Using classified text and deep learning algorithms to identify medical risk and provide early warning

• US 9,754,220 B1
• Filed: 01/23/2017
• Issued: 09/05/2017
• Est. Priority Date: 07/01/2016
• Status: Expired due to Fees

• Current Assignee: IP3 2017, Series 200 of Allied Security Trust I (Allied Security Trust)
• Original Assignee: Intraspexion, Inc.
• Inventors: Brestoff, Nelson E., Parker, Jonathan Brestoff
• Primary Examiner(s): Holmes, Michael B
• Application Number:
  US15/413,335
• Time in Patent Office: 225 Days
• Field of Search: 706 12
• CPC Class Codes: G06F 40/40 Processing or translation of natural language natural language analysis G06F40/20; semantic analysis G06F40/30
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 3/044 Recurrent networks, e.g. Hopfield networks
• CPC Class Codes: G06N 3/08 Learning methods
• CPC Class Codes: G06Q 10/0635 Risk analysis of enterprise or organisation activities
• CPC Class Codes: G16H 10/60 for patient-specific data, e.g. for electronic patient records
• CPC Class Codes: G16H 50/30 for calculating health indices; for individual health risk assessment

Abstract

Deep learning is used to identify specific, potential risks of missed diagnosis for a patient and reporting the risk to healthcare provider. The system involves mining and using existing electronic health records for specific medical diagnosis to train one or more deep learning algorithms, and then examining the internal electronic health record of the patient with the trained algorithm, to generate a scored output that will enable a healthcare provider to be alerted to potential risks of a missed diagnosis.

Claims

1. A method of using classified text and deep learning algorithms to identify medical risk and provide early warning comprising:

obtaining one or more training datasets for textual data corresponding to one or more risk classifications, wherein said risk classification comprises one or more specific medical diagnoses of interest;

training one or more deep learning algorithms using said one or more training datasets;

obtaining and indexing an internal electronic health record (EHR) of a patient;

applying said one or more deep learning algorithms to said internal EHR to identify and report any one of said one or more specific medical diagnoses of interest;

determining if said identified one of said one or more specific medical diagnoses of interest is a false positive or a true positive; and

re-training said one or more deep learning algorithms if said identified one of said one or more specific medical diagnoses of interest is a false positive.

2. The method of claim 1, wherein said one or more deep learning algorithms is a framework for natural language processing of text.

3. The method of claim 1, wherein said one or more deep learning algorithms is a recurrent neural network with a multiplicity of layers and various features, including but not limited to long short-term memory or gated recurrent units.

4. The method of claim 1, wherein the one or more deep learning algorithms have been trained with different classifiers using previously classified data sourced and provided by a subject matter expert to become models for the specific medical diagnoses of interest.

5. The method of claim 1, wherein each one of said one or more deep learning algorithms has also been trained with one or more datasets unrelated to the specific medical diagnosis of interest.

6. The method of claim 1, wherein each one of said one or more deep learning algorithms scores the EHR textual data of a patient for accuracy according to one or more deep learning models for one or more specific medical diagnoses of interest and reports said scores.

7. The method of claim 6, wherein said report comprises the result of accessing one or more databases related to said one or more specific medical diagnoses to report one or more tests or clinical features that may help to evaluate the presence, absence, or likelihood of said one or more specific medical diagnoses.

8. The method of claim 1, wherein said report comprises providing the report to one or more designated physicians, research personnel or other personnel providing care to a patient.

9. The method of claim 1, wherein said report may be limited to scores which surpass one or more specified thresholds associated with each of said one or more deep learning algorithms.

10. The method of claim 1, wherein a healthcare provider or research personnel obtains one or more of said reports through a graphical user interface.

11. The method of claim 1, wherein said one or more training datasets is obtained by mining one or more Electronic Health Records for each specific medical diagnosis.