Back to results
US 2018/0121618 A1Published application
System and method for extracting oncological information of prognostic significance from natural language
AssigneeCOTA, Inc.
InventorsStephen August Smith, Dilip Raj, Ali Hasan, Idan Waisman, Cory Parent
Life of this patent
Estimating the term…
Abstract
A system, nontransitory computer readable medium, and method are provided for extracting data from unstructured medical text. Data points are identified in unstructured medical text, where the data points are determined using a dictionary database. A value associated with each of the data points is determined from the unstructured medical text. Each of the data points is mapped to its respective value for extraction from the unstructured medical text.
Claims30
A method for selectively extracting objective oncologic data of diagnostic or prognostic significance from subjective unstructured medical text in a natural language, the method comprising:
- (A) Receiving input comprising unstructured medical text including unstructured oncologic text;
- (B) Processing the unstructured medical text to generate attribute-value pairs, the processing including:
- (a) selectively identifying, in the unstructured medical text, data points and equivalents of the data points that are significant for diagnosis and prognosis as extracted data points, the identification based on a database comprising a dictionary of medical knowledge that includes data points significant to diagnosis and prognosis and equivalents of the data points;
- (b) selectively extracting from the unstructured medical text all facts of known importance associated with words or phrases in the unstructured medical text that are syntactically or semantically dependent on or related to the extracted data points;
- (c) associating the extracted data points with the words or phrases in the unstructured medical text that are syntactically or semantically dependent on, or related to, the extracted data points;
- (d) mapping each extracted data point (attribute) with its syntactically or semantically dependent word or phrase (value) in (c) to generate attribute-value pairs;
- and
- (e) for at least some of the attribute-value pairs, standardizing the attribute in the attribute-value pair according to a code that represents the attribute;
- (C) Validating the standardized attribute-value pairs to ensure oncologic integrity, the validation including:
- (i) identifying inconsistencies between different standardized attribute-value pairs using an oncologic model;
- and
- (ii) identifying incorrect values by identifying values that lay outside a possible range for a corresponding attribute or that are not in a possible set of values for the corresponding attribute;
- (D) grouping the standardized attribute-value pairs into one or more lists of one or more grouped attribute-value pairs based on the oncologic model;
- and
- (E) outputting at least one list of grouped validated attribute-value pairs.
12 more elements
A non-transitory computer readable storage medium storing computer program instructions for selectively extracting objective oncologic data of diagnostic or prognostic significance from unstructured medical text in a natural language, which, when executed on a processor, cause the processor to perform operations comprising:
- (A) Receiving input comprising unstructured medical text including unstructured oncologic text;
- (B) Processing the unstructured medical text to generate attribute-value pairs, the processing including:
- (a) selectively identifying in the unstructured medical text data points and equivalents of the data points that are significant for diagnosis and prognosis as extracted data points, the identification based on a database comprising a dictionary of medical knowledge that identifies data points significant to diagnosis and prognosis and equivalents of the data points;
- (b) selectively extracting from the unstructured medical text all facts of known importance associated with words or phrases in the unstructured medical text that are syntactically or semantically dependent on or related to the extracted data points;
- (c) associating the extracted data points with the words or phrases in the unstructured medical text that are syntactically or semantically dependent on, or related to, the extracted data points;
- (d) mapping each extracted data point (attribute) with its syntactically or semantically dependent word or phrase (value) in (c) to generate attribute value pairs;
- and
- (e) for at least some of the attribute-value pairs, standardizing the attribute in the attribute-value pair according to a code that represents the attribute;
- (C) Validating the standardized attribute-value pairs to ensure oncologic integrity, the validation including:
- (i) identifying inconsistencies between different standardized attribute-value pairs using an oncologic model;
- and
- (ii) identifying incorrect values by identifying values that lay outside a possible range for a corresponding attribute or that are not in a possible set of values for the corresponding attribute;
- (D) grouping the standardized attribute-value pairs into one or more lists of one or more grouped attribute-value pairs based on the oncologic model;
- and
- (E) outputting at least one list of grouped validated attribute-value pairs.
12 more elements
A system for selectively extracting objective oncologic data of diagnostic or prognostic significance from unstructured medical text in a natural language, comprising:
- a first database comprising a dictionary of medical knowledge that identifies data points as significant to diagnosis and prognosis and equivalents of the data points;
- a second database comprising personal health information data for a population of human subjects;
- wherein the first and second database are communicatively linked using a common patient identifier and through the use of database access using the common patient identifier;
- and
- an extraction system comprising:
- a computer server including a processor comprising a clinical outcome tracking and analysis module communicatively linked to the first database, the second database, and the network;
- and
- a memory to store computer program instructions, the computer program instructions when executed on the processor cause the processor to perform operations comprising:
- (A) Receiving input comprising unstructured medical text including unstructured oncologic text;
- (B) Processing the unstructured medical text to generate attribute-value pairs, the processing including:
- (a) selectively identifying in the unstructured medical text data points and equivalents of the data points that are significant for diagnosis and prognosis as extracted data points, the identification based on the dictionary of medical knowledge of the first database;
- (b) selectively extracting from the unstructured medical text all facts of known importance associated with words or phrases in the unstructured medical text that are syntactically or semantically dependent on or related to the extracted data points;
- (c) associating each extracted data point with the words or phrases in unstructured medical text that are syntactically or semantically dependent on, or related to, the extracted data point;
- (d) mapping each extracted data point (attribute) with its syntactically or semantically dependent word or phrase (value) in (c) to generate attribute value pairs;
- and
- (e) for at least some of the attribute-value pairs, standardizing the attribute in the attribute-value pair according to a code that represents the attribute;
- (C) Validating the standardized attribute-value pairs to ensure oncologic integrity by:
- (i) identifying inconsistencies between different standardized attribute-value pairs using an oncologic model;
- and
- (ii) identifying incorrect values by identifying values that lay outside a possible range for a corresponding attribute or that are not in a possible set of values for the corresponding attribute;
- (D) grouping the standardized attribute-value pairs into one or more lists of one or more grouped attribute-value pairs based on the oncologic model;
- and
- (E) outputting at least one list of grouped validated attribute-value pairs.
20 more elements
Dependencies are inferred from the claim text.