Hypothesis-driven, real-time analysis of physiological data streams using textual representations

Patent metadata

Field	Value
Publication number	US-9292576-B2
Application number	US-201213570680-A
Country	US
Kind code	B2
Filing date	Aug 9, 2012
Priority date	Aug 9, 2012
Publication date	Mar 22, 2016
Grant date	Mar 22, 2016

Abstract

A method of analyzing physiological data streams. According to the method, physiological data is received into a computerized machine. The physiological data comprises numerical data and medical symptoms of a patient. Features are extracted from the physiological data based on development of the physiological data over a period of time. The features are converted into a textual representation using natural language generation. Input terms for an information retrieval system operating on the computerized machine are automatically generated based on the features. The input terms are input to the information retrieval system. A corpus of data is automatically searched to retrieve results to the input terms using the information retrieval system.

First claim

What is claimed is:

1. A method comprising:
receiving physiological data, associated with a patient having an unknown medical condition, into a computerized device, said physiological data comprising streams of medical data obtained by monitoring said patient and medical symptoms reported by said patient;
extracting numerical data from said physiological data based on development of said streams of medical data over a period of time, using said computerized device;
extracting features from said numerical data, using said computerized device, said features comprising a property of said physiological data being observed;
converting said features extracted from said physiological data into a textual representation using natural language generation, using said computerized device, said natural language generation converting said features into words to use as a query for an information retrieval system operating on said computerized device;
automatically generating said query for said information retrieval system based on said textual representation and said medical symptoms reported by said patient, using said computerized device, at least one said query being generated for each feature extracted from said physiological data;
inputting said query to said information retrieval system, using said computerized device;
automatically searching a corpus of data to retrieve results to said query, using said information retrieval system, said corpus of data comprising structured and unstructured data, said results indicating possible medical conditions of said patient in said period of time;
comparing said results obtained from said information retrieval system to said medical symptoms reported by said patient, using said computerized device;
generating hypotheses related to said possible medical conditions of said patient based on said comparing said results obtained from said information retrieval system, using said computerized device; and
recommending a medical test or analysis based on said hypotheses to confirm said possible medical conditions of said patient, using said computerized device.

2. The method according to claim 1 , further comprising:
ranking said results obtained from said information retrieval system based on importance scores retrieved with said results, using said computerized device;
formulating said hypotheses related to said possible medical conditions of said patient to generate analytical patterns, using said computerized device; and
identifying relationships among said physiological data based on said analytical patterns, using said computerized device.

3. The method according to claim 1 , further comprising generating confidence scores for said hypotheses, using said computerized device.

4. The method according to claim 3 , further comprising ranking said hypotheses according to said confidence scores for said hypotheses and recommending additional analyses based on rank order of said hypotheses, using said computerized device.

5. The method according to claim 3 , further comprising:
storing, in a non-transitory storage medium, a history of said results to said query, said hypotheses, and said confidence scores for said hypotheses, using said computerized device; and
correlating previously generated hypotheses and corresponding confidence scores stored in said non-transitory storage medium with at least one analysis based on said hypotheses, using said computerized device.

6. The method according to claim 1 , further comprising displaying said results to said query and a link to said corpus of data indicating how said corpus of data contributed to said results on a user interface, using said computerized device.

7. A method comprising:
extracting features from physiological data associated with a patient having an unknown medical condition using a computerized device, said features being based on development of said physiological data over a period of time, said physiological data comprising streams of medical data obtained by monitoring said patient and medical symptoms reported by said patient;
generating at least one query based on said features, using said computerized device, at least one said query being generated for each feature;
inputting said at least one query to a textual query engine operating on said computerized device, said textual query engine automatically searching a corpus of data comprising sources of structured and unstructured data;
retrieving results to said at least one query using said textual query engine operating on said computerized device, said results indicating a possible medical condition of said patient in said period of time;
comparing said results to said medical symptoms reported by said patient;
generating hypotheses related to said possible medical condition of said patient based on said results obtained from said textual query engine, using said computerized device; and
translating said hypotheses into at least one additional analysis based on said possible medical condition to confirm said possible medical condition of said patient, using said computerized device.

8. The method according to claim 7 , further comprising:
ranking said results obtained from said textual query engine based on importance scores retrieved with said results, using said computerized device;
formulating said hypotheses related to said possible medical condition of said patient to generate analytical patterns, using said computerized device; and
identifying relationships among said physiological data based on said analytical patterns, using said computerized device.

9. The method according to claim 7 , said textual query engine comprising one of a search engine and a question answering (QA) system.

10. The method according to claim 7 , said results from said textual query engine comprising documents, web pages, and other text-based knowledge representations.

11. The method according to claim 7 , further comprising generating confidence scores for said hypotheses related to said possible medical condition of said patient, using said computerized device.

12. The method according to claim 11 , further comprising:
storing, in a non-transitory storage medium, a history of said at least one query, said results, said hypotheses, and said confidence scores for said hypotheses, using said computerized device; and
correlating previously generated hypotheses and corresponding confidence scores stored in said non-transitory storage medium with said at least one additional analysis based on said hypotheses, using said computerized device.

13. The method according to claim 7 , further comprising outputting said at least one additional analysis using a user interface on said computerized device.

14. The method according to claim 13 , further comprising displaying said results and a link to a corpus of data indicating how said corpus of data contributed to said hypotheses on said user interface.

15. A method comprising:
receiving streams of physiological data associated with a patient having an unknown medical condition, into a computerized device, said physiological data comprising numerical data and medical symptoms reported by said patient;
converting time series data from said streams of physiological data into words based on natural language generation, using said computerized device;
formulating textual queries from said words and said medical symptoms reported by said patient, using said computerized device, at least one query being generated for each feature extracted from said streams of physiological data;
outputting said textual queries to external source

Assignees

• Biem Alain E
• Dinger Timothy R
• Halim Nagui
• Luo Gang
• Sow Daby M
• Turaga Deepak S
• IBM

Inventors

• Biem Alain E
• Dinger Timothy R
• Halim Nagui
• Luo Gang
• Sow Daby M
• Turaga Deepak S

Classifications

• G06F16/24568

  Data stream processing; Continuous queries · CPC title

• G06N5/02

  Knowledge representation; Symbolic representation · CPC title

• G16H70/20

  relating to practices or guidelines · CPC title

• G06F19/345

  Physics · mapped topic

• G06F17/30516 Primary

  Physics · mapped topic