Breath by breath analysis of snore sounds for the diagnosis of Apnoea
Speaker: Suren Rathnayake, ITEE
When: 2008-05-21 10:00:00
Venue: 78-622
Host: Dr. Udantha Abeyratne
Abstract:Obstructive sleep apnea-hypopnea syndrome (OSAHS) is a sleep disordered
breathing disease, which has prevalent estimates of at least 2% of the
population showing signs of which a large percentage undiagnosed. OSAHS is
associated with adverse effects on health and behavioural outcomes.
Reference standard for OSAHS diagnosis is Polysomnography, where over
fifteen signals are recorded from a subject undergoing a typical overnight
session. Positive Airway Pressure therapy is the Standard treatment to
OSAHS, which aims to alleviate obstructions in the upper airways through air
supplied with increased inspiration air pressure. Efficacy of Positive
Airway Pressure therapy is automated machines that change the pressure to
the actual needs for better patient compliance. Present machines do not
actively detect events of sleep disordered breathing before they actually
occur. Efficacy of the proposed work is to identify OSAHS events before they
occur to avoid them. We model respiratory dynamics based on dynamical
systems theory to predict signals into future. Previous work on prediction
of respiratory time series reports high mean squared errors in predictions,
attributed to distortions in measurements. Distortions such as signal
losses, artifacts, and electrical interferences are often present in
Polysomnography measurements. We identify measurement distortions and
restore data though a prediction of signals both across time and
measurements spaces. Measurements identified as reliable will be used in
developing and validating algorithms to predict respiration. Distortions
further affect diagnosis of OSAHS in both manual and automated analysis, and
also the scorer intra/inter-rater variability. Distortions and data
restoration procedure has the potential to reduce manual labour associated
in sleep scoring and decrease scorer intra/inter-rater variability.
Improvements to automated diagnostic algorithms will also be investigated
either using distortion identification and data restoration as a
pre-processing stage for present methods, or though modelling respiratory
dynamics during un-disordered breathing enabling to identify disordered
breathing.
Biography:(biography unavailable)
Type: Ph.D confirmation
Contact:Dr. Udantha Abeyratne, seminar host (udantha@itee.uq.edu.au)
or Guido Governatori (ITEE seminar co-ordinator)
(guido@itee.uq.edu.au)