Reduction of classification error rates and monitoring system using an artificial class

• US 7,840,507 B2
• Filed: 04/08/2010
• Issued: 11/23/2010
• Est. Priority Date: 02/21/2007
• Status: Active Grant

• Current Assignee: Cyberonics, Inc
• Original Assignee: NeuroVista Corporation
• Inventors: Snyder, David
• Primary Examiner(s): Holmes; Michael
• Application Number:
  US12/756,953
• Publication Number:
  US 20100198763A1
• Time in Patent Office: 229 Days
• Field of Search: 706/20
• US Class Current: 706/20
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G16H 40/63 for local operation

Abstract

Systems and methods for enhancing the accuracy of classifying a measurement by providing an artificial class. Seizure prediction systems may employ a classification system including an artificial class and a user interface for signaling uncertainty in classification when a measurement is classified in the artificial class.

Claims

1. A seizure advisory system comprising a user interface, the user interface comprising:

an output interface configured to provide an indication corresponding to a result of classification of data from a subject; and

an output interface configured to provide an indication of uncertainty when a result of classification of data from a subject is indicative of uncertainty in the classification.

2. A system as recited in claim 1, wherein the classification of data from a subject comprises a plurality of classes corresponding to neurological conditions of the subject.

3. A system as recited in claim 1, wherein the classification of data from a subject comprises training a classifier using unsupervised learning on a training set of data from a subject.

4. A system as recited in claim 2, wherein the classes corresponding to neurological conditions of the subject comprise a class corresponding to a heightened susceptibility of the subject to a seizure.

5. A system as recited in claim 2, wherein the classes corresponding to neurological conditions of the subject comprise a class corresponding to a reduced susceptibility of the subject to a seizure.

6. A system as recited in claim 2, further comprising:

a body configured to generate a signal indicative of the result of classification of data from the subject.

7. A system as recited in claim 6, wherein the body is a handheld device.

8. A system as recited in claim 6, wherein the body is configured to be implanted in the subject.

9. A seizure advisory system comprising:

at least one sensor input that receives at least one electrical signal indicative of brain activity;

a processing unit configured to determine a measurement feature vector based at least in part on the at least one electrical signal;

a classifier comprising a plurality of empirical classes, each empirical class associated with at least one feature vector from a training set of feature vectors and being associated with an estimated neural condition and an other class independent of each empirical class; and

an output device for communicating information related to a result of applying the classifier to the measurement feature vector.

10. A seizure advisory system as recited in claim 9, wherein the output device is configured to:

provide a signal indicative of an estimated neural condition in response to a result of applying the classifier to the measurement feature vector being a class associated with the estimated neural condition; and

provide a signal indicative of uncertainty in estimating a neural condition in response to a result of applying the classifier being the other class.

11. A seizure advisory system as recited in claim 10, wherein the output device, is configured to provide a seizure warning in response to the result of applying the classifier being a class identified with a pre-ictal condition or a class identified with a pro-ictal condition.

12. A seizure advisory system as recited in claim 10, wherein the other class is associated with a plurality of artificially generated feature vectors having a specified distribution.