Quantitative Classification of Swallowing Through Accelerometry

agerasiaetherealΤεχνίτη Νοημοσύνη και Ρομποτική

24 Νοε 2013 (πριν από 3 χρόνια και 9 μήνες)

83 εμφανίσεις

Quantitative Classification of Swallowing
T
hrough Accelerometry

Céleste Mérey (M.H.Sc. Candidate), Ervin Sejdić (Ph.D), Tom Chau (Ph.D, P.Eng)

Institute of Biomaterials and Biomedical Engineering, University of Toronto

Bloorview Research Institute,
Holland Bloorview Kids Rehabilitation Hospital


Objective and Target Population:

The purpose of this study is to investigate the potential use
of accelerometry to provide a consistent, quantitative means of monitoring swallowing in the
dysphagic pediatric
population, thus preventing aspiration.

Background:

Children with neurological and muscular impairments, such as cerebral palsy, can
experience both physical and psycho
-
social consequences as a result of the inability to swallow.
Dysphagia is not easily d
etected or observed. Current methods of evaluation are not suitable for
regular, bedside use. For example, the gold standard, videofluroscopic swallowing studies
(VFSS), requires specialized clinical experience and expensive institutionally
-
based equipment
.
Recent studies have delivered promising results in the use of cervical accelerometry to detect
abnormal swallowing. This device may offer a portable, low
-
cost and non
-
invasive method of
monitoring swallowing.

Methods:

First, data was collected from thir
ty participants with a prior diagnosis of dysphagia
from the feeding clinic at Holland Bloorview Kids Rehabilitation Hospital. In conjunction with the
VFSS, signals were collected, synchronized and stored from a dual axis accelerometer,
respiratory belt an
d head
-
mounted tri
-
axial accelerometer along with the video recording of the
x
-
ray swallowing images. Retrospective clinical evaluation of the video recordings served to
label: 1) the timestamps of the swallows in the accelerometry signals and 2) the natur
e of the
swallow: healthy or unhealthy.

Second, signal processing and analysis methods will be applied to this data. First, the
accelerometry signals were manually segmented to segregate the swallows. Pre
-
processing
techniques were applied to the swallows
to reduce the noise and disturbances in the signals.
Feature selection and classification techniques will be investigated in an iterative approach in
order to determine an optimal method for identifying a healthy versus unhealthy swallow using
accelerometr
y signals.

Results:

Current results have demonstrated that pre
-
processing techniques applied to the adult
population can be used towards signals generated by the dysphagic pediatric population.

Early analysis of the data collected from this study reveals
a signal that contains many
disturbances which convolute the ability to visually discern a swallow from other movement or
noise. Work is underway in applying classification methods in order to meet the primary
objective of detecting an unhealthy swallow.

A
nticipated Contributions:

The anticipated contributions of this thesis towards the use of
pediatric swallow accelerometry will be: (1) evaluation of previously developed accelerometric
segmentation and denoising algorithms on pediatric data; (2) the identi
fication of discriminatory
features from pediatric accelerometry signals, and (3) the development of a classification
approach for dual
-
axis pediatric cervical accelerometry
.