Affiliations: DeustoTech-Life Unit (eVida), DeustoTech Institute of Technology, University of Deusto, Bilbao, Spain
Correspondence:
[*]
Corresponding author: Amaia Méndez Zorrilla, DeustoTech-Life Unit (eVida), DeustoTech Institute of Technology, Avda. Universidades, 24. 48007, Bilbao, Spain. E-mail:[email protected]
Abstract:
BACKGROUND: Given the importance of the voice in our daily lives,
any study focused on its pathologies and the way of caring and promoting the
health of them is of common interest. OBJECTIVE: This paper describes a method to automatically aid
indetecting vocal folds benign pathologies based on glottal space
segmentation vocal fold video sequences captured by a laryngoscope. METHODS: The proposed algorithm is based on automatic segmentation
supported by Gabor filters, and features classification with Principal
Component Analysis (PCA) to achieve the expected results. RESULTS: The authors wish to emphasize that the proposed algorithm
is independent from the images' resolution and zoom, but their quality
depends on specialist experience with the instrumentation. Segmentation
block provides good results for 95% of images and classification block
distinguishes successfully between pathological and healthy images in
92.1% of cases. The proposed system's findings have been compared with
the diagnosis made by doctors and it obtains the same results in all the 45
sequences. CONCLUSIONS: One of the proposed study's key elements has been
which objective measurements are of significance for the specialist. In this
case, it is those that enable the specialist to calculate the size of the
pathology (previously classified automatically) that he/she is observing,
thus enabling them to provide the patient with more information or to
prescribe treatment and even measure its development.
Keywords: Gabor filters, principal component analysis, vocal folds, stroboscope, glottal space
