Procedure for extracting information from a heart sound signal
AssigneeBang & Olufsen Medicom A/S
InventorFlemming Munk
Life of this patent
Estimating the term…
Abstract
The invention relates to a procedure for extracting features from phonocardiographic signals without the use of synchronizing information from electrocardiographic signals. The features extracted are the timing and value of first and second heart sounds and various combinations of timing and value of signal components constituting heart murmur. Such combinations are directly related to various heart conditions, which are more easily diagnoseable by a medically trained person when assisted by the signal extraction according to the invention. The features are extracted by a novel combination of energy/time relationships for the heart signal and various novel classification schemes.
Claims9
A procedure for extracting information from a phonocardiographic signal obtained from a transducer and subjected to signal processing,
- characterised in that first and second heart sounds are detected and placed correctly on a time axis by performing the steps of feature extraction and classification based on the energy distribution over time in a feature time function by the steps of bandpass filtering, followed by instantaneous power and lowpass filtering of the original phonocardiographic signal.
A procedure used for extracting murmur information,
- characterised in that it comprises the following steps:
- obtaining a digital representation of heart sound for a predetermined number of seconds,
- identifying the time of occurence of the first and second heart sounds in each cycle,
- windowing the parts of heart sounds falling between the first and second heart sounds, and second and first heart sounds, respectively
- decomposition of the signals into a predetermined first number n1 of frequency bands, each band being decomposed into a predetermined second number n2 of time-slices
- obtaining a systole (SP) and a diastole (DP) power vector consisting of the sum of n1 powers measured in each of the n2 time slices
- for each combination of a frequency band and a time slice, the power values from the different systoles are compared, and the median value is chosen to be the standard value for a power vector
- obtaining a systole (SMF) and a diastole (DMF) mean frequency vector by weighting the power value for each of n1 frequency bands with the mean frequency of the corresponding band, summing the results and dividing the sum by the corresponding element in the respective systole or diastole power vector
- while using the time of occurence of the intensity vectors of the various classes for classifying the time distribution of murmurs.
Dependencies are inferred from the claim text.