Identifying potential patient candidates for clinical trials

Patent metadata

Field	Value
Publication number	US-10162866-B2
Application number	US-201715672717-A
Country	US
Kind code	B2
Filing date	Aug 9, 2017
Priority date	Jun 3, 2016
Publication date	Dec 25, 2018
Grant date	Dec 25, 2018

Abstract

A computer system gleans data from patient records and clinical trial descriptions using NLP techniques. NLP annotation data is used to generate clinical trial feature vectors and patient feature vectors. Clinical trial feature vectors and patient feature vectors are compared to match appropriate patient candidates with clinical trial openings.

First claim

What is claimed is:

1. A computer-implemented method for matching clinical trial openings with candidates from a patient population, the method comprising: identifying a first clinical trial description;
generating annotations of the first clinical trial description, based on natural language processing techniques including Unstructured Information Management Architecture (UIMA), wherein the UIMA includes a language identification module using n-gram models for determining the language of the first clinical trial description; a linguistic analysis module for identifying parts of speech, syntactic position, and function; a dictionary module for defining terms and identifying synonyms; a named entity recognition module for identifying names, locations, and companies; a pattern recognition module for interpreting phrases and strings of numbers; a classification module for classifying the first clinical trial description according to content; and one or more custom annotators for identifying clinical features, disease states, diagnoses, one or more patient medications, one or more patient demographics, a patient-provider relationship, and a stage of disease progression; and wherein the annotations populate text indices, triple stores, and relational databases;
generating a first trial feature vector based on the annotations of the first clinical trial description, the first trial feature vector comprising a first array of values in a matrix, wherein the first array of values represents disease states, diagnoses, one or more patient medications, one or more patient demographics, a patient-provider relationship, and a stage of disease progression;
determining that a first value in the first array of values should be prioritized;
transforming the first value into a weighted vector value, wherein the weighted vector value is used to generate a second trial feature vector;
retrieving the first patient record and other patient records stored in a collection database, using a crawler, based on similarities between the annotations of the clinical trial description and contents of the patient records;
applying natural language processing techniques to the first patient record, the natural language processing techniques including Unstructured Information Management Architecture (UIMA), wherein the UIMA includes a language identification module using n-gram models for determining the language of the first patient record; a linguistic analysis module for identifying parts of speech, syntactic position and function; a dictionary module for defining terms and identifying synonyms; a named entity recognition module for identifying names, locations, and companies; a pattern recognition module for interpreting phrases and strings of numbers; a classification module for classifying the first patient record according to content; and one or more custom annotators for identifying clinical features, disease states, diagnoses, one or more patient medications, one or more patient demographics, a patient-provider relationship, and a stage of disease progression; wherein the UIMA generates annotations of the first patient record, the annotations populating one or more of text indices, triple stores, and relational databases;
generating a first patient feature vector based on the annotations of the first patient record, the first patient feature vector comprising the second array of values in a matrix, wherein the second array of values represents disease states, diagnoses, one or more patient medications, one or more patient demographics, a patient-provider relationship, and a stage of disease progression;
generating a comparison value based on the second trial feature vector and the first patient feature vector, wherein the comparison value represents a logical distance between the clinical trial feature vector and the patient feature vector;
displaying, via an interactive user interface, information about a patient associated with the patient record to a user based on the comparison value, wherein the information about the patient, including at least the comparison value, an identification of the patient, and a contact information for the patient, is displayed with information about other patients based on a plurality of comparison values generated from a plurality of patient records associated with the other patients, and wherein the interactive user interface allows the user to sort the information about the patient and the information about other patients based on sorting options comprising comparison values, patient name, and patient contact information;
receiving, through the interactive user interface, a threshold comparison value from the user; and
displaying, via the interactive user interface, information about a set of patients, the set of patients associated with comparison values that meet the threshold comparison value, and wherein the set of patients is filtered and sorted based on patient name, contact information, geographic region, and disease type.

Assignees

• IBM

Inventors

• Fink Patrick W
• Mcneil Kristin E
• Parker Philip E
• Werts David B

Classifications

• G06F40/295

  Named entity recognition · CPC title

• G16H10/20 Primary

  for electronic clinical trials or questionnaires · CPC title

• G16H10/60

  for patient-specific data, e.g. for electronic patient records · CPC title

• G06F40/268

  Morphological analysis · CPC title

• G16H40/63

  for local operation · CPC title