Developing an unbiased robust algorithm for objective diagnostic classification of most common types of dementia

Alzheimer’s disease (AD) and Alzheimer’s with cerebrovascular disease (AD-CVD) are the two most common types of dementia in elderly population. Differential diagnosis of dementia type in early stage is challenging due to overlapping symptoms and mixed etiologies. Current diagnostic techniques are invasive, expensive, or lack independent validation. An early detection of dementia type helps enabling better personalized treatments. Electrovestibulography (EVestG) showed promising preliminary results in early detection of dementia types. This proposal presents three research projects to apply advanced signal processing and machine learning techniques on EVestG data to design diagnostic classification models. This proposal will investigate supervised and unsupervised classifiers, each sensitive to AD and AD-CVD, as well as to healthy controls, then, compares them to provide a robust classifier with high predictive performance. The outcome of this work would be development of a quick, inexpensive, and objective tool with high classification power for diagnosis of dementia types using EVestG technology.

Faculty Supervisor:

Zahra Kazem-Moussavi;Brian Lithgow

Student:

Zeinab Alsadat Dastgheib

Partner:

SAM

Discipline:

Engineering - biomedical

Sector:

Professional, scientific and technical services

University:

University of Manitoba