 to determine whether there are more dimensions of predicted influence vectors to be processed with the current category model.

[0088] Upon determining that there are no more dimensions, in some embodiments, the process may continue to select a subset of categories based on the total score, as indicated by box 74. Some embodiments may then obtain the patient's comorbidities, as indicated by box 76, and filter the subset based on the comorbidities, as indicated by box 78. Again, the term filtering as used herein may include adjusting the corresponding scores or removing certain categories of drugs from other considerations. Some embodiments may then obtain the patient's demographic attributes, as indicated by box 80, for example, through the above-mentioned techniques, and filter the subset based on the demographic attributes, as indicated by box 82. Some embodiments may also filter the subset based on pairwise incompatibilities, for example, through the above-mentioned compatibility filter 36, as indicated by box 84. Some embodiments may continue to rank the filtered subset, as indicated by box 86, and present a sequence of recommended categories to be administered to the patient based on the ranking, for example, on the above-mentioned display 18 through the user interface generator 40, as indicated by box 88.

[0089] In some embodiments, some of the following steps may be performed by the physician when interacting with the user interface. In some embodiments, the physician may determine whether there are more steps in the recommended sequence, as indicated by frame 90. When it is determined that there are more steps, the physician may select the next category of drugs in the sequence, as indicated by frame 92. For example, within a certain duration that may span more than a week or a month and in some cases across multiple office visits, the physician may then administer the selected category of drugs to the patient, as indicated by frame 94. The physician may periodically determine whether to retest the patient, as indicated by frame 96. When it is determined not to retest, the process may return to frame 90 to determine whether there are other steps in the sequence administered to the patient. When it is determined that there are no more steps in the sequence or when it is determined not to retest, as indicated by frame 90 or frame 96, if the patient is healthy, some embodiments may terminate the process, or return to frame 52 to update the sequence based on the updated test of the patient.

[0090] In some embodiments, the above technology can be implemented by specific methods described in the following technical documents. It should be emphasized that this technical document describes a subset of the embodiments and is not intended to characterize all implementation forms of the method, which does not mean that any other description is so restrictive:

[0091] Figure 3 1000 is a diagram illustrating an exemplary computing system 1000 according to an embodiment of the present technology. Portions of the systems and methods described herein may include one or more computer systems similar to computing system 1000, or may be executed on one or more computer systems similar to computing system 1000. In addition, the processes and modules described herein may be executed by one or more processing systems similar to computing system 1000.

[0092] The computing system 1000 may include one or more processors (e.g., processors 1010a to 1010n) coupled to a system memory 1020, an input / output I / O device interface 1030, and a network interface 1040 via an input / output (I / O) interface 1050. The processor may include a single processor or multiple processors (e.g., a distributed processor). The processor may be any suitable processor capable of executing or otherwise executing instructions. The processor may include a central processing unit (CPU) that executes program instructions to perform arithmetic, logic, and input / output operations of the computing system 1000. The processor may execute code (e.g., processor firmware, a protocol stack, a database management system, an operating system, or a combination thereof) that creates an execution environment for program instructions. The processor may include a programmable processor. The processor may include a general-purpose microprocessor or a special-purpose microprocessor. The processor may receive instructions and data from a memory (e.g., system memory 1020). The computing system 1000 may be a single processor system including one processor (e.g., processor 1010a), or a multiprocessor system including any number of suitable processors (e.g., 1010a to 1010n). Multiple processors may be used to provide parallel or sequential execution of one or more parts of the technology described herein. The processes described herein, such as logic flows, may be performed by one or more programmable processors, which may perform functions by operating on input data and generating corresponding outputs. The processes described herein may be performed by dedicated logic circuits, and the devices may also be implemented as dedicated logic circuits, such as FPGAs (field programmable gate arrays) or ASICs (application-specific integrated circuits). The computing system 1000 may include multiple computing devices (e.g., distributed computer systems) to implement various processing functions.

[0093] The I / O device interface 1030 may provide an interface for connecting one or more I / O devices 1060 to the computer system 1000. An I / O device may include a device that receives input (e.g., from a user) or outputs information (e.g., to a user). The I / O device 1060 may include, for example, a graphical user interface presented on a display (e.g., a cathode ray tube (CRT) or liquid crystal display (LCD) monitor), a pointing device (e.g., a computer mouse or trackball), a keyboard, a keyboard, a touch pad, a scanning device, a voice recognition device, a gesture recognition device, a printer, an audio speaker, a microphone, a camera, etc. The I / O device 1060 may be connected to the computer system 1000 via a wired connection or a wireless connection. The I / O device 1060 may be connected to the computer system 1000 from a remote location. The I / O device 1060 located on a remote computer system may be connected to the computer system 1000, for example, via a network and a network interface 1040.

[0094] The network interface 1040 may include a network adapter that provides a connection of the computer system 1000 to a network. The network interface 1040 may facilitate data exchange between the computer system 1000 and other devices connected to the network. The network interface 1040 may support wired or wireless communications. The network may include an electronic communication network, such as the Internet, a local area network (LAN), a wide area network (WAN), a cellular communication network, etc.

[0095] System memory 1020 may be configured to store program instructions 1100 or data 1110. Program instructions 1100 may be executed by a processor (e.g., one or more of processors 1010a to 1010n) to implement one or more embodiments of the present technology. Instructions 1100 may include a module of computer program instructions for implementing one or more techniques described herein regarding various processing modules. Program instructions may include a computer program (which is referred to as a program, software, software application, script, or code in some forms). The computer program may be written in a programming language, which includes a compiled language or an interpreted language, or a declarative language or a procedural language. The computer program may include units suitable for use in a computing environment, including units as independent programs, modules, components, or subroutines. A computer program may or may not correspond to a file in a file system. A program may be stored in a portion of a file that holds other programs or data (e.g., one or more scripts stored in a markup language document), in a single file dedicated to the program in question, or in multiple collaborative files (e.g., files storing one or more modules, subroutines, or code portions). A computer program can be deployed to be executed on one or more computer processors located locally at one site or distributed across multiple remote sites and interconnected by a communication network.

[0096] The system memory 1020 may include a tangible program carrier having program instructions stored thereon. The tangible program carrier may include a non-transitory computer-readable storage medium. The non-transitory computer-readable storage medium may include a machine-readable storage device, a machine-readable storage substrate, a storage device, or any combination thereof. The non-transitory computer-readable storage medium may include a non-volatile memory (e.g., flash memory, ROM, PROM, EPROM, EEPROM memory), a volatile memory (e.g., a random access memory (RAM), a static random access memory (SRAM), a synchronous dynamic RAM (SDRAM)), a mass storage device (e.g., a CD-ROM and / or DVD-ROM, a hard disk drive), etc. The system memory 1020 may include a non-transitory computer-readable storage medium, which may have program instructions stored thereon, which may be executed by a computer processor (e.g., one or more of processors 1010a to -1010n) to cause the subject and functional operations described herein. The memory (e.g., the system memory 1020) may include a single storage device and / or multiple storage devices (e.g., a distributed storage device). Instructions or other program codes that provide the functions described herein may be stored on a tangible, non-transitory computer-readable medium. In some cases, the entire set of instructions may be stored on the medium at the same time, or in some cases, different portions of the instructions may be stored on the same medium at different times.

[0097] The I / O interface 1050 may be configured to coordinate I / O traffic between the processors 1010a-1010n, the system memory 1020, the network interface 1040, the I / O devices 1060, and / or other peripheral devices. The I / O interface 1050 may perform protocol, timing, or other data conversions to convert data signals from one component (e.g., the system memory 1020) into a format suitable for use by another component (e.g., the processors 1010a to 1010n). The I / O interface 1050 may include support for devices attached via various types of peripheral buses, such as a variant of the Peripheral Component Interconnect (PCI) bus standard or the Universal Serial Bus (USB) standard.

[0098] Embodiments of the techniques described herein may be implemented using a single instance of computer system 1000 or multiple computer systems 1000 configured to host different portions or instances of the embodiments. Multiple computer systems 1000 may provide for parallel or sequential processing / execution of one or more portions of the techniques described herein.

[0099] Those skilled in the art will appreciate that computer system 1000 is only illustrative and is not intended to limit the scope of the technology described herein. Computer system 1000 may include any combination of equipment or software, which may perform the technology described herein or otherwise provide the execution of the technology described herein. For example, computer system 1000 may include or be a combination of: cloud computing system, data center, server rack, server, virtual server, desktop computer, laptop computer, tablet computer, server device, client device, mobile phone, personal digital assistant (PDA), mobile audio or video player, game console, vehicle-mounted computer or global positioning system (GPS) etc. Computer system 1000 may also be connected to other devices not shown, or may be used as an independent system. In addition, the functions provided by the components shown may be combined in fewer components or distributed in additional components in some embodiments. Similarly, in some embodiments, the functions of some components in the components shown may not be provided, or other additional functions may be available.

[0100] Those skilled in the art will also understand that, although various items are shown as being stored in memory or stored on storage when in use, for the purpose of memory management and data integrity, these items or a portion of these items can be transmitted between memory and other storage devices. Alternatively, in other embodiments, some or all of the software components can be executed in a memory on another device and communicated with the computer system shown via inter-computer communication. Some or all of the system components or data structures can also be stored (for example, as instructions or structured data) on a computer-accessible medium or portable article to be read by an appropriate drive, and various examples thereof are as described above. In some embodiments, instructions stored on a computer-accessible medium separated from the computer system 1000 can be transmitted to the computer system 1000 via a transmission medium or a signal such as an electrical, electromagnetic or digital signal, and transmitted via a communication medium such as a network or wireless link. Various embodiments may also include receiving, sending or storing instructions or data implemented according to the foregoing description on a computer-accessible medium. Therefore, the present technology can be practiced with other computer system configurations.

[0101] In the block diagram, the components shown are depicted as discrete functional blocks, but the embodiments are not limited to systems in which the functions described herein are organized as shown. The functions provided by each of the components may be provided by software or hardware modules organized in a manner different from the currently depicted manner, for example, such software or hardware may be mixed, combined, replicated, decomposed, distributed (e.g., in a data center or geographically), or otherwise organized. The functions described herein may be provided by one or more processors of one or more computers, which execute code stored on a tangible non-transitory machine-readable medium. In some cases, although the singular term "medium" is used, the instructions may be distributed on different storage devices associated with different computing devices, for example, where each computing device has a different subset of instructions, and the implementation is consistent with the singular term "medium" herein. In some cases, a third-party content delivery network may host part or all of the information transmitted over the network, in which case, with respect to information (e.g., content) provided in accordance with the provision or otherwise provided, the information may be provided by sending instructions to retrieve the information from the content delivery network.

[0102] The reader should understand that this application describes several independently useful technologies. Instead of separating these technologies into multiple independent patent applications, the applicant combined these technologies into one file because their related subject matter helps the economy in the application process. However, the unique advantages and aspects of such technologies should not be lumped together. In some cases, the embodiments solve all the defects mentioned herein, but it should be understood that these technologies are independently useful, and some embodiments only solve a subset of such problems or provide other unmentioned benefits that are obvious to those skilled in the art who review this disclosure. Due to cost constraints, some of the technologies disclosed herein may not be claimed for protection at present, and may be claimed in later applications, such as continued applications or by amending the present claims. Similarly, due to space limitations, neither the abstract nor the content of the invention section of this document should be regarded as a comprehensive list containing all such technologies or all aspects of such technologies.

[0103] It should be understood that this description and the accompanying drawings are not intended to limit the present technology to the specific form disclosed, but rather, are intended to cover all modifications, equivalents, and alternative forms that fall within the spirit and scope of the present technology defined by the appended claims. In view of this description, other modifications and alternative embodiments of various aspects of the present technology will be apparent to those skilled in the art. Therefore, the description and the accompanying drawings should be interpreted as being merely illustrative and for the purpose of teaching those skilled in the art the general manner of implementing the present technology. It should be understood that the forms of the present technology shown and described herein will be considered as examples of embodiments. Elements and materials may replace the elements and materials shown and described herein, parts and processes may be reversed or omitted, and certain features of the present technology may be used independently, all of which will be apparent to those skilled in the art after benefiting from the description of the present technology. The elements described herein may be changed without departing from the spirit and scope of the present technology described in the claims below. The titles used herein are for organizational purposes only and are not meant to limit the scope of this description.

[0104] As used throughout this application, the word "may" is used in a permissive sense (i.e., meaning that it is possible), rather than in a mandatory sense (i.e., meaning that it must be used). The words "include", "include" and "includes" etc. mean to include but are not limited to. As used throughout this application, the singular forms "one", "an" and "said" include plural indicators unless the content explicitly states otherwise. Therefore, for example, although other terms and phrases are used for one or more elements, such as "one or more", reference to "an element" or "an element" includes a combination of two or more elements. Unless otherwise specified, the term "or" is non-exclusive, i.e., including both "and" and "or". Terms describing conditional relationships such as "in response to X, Y", "on X, Y", "if X, Y", "when X, Y", etc., include the causal conditions in which the antecedent is a necessary causal condition, the antecedent is a sufficient causal condition, or the antecedent is a contributing causal condition of the latter, such as "state X occurs when condition Y is obtained" refers to "X only occurs when Y is obtained" and "X occurs when Y and Z are obtained". This conditional relationship is not limited to the consequent immediately following the antecedent, because some consequents may be delayed, and in the conditional statement, the antecedent is related to its consequent, for example, the antecedent is related to the possibility of the consequent occurring. Unless otherwise specified, a statement that maps multiple attributes or functions to multiple objects (e.g., one or more processors that perform steps A, B, C, and D) includes both all such attributes or functions mapped to all such objects and subsets of attributes or functions mapped to subsets of attributes or functions (e.g., all processors each perform steps A to D, and processor 1 performs step A, processor 2 performs step B and part of step C, and processor 3 performs part of step C and step D). Similarly, a reference to a "computer system" that performs step A and a "computer system" that performs step B may include the same computing device within a computer system that performs these two steps or different computing devices within a computer system that performs steps A and B. In addition, unless otherwise specified, a statement that a value or action is "based on" another condition or value includes both the case where the condition or value is the only factor and the case where the condition or value is one of multiple factors. Unless otherwise stated, a statement that "each" instance of a set has certain properties should not be read as excluding the case where some other identical or similar member of the larger set does not have that property, i.e., each does not necessarily mean every and every. Limitations on the order of enumerated steps should not be read into claims unless explicitly stated, for example, using explicit language such as "after performing X, performing Y", in contrast to statements that could be improperly argued to imply an order limitation such as "perform X on the item, perform Y on the item that has performed X", which is used to make the claim more readable rather than to specify an order.Statements referring to "at least Z of A, B, and C," etc. (e.g., "at least Z of A, B, or C") refer to at least Z of the listed categories (A, B, and C) and do not require at least Z elements in each category. Unless otherwise specifically stated, it will be apparent from the discussion that throughout this specification, discussions using terms such as "processing," "computing," "calculating," "determining," etc. refer to the actions or processes of a specific device, such as a special purpose computer or similar special purpose electronic processing / computing equipment. Features described with reference to geometric configurations, such as "parallel," "perpendicular / orthogonal," "square," "cylindrical," etc., should be interpreted to include items that substantially embody the characteristics of the geometric configurations, such as reference to "parallel" surfaces includes substantially parallel surfaces. The range of allowable deviations of these geometric structures from the Platonic ideal will be determined by reference to the ranges in the specification, and if such ranges are not stated, then by reference to the industry specifications for the field of use, and if such ranges are not defined, then by reference to the industry specifications for the field of manufacture of the specified features, and if such ranges are not defined, then features that substantially embody the geometric configuration should be interpreted as including those features within 15% of the defined attributes of the geometric configuration. If the terms "first", "second", "third", "given", etc. are used in the claims, they are used to distinguish or otherwise identify, and not to express an ordinal limitation or a numerical limitation. As is common usage in the art, data structures and formats described with reference to significant human use do not need to be presented in a human-understandable format to constitute the described data structures or formats, for example, text does not need to be presented or even encoded in Unicode or ASCII to constitute text; images, maps, and data visualizations do not need to be displayed or decoded to constitute images, maps, and data visualizations, respectively; speech, music, and other audio do not need to be emitted or decoded through speakers to constitute speech, music, or other audio, respectively. Computer-implemented instructions, commands, and the like are not limited to executable code, and may be implemented in the form of data that causes the function to be invoked, such as in the form of parameters to a function or API call.

[0105] In this patent, to the extent that any U.S. patent, U.S. patent application, or other material (e.g., article) has been incorporated by reference, the text of such material is incorporated by reference only to the extent that there is no conflict between such material and the statements and drawings presented herein. In the event of such a conflict, the text of this document controls, and the terms in this document should not be construed narrowly based on how such terms are used in other materials incorporated by reference.

[0106] The present technology will be better understood with reference to the following examples:

[0107] 1. A method, comprising: obtaining, by a computer system, test data quantifying a patient's cardiovascular condition, the test data specifying patient attributes on a plurality of cardiovascular dimensions of the patient, a first subset of the plurality of cardiovascular dimensions being independent cardiovascular dimensions, and a second subset of the plurality of cardiovascular dimensions being dependent cardiovascular dimensions; determining, by the computer system, a plurality of normalized differences between the test data and a target standard on each of the cardiovascular dimensions, the plurality of normalized differences quantifying different aspects of hypertension in the patient; determining, by the computer system, a predicted impact vector for each of a plurality of different classes of drugs having hypertension as an indication, wherein: the predicted impact vectors each correspond to a different corresponding class of drugs in the different classes of drugs, and the predicted impact vectors each have a plurality of values, The multiple values ​​quantify the corresponding effects of the corresponding categories of drugs on the corresponding dimensions in the third set of the patient's cardiovascular dimensions, each of the values ​​is based on the corresponding normalized difference on the corresponding dimensions in the third set and the corresponding strength of the effects of the corresponding categories of drugs on the corresponding dimensions, the corresponding strength effects are determined based on the respective multidimensional models of the corresponding categories of drugs, and the third set overlaps with at least the first subset and the second subset; through the computer system, based on the values ​​of the corresponding predicted impact vectors of the corresponding categories of drugs, the total score of each corresponding category of drugs in the different categories of drugs is determined; through the computer system, the different categories of drugs are ranked based on the total scores to form a ranked list of the different categories of drugs; and through the computer system, based on the ranking, a recommended sequence for administering the categories of drugs to the patient is output.

[0108] 2. The method according to embodiment 1 comprises: accessing the pairwise compatibility of the different categories of drugs through the computer system; and based on the pairwise compatibility, detecting incompatible pairs of the different categories of drugs in the ranking list; and removing at least one member of each detected incompatible pair from the ranking list through the computer system to form a filtered ranking list.

[0109] 3. A method according to embodiment 2, wherein: removing at least one member from each detected pair by the computer system includes removing a lower ranked member from each detected pair.

[0110] 4. The method of embodiment 2, wherein forming the filtered ranking list comprises: obtaining comorbidity or demographic information of the patient through a computer system; and adjusting the filtered ranking list based on the obtained comorbidity or demographic information through the computer system.

[0111] 5. The method of embodiment 4, comprising: causing, by the computer system, a user interface to be presented, the user interface indicating a warning for a given class of drugs in the classes of drugs in response to the comorbidity or demographic information.

[0112] 6. The method according to Example 2 includes: administering a first category of drugs with a first ranking in the filtered ranking list; and after administering the first category of drugs, administering a second category of drugs with a second ranking in the filtered ranking list, wherein the second ranking is lower than the first ranking in the ranking.

[0113] 7. The method described in any one of embodiments 1-6 further includes: obtaining, through the computer system, comorbidities or demographic attributes of the patient, wherein: a given value in the value of the given predicted influence vector in the predicted influence vector is based on the comorbidity or demographic attribute, at least some of the values ​​in the given predicted influence vector in the predicted influence vector are not affected by the comorbidity or demographic attribute, and at least some of the predicted influence vectors are not affected by the comorbidity or demographic attribute.

[0114] 8. A method according to any one of embodiments 1-7, wherein: the ranking is based on a predicted influence vector value corresponding to each of the dimensions of the first subset.

[0115] 9. The method of embodiment 8, wherein: the ranking is not based on predicted influence vector values ​​corresponding to any one of the dimensions of the second subset.

[0116] 10. The method of embodiment 8, wherein the dimensions of the first subset include at least two of the following dimensions: systolic blood pressure, pulse pressure, cardiac index, or heart rate.

[0117] 11. The method of embodiment 8, wherein the dimensions of the first subset include each of the following dimensions: systolic blood pressure, pulse pressure, cardiac index, and heart rate.

[0118] 12. A method according to any one of embodiments 1-11, wherein the category of drugs includes at least six of the following categories of drugs: diuretics, beta blockers, ACE inhibitors, angiotensin II receptor blockers, calcium channel blockers, alpha blockers, alpha2 receptor agonists, combined alphabeta blockers, central nervous system agonists, peripheral adrenergic inhibitors, or vasodilators.

[0119] 13. The method of any one of embodiments 1-11, wherein the class of drugs includes each of the following classes of drugs: diuretics, beta-blockers, ACE inhibitors, angiotensin II receptor blockers, calcium channel blockers, alpha blockers, alpha-2 receptor agonists, combined alpha beta blockers, central agonists, peripheral adrenergic inhibitors, and vasodilators.

[0120] 14. A method according to any one of embodiments 1-13, wherein at least some of the multiple normalized differences are z-scores that specify a number of standard deviations of the corresponding patient attribute in the test data from a population mean.

[0121] 15. A method according to any one of embodiments 1-13, wherein a given score in the total score for a given class of drugs in the class of drugs is based on a weighted sum of the values ​​of the corresponding predicted impact vectors of the given class of drugs in the class of drugs.

[0122] 16. A method according to any one of embodiments 1-15, wherein at least some of the values ​​are each based on the product of the corresponding normalized difference value of the corresponding dimension in the third set and the corresponding strength of the effect of the corresponding category of drug on the corresponding dimension.

[0123] 17. A method according to any one of embodiments 1-16, wherein a given multidimensional model of a given class of drugs in the class of drugs specifies a non-planar surface of five dimensions or higher, and the five dimensions or higher include: an output dimension corresponding to the intensity of the impact of the given class of drugs in the class of drugs, and four or more cardiovascular dimensions of the patient.

[0124] 18. The method according to any one of embodiments 1-17 comprises: repeating, by the computer system, the method of embodiment 1 using an updated instance of test data obtained after administering a portion of a sequence of a class of drugs to a patient; and changing, by the computer system, the sequence as a result of the repetition.

[0125] 19. A tangible, non-transitory machine-readable medium storing instructions that, when executed by a data processing device, cause the data processing device to perform operations, the operations comprising: computer-implemented operations according to any one of embodiments 1-18.

[0126] 20. A system comprising: one or more processors; and a memory storing instructions, which when executed by the processors cause the processors to implement operations, the operations comprising: computer-implemented operations according to any one of embodiments 1-18.

Claims

1. A method of patient management and care, comprising: obtaining, by a computer system, test data quantifying a cardiovascular condition of a patient, the test data specifying attributes of the patient in a plurality of cardiovascular dimensions of the patient, a first subset of the plurality of cardiovascular dimensions being independent cardiovascular dimensions and a second subset of the plurality of cardiovascular dimensions being dependent cardiovascular dimensions; determining, by the computer system, a plurality of normalized differences between the test data and a target standard in each of the cardiovascular dimensions, the plurality of normalized differences quantifying different aspects of hypertension in the patient; By the computer system, a predicted impact vector for each class of drugs in a plurality of different classes of drugs having hypertension as an indication is determined, wherein: The predicted impact vectors each correspond to a different corresponding class of drugs among the different classes of drugs, The predicted impact vectors each have a plurality of values ​​quantifying a respective impact of the corresponding class of drugs on a corresponding dimension in a third set of cardiovascular dimensions of the patient, The values ​​are each based on a corresponding normalized difference on a corresponding dimension in the third set and a respective strength of the effect of the corresponding class of drugs on the corresponding dimension, determining the corresponding intensity impact based on the corresponding multidimensional model of the corresponding class of drugs, and The third set overlaps with at least the first subset and the second subset; determining, by the computer system, a total score for each corresponding class of drugs in the different classes of drugs based on the value of the corresponding predicted impact vector for the corresponding class of drugs; Ranking the different categories of drugs based on the total scores by the computer system to form a ranked list of the different categories of drugs; and Outputting, by the computer system, a recommended sequence for administering the class of drugs to the patient based on the ranking.

2. The method according to claim 1, comprising: accessing, by the computer system, pairwise compatibilities of the different classes of drugs; as well as Based on the pairwise compatibilities, detecting incompatible pairs of the drugs of different categories in the ranked list; as well as At least one member of each detected incompatible pair is removed from the ranked list by the computer system to form a filtered ranked list.

3. The method according to claim 2, wherein: Removing, by the computer system, at least one member from each detected pair includes removing a lower ranked member from each detected pair.

4. The method according to claim 2, wherein: The filtered ranking list includes: obtaining, through the computer system, comorbidity or demographic information of the patient; adjusting, by the computer system, the filtered ranked list based on the obtained comorbidity or demographic information; and A user interface is caused to be presented, by the computer system, the user interface indicating a warning for a given one of the classes of drugs in response to the comorbidity or demographic information.

5. The method according to claim 2, comprising: administering a first-ranked first-category drug in the filtered ranked list; as well as After administering the first category of drugs, administering a second category of drugs ranked second in the filtered ranked list, the second ranking being lower in the ranking than the first ranking.

6. The method according to any one of claims 1 to 5, comprising: obtaining, by the computer system, comorbidities or demographic attributes of the patient, in: A given one of the values ​​of a given one of the predicted influence vectors is based on the comorbidity or demographic attribute, at least some of the values ​​of the given one of the predicted influence vectors are not affected by the comorbidity or demographic attribute, and At least some of the predicted influence vectors are not influenced by the comorbidity or demographic attributes.

7. The method according to any one of claims 1 to 5, wherein: The ranking is based on predicted influence vector values ​​corresponding to each of the dimensions of the first subset.

8. The method according to claim 7, wherein: The ranking is not based on predicted influence vector values ​​corresponding to any of the dimensions of the second subset.

9. The method according to claim 7, wherein: The first subset of dimensions includes each of the following dimensions: systolic blood pressure, pulse pressure, cardiac index, and heart rate.

10. The method according to any one of claims 1 to 5, wherein: The class of drugs includes at least six of the following classes of drugs: diuretics, beta blockers, Angiotensin-converting enzyme (ACE) inhibitors, Angiotensin II receptor blockers (ARBs), calcium channel blockers, alpha blockers, Alpha-2 receptor agonists, Combined with αβ blockers, Central nervous system agonists, Peripheral adrenergic inhibitors, or Vasodilator.

11. The method according to any one of claims 1 to 5, wherein: A given multidimensional model for a given class of drugs in the classes of drugs specifies a non-planar surface in five or more dimensions, the five or more dimensions comprising: an output dimension corresponding to the strength of the effect of said given class of drugs in said class of drugs, and Four or more cardiovascular dimensions of the patient.

12. The method according to any one of claims 1 to 5, comprising: The step of determining a sequence of classes of drugs to be administered to a patient.

13. The method according to any one of claims 1 to 5, comprising: repeating, by the computer system, the method of claim 1 using an updated instance of the test data obtained after a portion of a sequence of a class of drugs is administered to a patient; and As a result of said repetition, said sequence is altered, by said computer system.

14. A tangible, non-transitory machine-readable medium storing instructions which, when executed by one or more processors, implement the method according to any one of claims 1 to 13.

15. An artificial intelligence system comprising: one or more processors; and a memory storing instructions which, when executed by at least some of the processors, cause the processors to implement the method according to any one of claims 1 to 13.

Images