Researchers Submit Patent Application, “Predictive System For Generating Clinical Queries”, for Approval (USPTO 20230281253): Patent Application

2023 SEP 26 (NewsRx) -- By a News Reporter-Staff News Editor at Insurance Daily News -- From Washington, D.C., NewsRx journalists report that a patent application by the inventors Arbona, Joaquin Palancar (Madrid, ES); Duishoev, Nurlanbek (Durham, NC, US); Glass, Lucas (Devon, PA, US); Morgan, Kristy (Chapel Hill, NC, US); Sakhrani, Shyam (London, GB), filed on March 24, 2023, was made available online on September 7, 2023.

No assignee for this patent application has been made.

News editors obtained the following quote from the background information supplied by the inventors: “As part of the healthcare process, physicians or other medical care providers may perform clinical trials, programs, and other activities to evaluate subject safety and efficacy of a pharmaceutical drug or other medical treatment option. The use of health-related trial programs can help to identify novel treatment options for improving overall patient health and reducing health system costs. A clinical trial or program can be a single research study or multiple research studies that prospectively assigns human participants/subjects or groups of human subjects to one or more health-related interventions to evaluate the effects on health outcomes.”

As a supplement to the background information on this patent application, NewsRx correspondents also obtained the inventors’ summary information for this patent application: “As part of the healthcare process, physicians or other medical care providers may perform trials, programs, and other activities to evaluate the efficacy of a particular pharmaceutical drug or other medical treatment option. Conducting health-related clinical trials can help to identify medical treatment options for improving overall patient health and reducing health system costs. Clinical trials and other controlled programs are generally conducted by one or more investigators at medical facilities in different geographic locations that interact with study subjects to evaluate the efficacy of a drug treatment option. In some instances a physician for a patient can be associated with a clinical trial and the physician can refer a patient as a candidate for participation in a trial based on a diagnosed condition of the patient. An investigator, a geographic location, or both, can form an entity that executes a program.

“Based on the above context, this document describes a computing system that uses specific computing rules or instructions (e.g., a unique algorithm) to predict or generate commands based on a received user input. To generate the commands, the system is configured to train a predictive model using one or more learning algorithms (e.g., deep learning algorithms). The predictive model is used to process terms that are recognized and extracted using a natural language processor (NLP) in an entity module of the system. The predictive model can be trained to semantically understand relevant terms (e.g., medical and clinical terms) and their relations to other medical terms. Terms can be extracted from information sources such as textbooks and online resources, or from unstructured datasets such as electronic medical data for multiple healthcare patients.

“An encoding module uses one or more neural network models to encode and link the extracted terms to a particular medical entity, such as a disease entity, a drug entity, a medical procedure entity, or various other types of entities. The system leverages the predictive model’s learned inferences about the encoded medical terms to generate a command based on a received query from a user. For example, a parsing engine can automatically translate the extracted terms into a machine-readable command that is processed against a medical database to obtain an accurate response to the user query. Hence, at least one goal of the predictive system is to accurately interpret, in a manner that is computationally efficient, a user query that includes health-related information about a patient or set of patients.

“For example, the user query represents user input, to the predictive system, that specifies a list(s) of patient attributes. The query/user input can be in a human-readable format. The described techniques enable the system to quickly and efficiently generate a corresponding command in a computer-readable format. The computer-readable command is then used to query different electronic health records (EHR) to identify patients (e.g., trial subjects) that satisfy a given condition(s) based on the attributes specified in the list. For example, the command can be used to query historical medical records to derive insights and information without manual intervention (e.g., from a human-operator). The derived insights can include accurate estimation of an eligible patient population for clinical trial participation and estimation of a propensity for adverse events).

“One aspect of the subject matter described in this specification can be embodied in a computer-implemented method that includes: obtaining a first set of data including multiple terms; determining that a term of the multiple terms describes a medical entity; responsive to determining that the term describes the medical entity, linking the medical entity with a category based on an encoding scheme for the category; responsive to receiving a query, generating a machine-readable command by parsing the query against terms in the first set of data that describe the medical entity and based on the encoding scheme; using the machine-readable command to query multiple databases; obtaining a second set of data responsive to the received query when the machine-readable command is used to query the multiple databases; and providing the second set of data as an output for display at a user device.

“These and other implementations can each optionally include one or more of the following features. For example, in some implementations, determining whether a term describes the medical entity includes: generating a confidence score based on inferences of similarity between terms described in the first set of data and the medical entity; and determining the confidence score exceeds a threshold confidence score.

“In some implementations, linking the medical entity with the category includes: obtaining a listing of category codes for the category; determining a match between the term and corresponding category codes in the listing of category codes; and linking the medical entity with the specified category based on the match between the term that describes the medical entity and the corresponding category codes.

“In some implementations, linking the medical entity with the category includes: encoding the medical entity with corresponding category codes based on the encoding scheme for the specified category; and the encoding scheme for the specified category is a hierarchical encoding scheme including a hierarchy of levels.

“In some implementations, encoding the medical entity with corresponding category codes includes: quantifying content including the medical entity to be encoded; determining depths of levels in the hierarchy of levels for mapping the content; and associating the medical entity included in the content with the corresponding category codes for a particular depth level in the hierarchy of levels.

“In some implementations, the medical entity is a disease, and determining the match includes: generating a respective match score for each level in the hierarchy of levels; and determining that the respective match score exceeds a threshold match score.

“In some implementations, determining that the term describes the medical entity includes: performing a lookup of the term against information in an entity-specific dataset; and determining that the term describes the medical entity based on a match between the term and a first entry in the entity-specific dataset.

“In some implementations, the medical entity is associated with a healthcare condition that affects an individual, and the medical entity includes at least one of: one or more medical diseases; medical drugs for treating the one or more medical diseases; medical procedures associated with the one or more medical diseases; or data describing multiple medical findings that correspond to a healthcare condition of the individual.

“In some implementations, the entity specific dataset is generated based on data including at least one of: i) a predefined set of information describing multiple diseases; ii) a predefined set of information describing multiple drugs; iii) a predefined set of information describing multiple medical procedures; or iv) electronic medical data for multiple healthcare patients.

“In some implementations, obtaining the data that describes the terms relating to the multiple medical concepts includes: obtaining multiple unstructured data; and structuring the unstructured data to enable processing of the query against information in the multiple databases.

“Other implementations of this and other aspects include corresponding systems, apparatus, and computer programs, configured to perform the actions of the methods, encoded on computer storage devices. A computing system of one or more computers or hardware circuits can be so configured by virtue of software, firmware, hardware, or a combination of them installed on the system that in operation cause the system to perform the actions. One or more computer programs can be so configured by virtue of having instructions that, when executed by data processing apparatus, cause the apparatus to perform the actions.

“The subject matter described in this specification can be implemented to realize one or more of the following advantages. The described techniques provide a scalable computing system that is a fully automated end-to-end predictive solution for analyzing and parsing structured and unstructured datasets. Using the analyzing and parsing functions, a predictive model of the system is configured such that information in the datasets can be queried using a machine-readable command that is generated based on data inferences learned by the predictive model.

“The predictive system is configured to quickly and efficiently analyze multiple datasets that describe a variety of diseases and indications, drugs/treatment options, and medical procedures. For example, the efficiency of the system is evidenced by the use of three steps to generate the command: (1) extraction of entities that describe patient attributes; (2) mapping a condition to a standardized scientific entity name; and (3) interpreting relationships between different healthcare conditions, including whether the conditions are negated or not.

“Hence, the system provides a solution that specializes in recognizing and encoding medical language terms and requires little (or no) manual data curation to achieve a desired level of accuracy in the commands or machine-readable queries that are generated and processed to obtain a response to user input. The predictive system uses learning algorithms (e.g., deep learning algorithms) to determine relationships between relevant categories of information and uses the relations between the information categories to directly query medical and research databases.”

There is additional summary information. Please visit full patent to read further.”

The claims supplied by the inventors are:

“1. (canceled)

“2. A computer system-implemented method comprising: encoding, using one or more machine learning models, one or more terms to a level in an encoding scheme of a category associated with the one or more terms, wherein each machine learning model is associated with a particular level in the encoding scheme; in response to receiving a first query, generating a second query at least by parsing the first query using the encoding scheme; and providing a reply to the first query in response to querying one or more databases using the generated second query.

“3. The computer system-implemented method of claim 2, wherein the encoding scheme comprises a hierarchy of levels, each level of the hierarchy of levels reflects a sub-category of the category.

“4. The computer system-implemented method of claim 2, wherein encoding the one or more terms to the level in the encoding scheme of the category associated with the one or more terms further comprises: for each term of the one or more terms: determining, using a model, an entity for a term; providing the entity to each of the one or more machine learning models; in response to providing the entity to each of the one or more machine learning models, obtaining a confidence score from each of the one or more machine learning models; selecting an output confidence score that exceeds the other confidence scores; and encoding, using the one of the machine learning models that produced the selected output confidence score, the entity to a level in the encoding scheme of the category that is associated with the one of the machine learning model.

“5. The computer system-implemented method of claim 4, wherein determining, using the model, the entity for the term comprises: generating, by the model, a confidence score for each of the one or more terms that describe the entity; comparing, by the model, the confidence score for each of the one or more terms to a threshold value; and in response to determining that the confidence score for each of the one or more terms exceeds the threshold value, determining, by the model, the entity for the term.

“6. The computer system-implemented method of claim 4, further comprising: obtaining a listing of category codes for the category; determining a match between the term from each of the one or more terms and corresponding category codes in the listing of category codes; and linking the entity with the category based on the match between the term that describes the entity and the corresponding category codes.

“7. The computer system-implemented method of claim 6, wherein linking the entity with the category comprises encoding the entity with corresponding category codes based on the encoding scheme for the category.

“8. The computer system-implemented method of claim 2, wherein generating the second query at least by parsing the first query using the encoding scheme comprises: identifying one or more terms in the first query; for each of the one or more terms identified in the first query: identifying an entity described by the term based on the encoding scheme for linking the entity to the category; and generating a machine readable command using each of the identified entities.

“9. The computer system-implemented method of claim 8, wherein providing the reply to the first query in response to querying the one or more databases using the generated second query comprises: querying the one or more databases using the generated machine readable command; in response to querying the one or more databases, receiving one or more data elements; generating the reply using the one or more data elements; and providing the reply to a client device that transmitted the first query.

“10. A system comprising: one or more computers and one or more storage devices storing instructions that are operable, when executed by the one or more computers, to cause the one or more computers to perform operations comprising: encoding, using one or more machine learning models, one or more terms to a level in an encoding scheme of a category associated with the one or more terms, wherein each machine learning model is associated with a particular level in the encoding scheme; in response to receiving a first query, generating a second query at least by parsing the first query using the encoding scheme; and providing a reply to the first query in response to querying one or more databases using the generated second query.

“11. The system of claim 10, wherein the encoding scheme comprises a hierarchy of levels, each level of the hierarchy of levels reflects a sub-category of the category.

“12. The system of claim 10, wherein encoding the one or more terms to the level in the encoding scheme of the category associated with the one or more terms further comprises: for each term of the one or more terms: determining, using a model, an entity for a term; providing the entity to each of the one or more machine learning models; in response to providing the entity to each of the one or more machine learning models, obtaining a confidence score from each of the one or more machine learning models; selecting an output confidence score that exceeds the other confidence scores; and encoding, using the one of the machine learning models that produced the selected output confidence score, the entity to a level in the encoding scheme of the category that is associated with the one of the machine learning model.

“13. The system of claim 12, wherein determining, using the model, the entity for the term comprises: generating, by the model, a confidence score for each of the one or more terms that describe the entity; comparing, by the model, the confidence score for each of the one or more terms to a threshold value; and in response to determining that the confidence score for each of the one or more terms exceeds the threshold value, determining, by the model, the entity for the term.

“14. The system of claim 12, further comprising: obtaining a listing of category codes for the category; determining a match between the term from each of the one or more terms and corresponding category codes in the listing of category codes; and linking the entity with the category based on the match between the term that describes the entity and the corresponding category codes.

“15. The system of claim 14, wherein linking the entity with the category comprises encoding the entity with corresponding category codes based on the encoding scheme for the category.

“16. The system of claim 10, wherein generating the second query at least by parsing the first query using the encoding scheme comprises: identifying one or more terms in the first query; for each of the one or more terms identified in the first query: identifying an entity described by the term based on the encoding scheme for linking the entity to the category; and generating a machine readable command using each of the identified entities.

“17. The system of claim 16, wherein providing the reply to the first query in response to querying the one or more databases using the generated second query comprises: querying the one or more databases using the generated machine readable command; in response to querying the one or more databases, receiving one or more data elements; generating the reply using the one or more data elements; and providing the reply to a client device that transmitted the first query.

“18. A non-transitory computer-readable medium storing software comprising instructions executable by one or more computers which, upon such execution, cause the one or more computers to perform operations comprising: encoding, using one or more machine learning models, one or more terms to a level in an encoding scheme of a category associated with the one or more terms, wherein each machine learning model is associated with a particular level in the encoding scheme; in response to receiving a first query, generating a second query at least by parsing the first query using the encoding scheme; and providing a reply to the first query in response to querying one or more databases using the generated second query.

“19. The non-transitory computer-readable medium of claim 18, wherein the encoding scheme comprises a hierarchy of levels, each level of the hierarchy of levels reflects a sub-category of the category.

“20. The non-transitory computer-readable medium of claim 18, wherein encoding the one or more terms to the level in the encoding scheme of the category associated with the one or more terms further comprises: for each term of the one or more terms: determining, using a model, an entity for a term; providing the entity to each of the one or more machine learning models; in response to providing the entity to each of the one or more machine learning models, obtaining a confidence score from each of the one or more machine learning models; selecting an output confidence score that exceeds the other confidence scores; and encoding, using the one of the machine learning models that produced the selected output confidence score, the entity to a level in the encoding scheme of the category that is associated with the one of the machine learning model.

“21. The non-transitory computer-readable medium of claim 20, wherein determining, using the model, the entity for the term comprises: generating, by the model, a confidence score for each of the one or more terms that describe the entity; comparing, by the model, the confidence score for each of the one or more terms to a threshold value; and in response to determining that the confidence score for each of the one or more terms exceeds the threshold value, determining, by the model, the entity for the term.”

For additional information on this patent application, see: Arbona, Joaquin Palancar; Duishoev, Nurlanbek; Glass, Lucas; Morgan, Kristy; Sakhrani, Shyam. Predictive System For Generating Clinical Queries. U.S. Patent Application Number 20230281253, filed March 24, 2023 and posted September 7, 2023. Patent URL (for desktop use only): https://ppubs.uspto.gov/pubwebapp/external.html?q=(20230281253)&db=US-PGPUB&type=ids

(Our reports deliver fact-based news of research and discoveries from around the world.)