Web proceedings papers


Maja Nikolova , Zorica Nanovic and Sonja Gievska


Introduction. Manually diagnosing neurodegenerative disorders like probable Alzheimer’s disease and associated dementias has proven to be an arduous task. Their diagnosis is achieved (or performed) by using a combination of neuropsychological testing and particular clinical diagnostic criteria.The use of machine learning algorythms in developing an automated diagnostic model based on linguistic obtained from verbal interviews can become a crucial aid in the diagnostic process of these particular disorders. Material and methods.Based on a clinical dataset from the Dementia Bank, which includes personal and demographic information, the findings of physical and other medical examinations, and transcripts of audio recordings (I.E.interviews of each patient), we developed various machine learning models to meet that purpose..The collection of data records included 99 participant in each of the two groups, the group with probable AD and the control group. The models are based on distinct syntactic and lexical linguistic biomarkers to be able to discriminate the group of patients with probable Alzheimer’s disease from a control group.. Results and Discussion.It was shown that patients with probable Alzheimer’s disease have particulary increased their use of lexical components, while and dramatically decreasing their use of syntactic components in their speech when compared to the healthy control group. The use of machine learning algorithms to identify linguistic biomarkers in the verbal utterances of an older group of patients is an adequate and powerful tool, according to experimental and statistical evaluation, since they may help in the clinical diagnosis of probable Alzheimer’s disease.


Neurodegenerative disorders, Alzheimer’s disease, Dementia Bank, Biomarkers