Cardiac Signal Template Generation Using Waveform Clustering

Abstract

Cardiac devices and methods using cardiac waveform clustering for template generation are described. A method of characterizing a cardiac response involves delivering pacing pulses to heart, the pulses having an energy greater than a capture threshold. Cardiac signals are sensed following the pulses. Cardiac signal characteristics, waveforms, and/or features are clustered into a plurality of clusters. A cardiac response template is formed using one or more of the plurality of clusters.

1 . A method of operating an implantable device to characterize a cardiac response to pacing, comprising:
 delivering pacing pulses to a heart sufficient in energy to effect capture;   sensing cardiac signals respectively following delivery of the pacing pulses;   extracting one more features from the cardiac signals;   clustering the features to define a plurality of clusters;   identifying one or more predominant clusters of the plurality of clusters and one or more less predominant clusters of the plurality of clusters; and   forming a captured response template using the predominant clusters.   
   
   
       2 . The method of  claim 1 , wherein the one or more features comprises one or more cardiac signal peaks. 
   
   
       3 . The method of  claim 1 , wherein clustering the one or more features of the cardiac signals comprises clustering the one or more features in batch mode. 
   
   
       4 . The method of  claim 1 , wherein clustering the one or more features of the cardiac signals comprises clustering the one or more features beat-by-beat. 
   
   
       5 . The method of  claim 1 , wherein clustering the features comprises clustering each feature of a cardiac signal independently from clustering other features of the cardiac signal. 
   
   
       6 . The method of  claim 1 , wherein forming the captured response template comprises forming one or more detection windows of the captured response template, each detection window having maximum and minimum time limits and maximum and minimum amplitude limits. 
   
   
       7 . The method of  claim 1 , further comprising forming a fusion template using the one or more less predominant clusters. 
   
   
       8 . The method of  claim 1 , further comprising forming an intrinsic activation template using the one or more less predominant clusters. 
   
   
       9 . The method of  claim 1 , further comprising forming a noise template using the one or more less predominant clusters. 
   
   
       10 . The method of  claim 1 , wherein at least one of the predominant clusters used to form the captured response template includes features extracted from cardiac signals of fusion beats. 
   
   
       11 . The method of  claim 1 , wherein clustering the features comprises clustering the features using one or both of K-means clustering and a self-organizing map algorithm. 
   
   
       12 . The method of  claim 1 , further comprising storing additional information related to one or more of pacing energy, pacing rate, and pacing interval with the captured response template. 
   
   
       13 . The method of  claim 1 , wherein:
 forming the captured response template comprises determining detection windows based on the cardiac feature clusters; and   further comprising adapting the detection windows to accommodate changes in a patient's cardiac waveform morphology.   
   
   
       14 . The method of  claim 13 , wherein adapting the detection windows comprises adapting the detection windows using features of a cardiac signal associated with a captured response. 
   
   
       15 . The method of  claim 13 , wherein adapting the detection windows comprises changing at least maximum amplitude of a detection window. 
   
   
       16 . A method of operating an implantable device to determine a template used to characterize a cardiac response to pacing, comprising:
 delivering pacing pulses to a heart sufficient in energy to effect capture;   sensing cardiac signals respectively following delivery of the pacing pulses;   extracting peak values from the cardiac signals;   clustering the peak values to define a plurality of clusters;   identifying one or more predominant clusters of the plurality of clusters and one or more less predominant clusters of the plurality of clusters; and   forming a captured response template using the predominant clusters.   
   
   
       17 . The method of  claim 16 , wherein:
 extracting the peak values comprises extracting positive peak values and negative peak values;   clustering the peak values to define a plurality of clusters comprises:
 clustering the positive peak values to define a plurality of positive peak clusters; and 
 clustering the negative peak values to define a plurality of negative peak clusters; 
 identifying the one or more predominant clusters comprises: 
 identifying one or more predominant positive peak clusters; and 
   identifying one or more predominant negative peak clusters; and   forming the captured response template comprises forming the captured response template using the one or more predominant positive peak clusters and the one or more predominant negative peak clusters.   
   
   
       18 . The method of  claim 17 , wherein forming the captured response template comprises:
 forming a positive peak detection window using the positive peak clusters; and   forming a negative peak detection window using the negative peak clusters.   
   
   
       19 . The method of  claim 18 , wherein the positive peak detection window has a first shape and the negative peak detection window has a second shape, different from the first shape. 
   
   
       20 . The method of  claim 16 , further comprising forming one or more of a fusion template, an intrinsic activation template, and a noise template using the one or more less predominant clusters.