Aplicação de aprendizagem de máquina no diagnóstico de declínio cognitivo e demência de Alzheimer baseado em testes cognitivos e marcadores genéticos

Abstract

Aid to the Diagnosis of Cognitive Decline and Alzheimer’s Dementia using Machine Learning based on Cognitive Tests and Genetic Markers A large number of solutions based on computer systems have recently been developed for the classification of cognitive abnormalities in the elderly, so that individuals at high risk of developing neurodegenerative diseases, such as Mild Cognitive Impairment (MCI) and Alzheimer’s Dementia (AD), can be Identified before the onset of disease. Several factors are related to these pathologies, making the diagnostic process a process of high complexity to be solved. This paper proposes the use of a computer model based on machine learning to perform data regression and pattern classification processes in a real database of elderly individuals. The proposal takes into account data on the gender, age, level of education of the individuals and the scores resulting from cognitive tests (Mental State Mini-Exam, Verbal and Semantic Fluency Test, Clinical Dementia Rating - And Dementia Determination Test - Ascertaining Dementia). Using nonlinear regression models, we can design classifiers to distinguish when aging is being healthy and / or pathological. The primary objective of this research is to use a regression model to analyze the data set to verify which parameters are most relevant to achieve high accuracy in the diagnosis of neurodegenerative disorders. One of the conclusions indicates that the diagnostic process based only on the results of the cognitive tests, can obtain a high rate of performance, compared to the use of all factors, including socio-cultural data. In this analysis, it is demonstrated that the use of cognitive tests produces better average values. Other analyzes were performed including genetic markers (CYP46A1 and ApoE4), without influencing the results in relation to the accuracy of the analyzes, comparing with the performance of the cognitive tests. Statistical analyzes show that the best performance in terms of sensitivity is above 97% when the settings have only cognitive tests. The approach presented can be encapsulated as a tool to support the clinical diagnostic process to identify patients with dementia or cognitive decline.