A deep convolutional neural network for automated vestibular disorder classification using VNG analysis. Issue 3 (3rd May 2020)

• Record Type: Journal Article
• Authors: Ben Slama, Amine; Mouelhi, Aymen; Sahli, Hanene; Zeraii, Abderrazek; Marrakchi, Jihene; Trabelsi, Hedi
• Is Part Of: Computer methods in biomechanics and biomedical engineering. Volume 8:Issue 3(2020)
• Page Start: 334
• Page End: 342
• Publication Date: 2020-05-03
• Subjects: Nystagmus disorder -- pupil tracking -- vestibular disorder -- VNG database -- convolutional neural networks; Imaging systems in biology -- Periodicals; Imaging systems in medicine -- Periodicals; Biomechanics -- Data processing -- Periodicals; Biomedical engineering -- Periodicals; 616.0757
• DOI: 10.1080/21681163.2019.1699165
• Languages: English
• ISSNs: 2168-1163

Abstract

ABSTRACT: Dizziness is a frequent syndrome of peripheral vestibular lesions. Identification of nystagmus disorder can be a useful sign to discriminate between diverse vestibular diseases. Through the use of videonystagmography (VNG) device accomplished in the clinical practice of ENT department, the assessment of the rotation eye movement response supplies objective, consistent and precise measurements in the therapeutic scheme. In fact, vestibular dysfunctions introduce an important variety in their features which includes different complications for common VNG examination method. This work introduces a new scheme to reach the classification of eye movement signals from optokinetic and caloric VNG tests. The proposed method offers quantitative assessment and uniform characteristics of recurrent-included disease. The rotation angle of eye movements is classified into two classes of vestibular diseases by the use of deep convolutional Neural Network (CNN) technique. The proposed approach is validated on three different categories: a factual incorporated meniere, neurite and healthy subjects. The employed VNG dataset contain patients admitted into both Charles Nicolle and La Rabta hospitals of Tunis. Compared to previous works, the results demonstrate that the proposed CNN-based method is efficient in enhancing dizziness analysis.