Apparatus for analyzing cardiac events
An apparatus for detecting cardiac events in electrograms has a feature extraction unit that derives features of the cardiac events for discriminating among different types of cardiac events. A clustering unit groups cardiac events with similar features into respective clusters defined by predetermined cluster features. The feature extraction unit determines a feature vector describing waveform characteristics of cardiac events in the electrogram by a wavelet transform. The clustering unit determines the distance between the feature vector and corresponding cluster feature vectors in order to assign the cardiac event in question to that cluster which results in a minimum distance, providing that the minimum distance is less than a predetermined threshold value. A heart stimulator provided with such a cardiac event detecting apparatus detects the occurrence of an arrhythmia and appropriately controls a pulse stimulator to treat the arrhythmia.
1. An apparatus for analyzing cardiac events, comprising:
a feature extraction unit, supplied with an electrocardiogram, said feature extraction unit configured for deriving features of cardiac events from said electrocardiogram and determining a feature vector, by a wavelet transform operating on said electrogram, describing waveform characteristics of cardiac events in said electrogram; and
a clustering unit provided with said feature vector, said clustering unit configured for determining a distance between said feature vector and corresponding cluster feature vectors and assigning a cardiac event in said electrogram to a particular cluster that results in a minimum distance, as long as said minimum distance is less than a predetermined threshold value, and generating an output signal that identifies the particular cluster to which the cardiac event has been assigned.
2. An apparatus as claimed in claim 1 wherein each of said clusters is defined by a cluster center μ i and a co-variance matrix Σ i for the cluster features of that cluster, and wherein said clustering unit determines a distance function D2 between each event feature vector and said cluster center μ i .
3. An apparatus as claimed in claim 2 wherein said clustering unit calculates said distance using Mahalanobis distance criterion.
4. An apparatus as claimed in claim 2 wherein said clustering unit determines said minimum distance by a grid search over a duration of the cardiac event.
5. An apparatus as claimed in claim 1 comprising an integrator that integrates said distance over a predetermined period of time.
6. An apparatus as claimed in claim 1 wherein said clustering unit updates said cluster feature according to a predetermined rule dependent on said minimum distance.
7. An apparatus as claimed in claim 1 wherein said clustering unit generates a new cluster if said minimum distance exceeds said predetermined threshold value, by setting features for said new cluster equal to the event features of the cardiac event that resulted in said minimum distance exceeding said predetermined threshold value.
8. An apparatus as claimed in claim 1 wherein said clustering unit terminates clusters that fail to have a predetermined number of cardiac events grouped therein within a predetermined time period.
9. An apparatus as claimed in claim 1 wherein said clustering unit performs a likelihood-based search sequence over said clusters to determine said minimum distance.
10. An apparatus as claimed in claim 1 wherein said clustering unit determines said minimum distance by a grid search only over clusters in which cardiac events have been grouped within a duration of the cardiac event.
11. An apparatus as claimed in claim 1 wherein said clustering unit calculates a distance of a cardiac event in question from a cluster in which a cardiac event was previously grouped.
12. An apparatus as claimed in claim 1 comprising a classifier that associates said clusters respectively with different cardiac rhythms according to predetermined rules.
13. A heart stimulator comprising:
a pulse generator configured to interact with a subject to deliver stimulation pulses to the subject;
an apparatus that analyzes cardiac events comprising a feature extraction unit, supplied with an electrocardiogram, said feature extraction unit configured for deriving features of cardiac events from said electrocardiogram and determining a feature vector, by a wavelet transform operating on said electrogram, describing waveform characteristics of cardiac events in said electrogram, and a clustering unit provided with said feature vector, said clustering unit configured for determining a distance between said feature vector and corresponding cluster feature vectors and assigning a cardiac event in said electrogram to a particular cluster that results in a minimum distance, as long as said minimum distance is less than a predetermined threshold value; and
an arrhythmia detection and control unit connected to said pulse generator that controls emission of stimulation pulses from said pulse generator dependent on detection of an arrhythmia dependent on the cluster in which the cardiac event is grouped.