Unit:
Κατεύθυνση Στατιστική και Επιχειρησιακή ΈρευναLibrary of the School of Science
Author:
Katachana Konstantina
Supervisors info:
Τρέβεζας Σάμης, Λέκτορας, Τμήμα Μαθηματικών, ΕΚΠΑ
Μελιγκοτσίδου Λουκία, Αναπληρώτρια Καθηγήτρια, Τμήμα Μαθηματικών, ΕΚΠΑ
Οικονόμου Αντώνης, Καθηγητής, Τμήμα Μαθηματικών, ΕΚΠΑ
Original Title:
Signal Processing and Statistical Analysis of the Cardiac Cycle via HSMMs
Translated title:
Signal Processing and Statistical Analysis of the Cardiac Cycle via HSMMs
Summary:
In this master thesis modern statistical techniques of heart sound segmentation
via hidden semi-Markov models (HSMMs) are presented. The
periodicity of the heartbeat can be broken down into cycles, each of which
consists of an S1 type sound, a systolic period, an S2 sound and a diastolic
period. Modeling this heart cycle periodicity through an HSMM gives very
important information about the health status of the heart and can be used
to detect abnormalities in its operation. Thus, in this work, appropriate
estimation and prediction techniques are presented in order to accurately
locate the fundamental heart sounds S1 and S2 which have a particular
importance in the detection of abnormalities. To achieve that, apart from
the original signal, four different signal features (Homomorphic envelogram,
Hilbert envelogram, Wavelet envelope and Power Spectral Density envelope)
are extracted and tested individually in all different modeling assumptions
under test. These are considered to be the observable output variables. The
hidden structure of the model concerns the different types of heart sounds
which alternate in time and can be considered as the states of an unobserved
semi-Markov chain. By modeling differently the state-dependent
sojourn time distributions, different HSMMs can be obtained and compared.
In this project, the Gamma distribution (as a proxy to the Negative Binomial)
and the Poisson distribution were tested and compared on their ability to
accurately identify the fundamental heart sounds S1 and S2, on the basis of
some annotated real data which serve as the reference data.
Main subject category:
Science
Keywords:
signal processing, HSMM, Cardiac cycle, sojourn time, heart sound segmentation
File:
File access is restricted only to the intranet of UoA.
Signal processing and statistical analysis of the cardiac cyle via HSMMs.pdf
2 MB
File access is restricted only to the intranet of UoA.
