System and method for inferring disease similarity by shape matching of ECG time series

• US 7,783,342 B2
• Filed: 04/21/2008
• Issued: 08/24/2010
• Est. Priority Date: 04/21/2008
• Status: Expired due to Fees

• Current Assignee: LinkedIn Corporation (Microsoft Corporation)
• Original Assignee: International Business Machines Corporation
• Inventors: Syeda-Mahmood, Tanveer F., Wang, Fei
• Primary Examiner(s): Schaetzle; Kennedy J
• Assistant Examiner(s): ANTHONY, JESSICA LYNN
• Application Number:
  US12/106,514
• Publication Number:
  US 20100081953A1
• Time in Patent Office: 855 Days
• Field of Search: 600508-510, 600/513, 600515-518, 600/523, 600/525, 128/920, 128/923, 128/925
• US Class Current: 600/523
• CPC Class Codes: A61B 5/35 by template matching
• CPC Class Codes: G06F 2218/16 by matching signal segments
• CPC Class Codes: Y10S 128/923 by comparison of patient data to other data

Abstract

A method for inferring disease similarity by similarity retrieval of electrocardiogram time-series, comprising: acquiring user ECG waveforms correspondingly depicting many cardiac cycles of the heart of many users stored in a database; pre-processing each of the user ECG waveforms through pre-processing steps to isolate sets of single cardiac cycles corresponding to different heart-rates detected for each of the user ECG waveforms, each single cardiac cycle within the many cardiac cycles of the heart of many users corresponds to one single heart-rate detected. acquiring patient ECG waveforms depicting multiple cardiac cycles of the heart of a query patient; pre-processing the patient ECG waveforms through pre-processing steps to isolate sets of single cardiac cycles corresponding to different heart-rates detected for each of the patient ECG waveforms of the query patient, each single cardiac cycle within the multiple cardiac cycles of the heart of the query patient corresponds to one single heart-rate detected.

Claims

1. A method for inferring disease similarity by shape matching electrocardiogram (ECG) time series, the method comprising:

acquiring a plurality of user ECG waveforms correspondingly depicting many cardiac cycles of the heart of many users stored in a database;

pre-processing each of the plurality of user ECG waveforms through a first plurality of pre-processing steps to isolate sets of single cardiac cycles corresponding to different heart rates detected for each of the plurality of user ECG waveforms, each single cardiac cycle within the many cardiac cycles of the heart of many users corresponds to one single heart rate detected;

acquiring one or more patient ECG waveforms depicting multiple cardiac cycles of the heart of a query patient;

pre-processing the one or more patient ECG waveforms through a second plurality of pre-processing steps to isolate sets of single cardiac cycles corresponding to different heart rates detected for each of the one or more patient ECG waveforms of the query patient, each single cardiac cycle within the multiple cardiac cycles of the heart of the query patient corresponds to one single heart rate detected; and

successively comparing each single cardiac cycle of the one or more patient ECG waveforms to each single cardiac cycle of the plurality of user ECG waveforms stored in the database based on a non-rigid shape matching technique that maximizes the match of the one or more patient ECG waveforms while allowing for gaps and insertions labels; and

further comprising;

recovering disease labels of one or more of the plurality of diseases from one or more of the plurality of user ECG waveforms that closely matches with the patient ECG waveform based on the comparison; and

generating a graph displaying a distribution of labels as a statistical report based on the matching.

2. The method as in claim 1, wherein the first plurality of pre-processing steps comprises:

obtaining the plurality of user ECG waveforms through multiple channels;

removing the noise in each of the plurality of user ECG waveforms without losing significant signal changes;

segmenting the plurality of user ECG waveforms into single cardiac cycles and determining the corresponding heart rates detected in each of the plurality of user ECG waveforms;

extracting a first plurality of fiducial shape features from each cardiac cycle isolated from the plurality of user ECG waveforms; and

extracting a first plurality of shape properties from the first plurality of fiducial features and storing the first plurality of shape properties in the database.

3. The method as in claim 1, wherein the second plurality of pre-processing steps comprises:

obtaining the one or more patient ECG waveforms through multiple channels;

removing the noise in the one or more patient ECG waveforms without losing significant signal changes;

segmenting the one or more patient ECG waveforms into single cardiac cycles and determining the corresponding heart rates detected in the one or more patient ECG waveforms;

extracting a second plurality of fiducial shape features from the one or more patient ECG waveforms; and

extracting a second plurality of shape properties from the second plurality of fiducial shape features.

4. The method as in claim 1, further generating a list of one or more of the plurality of user ECG waveforms that closely matches with the patient ECG waveform based on the comparison in a descending order of similarity.