Segmenting a cardiac acoustic signal
The invention relates to segmentation of cardiac acoustic signals, such as the heart sound signal, based on statistical algorithms. A duration-dependent Hidden Markov Model is disclosed which models the shifting states of the heart, based on the cardiac acoustic signal and the time spent in the given states relating to physiological events, e.g. the various states of the heart during the heart beat cycle.
1. A method of segmenting a cardiac acoustic signal, the method comprising:
obtaining the cardiac acoustic signal, the cardiac acoustic signal including at least one cardiac cycle with a plurality of states;
analyzing the cardiac acoustic signal by applying a Hidden Markov Model that takes into consideration a duration of each of the plurality of states; and
temporally segmenting the cardiac acoustic signal into a plurality of intervals, each of the plurality intervals being associated with a corresponding state of the plurality of states,
wherein a length of each of the intervals is determined using the Hidden Markov Model that takes into consideration a duration of the corresponding state with which each of the plurality of intervals is associated.
2. The method according to claim 1 , wherein the Hidden Markov Model models a duration of state of the plurality of states of the at least one cardiac cycle.
3. The method according to claim 2 , wherein the duration of state is defined by a probability distribution.
4. The method according to claim 1 , wherein the segmenting of the cardiac acoustic signal is a segmenting into segments related at least to first (S 1 ) and second (S 2 ) heart sounds.
5. The method according to claim 1 , wherein the Hidden Markov Model assumes a cyclical sequence of states with the following transition pattern: S 1 -state→silent systole (siSys) state→S 2 -state→silent diastole (siDia) state→S 1 -state.
6. The method according to claim 1 , wherein the segmenting of the cardiac acoustic signal reflects a most probable timing of a sequence of segments related at least to first (S 1 ) and second (S 2 ) heart sounds given the cardiac acoustic signal.
7. The method according to claim 1 , wherein prior to segmenting the cardiac acoustic signal, the cardiac acoustic signal is pre-processed.
8. The method according to claim 1 , wherein prior to segmenting the cardiac acoustic signal, the cardiac acoustic signal is processed to an envelope signal.
9. The method according to claim 8 , wherein the cardiac acoustic signal or the envelope signal is auto-correlated to extract the duration of each of the plurality of states.
10. The method according to claim 8 , wherein a probability distribution of a duration of state is estimated from the duration of each of the plurality of states extracted using auto-correlation.
11. The method according to claim 1 , wherein the cardiac acoustic signal is obtained in real-time, and the segmenting is performed in real-time.
12. The method according to claim 1 , wherein the segmenting of the cardiac acoustic signal is performed on a pre-recorded cardiac acoustic signal.
13. An apparatus for segmenting a cardiac acoustic signal, the apparatus comprising:
an input unit configured to receive the cardiac acoustic signal, the cardiac acoustic signal including at least one cardiac cycle with a plurality of states; and
a processing unit adapted to,
analyze the cardiac acoustic signal by applying a Hidden Markov Model that takes into consideration a duration of each of the plurality of states, and
temporally segment the cardiac acoustic signal into a plurality of intervals, each of the plurality intervals being associated with a corresponding state of the plurality of states,
wherein a length of each of the intervals is determined using the Hidden Markov Model that takes into consideration a duration of the corresponding state with which each of the plurality of intervals is associated.
14. A computer-readable medium having stored therein instructions that cause a processing unit to execute the method according to claim 1 .