Identifying errors in medical data

Patent metadata

Field	Value
Publication number	US-9754076-B2
Application number	US-201514860834-A
Country	US
Kind code	B2
Filing date	Sep 22, 2015
Priority date	Jul 23, 2015
Publication date	Sep 5, 2017
Grant date	Sep 5, 2017

Abstract

A computer processor may receive medical data including a report and an image. The computer processor may analyze the report using natural language processing to identify a condition and a corresponding criterion. The computer processor may also analyze the image using an image processing model to generate an image analysis. The computer processor may determine whether the report has a potential problem by comparing the image analysis to the criterion.

First claim

What is claimed is:

1. A computer implemented method for identifying errors in medical data, the method comprising:
receiving medical data comprising a report and an image;
analyzing the report, by a processor, using natural language processing (NLP) to identify a condition and a criterion, wherein the condition is a medical condition, and wherein the criterion includes diagnostic information that corresponds to the condition;
generating an image analysis, by the processor, by analyzing the image using an image processing model; and
determining whether the report has a potential problem by comparing at least the criterion to the image analysis.

2. The method of claim 1 , wherein there are a plurality of image processing models, each image processing model of the plurality of image processing models corresponding to a group of medical conditions, and wherein the image processing model used to generate the image analysis is selected based on the condition.

3. The method of claim 1 , the method further comprising:
providing, in response to determining that the report has a potential problem, a notification indicating the potential problem.

4. The method of claim 1 , wherein the report has a potential problem due to an optical character recognition (OCR) error.

5. The method of claim 1 , wherein the report has a potential problem due to an optical word recognition error.

6. The method of claim 1 , the method further comprising performing, prior to analyzing the report using natural language processing, optical character recognition on the report.

7. The method of claim 2 , wherein the plurality of image processing models includes a fracture analysis model and a tumor analysis model.

8. The method of claim 7 , wherein the fracture analysis model is configured to analyze X-Ray images to identify a presence, size, type, and location of bone fractures, and wherein the tumor analysis model is configured to analyze computerized axial tomography (CAT) scan images to identify the size and location of tumors.

9. The method of claim 7 , wherein the plurality of image processing models further include a model for detecting signs of diabetes, a model for detecting problems with the structure of a heart, a model for detecting torn ligaments, and a model for detecting bulging discs.

10. The method of claim 1 , wherein the image analysis includes an image condition and an image criterion, wherein the image condition is a medical condition identified in the image, and the image criterion is diagnostic information about the image condition.

11. The method of claim 10 , wherein determining whether the report has a potential problem includes comparing the condition to the image condition and the criterion to the image criterion.

12. The method of claim 1 , wherein the criterion includes a severity of the condition.

13. The method of claim 1 , wherein the condition is a bone fracture, and wherein the criterion is a Gustilo grade of the bone fracture.

14. The method of claim 1 , the method further comprising:
transmitting, in response to determining that the report contains a potential problem, a notification and the report to a user.

15. The method of claim 1 , wherein the condition is a bone fracture, wherein the criterion is a size of the bone fracture in the report, wherein the image analysis includes a size of the bone fracture identified in the image, and wherein the determining whether the report has a potential problem includes:
determining a difference between the size of the bone fracture identified from the report and a size of the bone fracture identified in the image;
comparing the difference to a threshold;
determining that the difference is greater than the threshold; and
determining, based on the difference being greater than the threshold, that the report has a potential problem.

16. The method of claim 1 , the method further comprising:
converting, prior to the analyzing the report using natural language processing, the report into machine-encoded text using optical character recognition.

17. A computer implemented method for identifying errors in medical data, the method comprising:
receiving medical data of a patient, the medical data comprising a medical report and an image;
converting the report into machine-encoded text using optical character recognition;
analyzing, by a processor, the converted report using natural language processing (NLP) to identify a condition and a criterion, wherein the condition is a medical condition, and wherein the criterion includes a severity that corresponds to the condition;
determining, based on the identified condition and from a plurality of image processing models, a particular image processing model that corresponds to the identified condition, wherein the plurality of image processing models includes a fracture analysis model and a tumor analysis model;
generating, by the processor, an image analysis by analyzing the image using the particular image processing model, the image analysis including a an image condition and an image criterion, wherein the image condition is a medical condition identified in the image, and the image criterion is a severity of the image condition;
determining that the condition and the image condition match;
comparing, in response to determining that the condition and the image condition match, the criterion to the image criterion;
determining that the criterion and the image criterion do not match; and
transmitting, in response to determining that the criterion and the image criterion do not match, a notification that the report has a potential problem along with the report to a user.

Assignees

• IBM

Inventors

• Biegert Keith P
• Bull Brendan C
• Contreras David
• Sizemore Robert C
• Smith Sterling R

Classifications

• G06V30/262

  using context analysis, e.g. lexical, syntactic or semantic context · CPC title

• G06F18/256

  of results relating to different input data, e.g. multimodal recognition · CPC title

• G06F18/22

  Matching criteria, e.g. proximity measures · CPC title

• G06F40/279

  Recognition of textual entities · CPC title

• G06F40/30

  Semantic analysis · CPC title