Document Type

Article

Language

eng

Publication Date

1-17-2019

Publisher

Institute of Electrical and Electronic Engineers (IEEE)

Source Publication

2018 17th IEEE International Conference on Machine Learning and Applications (ICMLA)

Source ISSN

9781538668054

Abstract

Given the high incidence of dizziness and its frequent misdiagnosis, we aim to create a clinical support system to classify the presence or absence of benign paroxysmal positional vertigo with high accuracy and specificity. This paper describes a three-phase study currently underway for classification of benign paroxysmal positional vertigo, which includes diagnosis by a specialist in a clinical setting. Patient background information is collected by a survey on an Android tablet and machine learning techniques are applied for classification. Decision trees and wrappers are employed for their ability to provide information about the question set. One goal of the study is to attain an optimal question set. Each phase of the study presents a unique set and style of questions. Results achieved in the first two phases of the survey indicate that our approach using decision trees with filters or wrappers does a good job of identifying benign paroxysmal positional vertigo.

Comments

Accepted version. 2018 17th IEEE International Conference on Machine Learning and Applications (2018): 332-337. DOI. © 2019 Institute of Electrical and Electronic Engineers(IEEE). Used with permission.

Povinelli_13347acc.docx (866 kB)
ADA Accessible Version

COinS