Method and apparatus for automatic arrhythmia classification with confidence estimation

Patent metadata

Field	Value
Publication number	US-10448853-B2
Application number	US-201615095298-A
Country	US
Kind code	B2
Filing date	Apr 11, 2016
Priority date	Apr 17, 2012
Publication date	Oct 22, 2019
Grant date	Oct 22, 2019

Abstract

An arrhythmia classification system receives cardiac data from an implantable medical device, performs automatic adjudication of each cardiac arrhythmia episode indicated by the cardiac data, and generates episode data representative of information associated with the episode. The episode data include at least an episode classification resulting from the automatic adjudication of the episode and a confidence level in the episode classification. In one embodiment, the episode data further include key features rationalizing the automatic adjudication of the episode.

First claim

We claim:

1. A medical system comprising:
a medical device configured to sense one or more cardiac signals indicative of one or more arrhythmia episodes and produce cardiac data representative of the one or more cardiac signals;
an arrhythmia analysis circuit communicatively coupled to the medical device, the arrhythmia analysis circuit configured to:
receive cardiac data transmitted from the medical device;
automatically produce a plurality of voting classifications for each episode by automatically executing, in response to receiving cardiac data transmitted from the medical device, a plurality of primary adjudication algorithms, wherein each primary adjudication algorithm produces a voting classification of the plurality of voting classifications;
automatically determine, via the arrhythmia analysis circuit, an episode classification of each episode based on a majority voting of the produced plurality of voting classifications;
perform automatic episode classification of each episode of the one or more arrhythmia episodes using the determined episode classification for each episode; and
generate episode data representative of information associated with each episode including the episode classification of each episode resulting from the automatic adjudication of each episode, wherein the information includes features rationalizing the automatic episode classification of each episode;

a memory circuit communicatively coupled to the arrhythmia analysis circuit, the memory circuit configured to store data including the cardiac data, the episode data and the episode classifications; and
a user interface communicatively coupled to the arrhythmia analysis circuit and the memory circuit, the user interface including a presentation device configured to present the information associated with each episode.

2. The system of claim 1 , wherein the arrhythmia analysis circuit is configured to determine a confidence level in each episode classification by executing a plurality of adjudication algorithms, wherein the plurality of adjudication algorithms used to determine the plurality of voting classification and the plurality of adjudication algorithms used to determine a confidence level are distinct sets of adjudication algorithms.

3. The system of claim 1 , wherein the arrhythmia analysis circuit comprises an automatic arrhythmia adjudicator configured to execute the adjudication algorithms each being a machine learning algorithm implementing a tachyarrhythmia adjudication algorithm.

4. The system of claim 1 , wherein the arrhythmia analysis circuit comprises a confidence analyzer configured to determine the confidence level in each episode classification using one or more secondary adjudication algorithms to produce respective voting classifications, the confidence level indicative of a percentage of the voting classifications produced using the secondary adjudication algorithms consistent with the determined episode classification.

5. The system of claim 1 , wherein the arrhythmia analysis circuit is further configured to generate episode features including features used by the automatic arrhythmia adjudicator to determine the episode classification for each episode.

6. The system of claim 5 , wherein the arrhythmia analysis circuit is configured to generate episode features including features allowing for manual adjudication of each episode.

7. The system of claim 1 , comprising a telemetry circuit configured to receive the cardiac data from the medical device.

8. The system of claim 7 , comprising a device including the telemetry circuit, a remote device including at least the user interface, and a communication network coupling between the device and the remote device.

9. A method for classifying cardiac arrhythmias, the method comprising:
receiving cardiac data representative of the one or more cardiac signals sensed by an implantable medical device, the one or more cardiac signals indicative of one or more arrhythmia episodes;
automatically producing a plurality of voting classifications for each episode by executing, in response to the received cardiac data, a plurality of adjudication algorithms, wherein each adjudication algorithm produces a voting classification of the plurality of voting classifications;
automatically determining, by a processing device, an episode classification of each episode based on a majority voting of the produced plurality of voting classifications;
performing automatic episode classification of each episode of one or more arrhythmia episodes using the determined episode classification;
generating, by the processing device, episode data representative of information associated with each episode including the episode classifications resulting from the automatic episode classification of each episode;
generating, by the processing device, episode features used to:
determine the episode classification for each episode and
rationalize the automatic episode classification of each episode;

rationalizing the automatic episode classification of each episode using the episode features;
storing the cardiac data, the episode data, the episode features, and the episode classifications in a memory device; and
presenting the information associated with each episode.

10. The method of claim 9 , wherein the plurality of voting classifications are produced by one or more primary adjudication algorithms of a plurality of adjudication algorithms.

11. The method of claim 10 , further comprising determining a confidence level in the determined episode classification using the cardiac data by executing one or more secondary adjudication algorithms of the plurality of adjudication algorithms to produce voting classifications, the confidence level indicative of a percentage of the voting classifications produced using the one or more secondary adjudication algorithms consistent with the determined episode classification, wherein the one or more primary adjudication algorithms and the one or more secondary adjudication algorithms are distinct sets of adjudication algorithms.

12. The method of claim 10 , wherein the plurality of adjudication algorithms comprises machine learning algorithms each selected to implement a tachyarrhythmia adjudication algorithm.

13. The method of claim 11 , comprising determining an amount of machine learning algorithms to be included in the plurality of adjudication algorithms based on one or more estimated measures of accuracy in the automatic adjudication and an estimated potential need for manual adjudication by a user.

14. The method of claim 12 , comprising determining the amount of machine learning algorithms to be included in the plurality of adjudication algorithms based on one or more factors selected from:
a risk factor being a proportion of arrhythmia episodes with classification identified as high confidence classifications that are misclassified during the automatic adjudication to a total number of classified arrhythmia episodes;
a specificity factor being a proportion of misclassified arrhythmia episodes with classifications identified as low confidence classifications to the total number of classified arrhythmia episodes; and
a service burden factor being a proportion of arrhythmia episodes with classification identified as low confidence classifications to the total number of classified arrhythmia episodes.

15. The method of claim 9 , comprising generating episode features including features used to determine the episode classification during the automatic adjudication of each episode.

16. The method of claim 15 , comprising ge

Assignees

• Cardiac Pacemakers Inc

Inventors

• Mahajan Deepa
• Pulliam Christopher
• Dong Yanting
• Perschbacher David L

Classifications

• G16H40/60

  for the operation of medical equipment or devices · CPC title

• G16H50/70

  for mining of medical data, e.g. analysing previous cases of other patients · CPC title

• G16H50/20

  for computer-aided diagnosis, e.g. based on medical expert systems · CPC title

• G16H50/30

  for calculating health indices; for individual health risk assessment · CPC title

• A61B5/7275

  Determining trends in physiological measurement data; Predicting development of a medical condition based on physiological measurements, e.g. determining a risk factor · CPC title