Apparatus and methods for analysis of cardiac device stored episodes containing sensed signals and waveforms

• US 7,263,399 B2
• Filed: 04/29/2002
• Issued: 08/28/2007
• Est. Priority Date: 04/29/2002
• Status: Active Grant

• Current Assignee: Medtronic Incorporated (Medtronic Plc)
• Original Assignee: Medtronic Incorporated (Medtronic Plc)
• Inventors: Carlson, Dan B.
• Primary Examiner(s): Bockelman; Mark
• Application Number:
  US10/135,910
• Publication Number:
  US 20030204146A1
• Time in Patent Office: 1,947 Days
• Field of Search: 600/509, 600/510
• US Class Current: 600/509
• CPC Class Codes: A61B 5/316 Modalities, i.e. specific diagnostic methods
• CPC Class Codes: A61B 5/318 Heart-related electrical modalities, e.g. electrocardiography [ECG]
• CPC Class Codes: A61B 5/349 Detecting specific parameters of the electrocardiograph cycle
• CPC Class Codes: A61B 5/7232 involving compression of the physiological signal, e.g. to extend the signal recording period

Abstract

Apparatus and methods for analyzing implantable medical device (IMD) data and to distinguish specific morphologies for review. EGM waveforms of a cardiac cycle can be grouped according to waveform shape similarities, preferably accomplished using clustering algorithms. The EGM waveform can be divided into segments represented by a one dimensional vector for each segment, and nearest neighbor vector clustering used to group the waveforms according to morphological similarity. The waveforms can be further grouped at a second level into chronologically contiguous similar EGM waveforms. Interval data for a cardiac cycle may also be grouped according to interval similarity and chronology. The chronologically contiguous waveform and interval groupings may be further grouped. Large quantities of EGM data are automatically analyzed and annotated to focus the attention of a cardiologist on likely areas of concern, reducing the need to visually inspect endless streams of EGM waveforms.

Claims

1. A method for analyzing stored cardiac EGM waveform data for a plurality of a first heart chamber cycles, the data including a plurality of waveform cycle segments, the data initially originating from an implanted device, the method comprising:

grouping the waveform cycle segments into a first set of cycle groups as a function of waveform shape similarity between the waveform cycle segments, wherein each of the first set group members have similar waveform shapes to the other members of the same first set group;

wherein the grouping includes clustering the waveform cycle segments based on waveform shape similarity;

wherein the waveform cycle segments are treated as vectors; and

wherein the clustering clusters includes using the Euclidean distance between the vectors.

2. A method as in claim 1, wherein the clustering includes nearest neighbor clustering.

3. A method as in claim 1, further comprising:

grouping the waveform cycle segments into a second set of cycle groups, wherein the second set group members are substantially chronologically contiguous to the other second set group members and are members of the same first set group.

4. A method as in claim 3, further comprising processing the first set groups by executing one of a step of marking, storing, displaying, and outputting the group memberships, and combinations thereof.

5. A method for analyzing stored cardiac EGM data from an implantable device, the method comprising:

retrieving the stored data, wherein the data includes EGM waveform data;

dividing the EGM waveform data into a plurality of cardiac cycle waveform segments, wherein each heartbeat waveform segment is associated with a heartbeat; and

grouping the EGM cardiac cycle waveform segments into waveform identity groups as a function of waveform shape similarities between the waveform segments with members of the same waveform identity group having similar waveform shapes to the other waveform identity group members;

wherein the grouping includes calculating a waveform shape distance between the cardiac cycle waveform segments and clustering the cardiac cycle waveform segments as a function of the shape distance between the heartbeat waveform segments;

wherein each cardiac cycle waveform segment is represented as a vector and the shape distance is a distance between the vectors; and

wherein the vector distance is a Euclidean distance.

6. A method as in claim 5, further comprising marking the EGM heartbeat waveform segments with an indicia of identity group membership and storing the marked EGM identity waveform segments in nonvolatile storage.

7. A method as in claim 5, further comprising the step of visually displaying the EGM heartbeat waveform segments together with a visual indicia of the waveform identity group membership for each EGM heartbeat waveform segment.

8. A method as in claim 5, further comprising collecting temporally adjacent heartbeat waveform segments having the same heartbeat waveform identity group membership into chronologically contiguous similar cardiac cycle waveform segments.

9. A method as in claim 8, wherein the temporally adjacent cardiac cycle waveform segments are located at a distance at no more than n cardiac cycles from each other, wherein n is selected from a group consisting of 1, 2, and 3 cardiac cycles, wherein n=0 indicates perfect adjacency.

10. A method as in claim 8, further comprising marking the collections with indicia of collection membership and storing the marked collections to nonvolatile storage.

11. A method as in claim 8, further comprising visually displaying the collections with a visual indication of cardiac cycle waveform segment collection membership.

12. In a computer, having a data storage device, a display device, and a processor for executing a program, a method for preprocessing stored cardiac data originating from an implantable device, the method comprising:

retrieving the stored data from the data storage device, wherein the data includes EGM waveform segment data for a plurality of beats for a first heart chamber;

clustering the EGM waveform segments into morphological similarity groups having similar morphology with other members of the same group; and

organizing morphologically similar and substantially chronologically contiguous cardiac cycle segments into collections of similar and substantially chronologically contiguous cycles

wherein the retrieved EGM waveform data is stored as a vector of wave amplitudes for each cardiac cycle segment;

wherein the clustering includes clustering the EGM waveform vectors using nearest neighbor vector clustering; and

wherein the clustering utilizes Euclid ian distance between the EGM waveform vectors to cluster the EGM waveforms.