Neuropsychological and electrophysiological measurements for diagnosis and prediction of dementia: a review on Machine Learning approach

Abstract

Introduction: Emerging and advanced technologies in the field of Artificial Intelligence (AI) represent promising methods to predict and diagnose neurodegenerative diseases, such as dementia. By using multimodal approaches, Machine Learning (ML) seems to provide a better understanding of the pathological mechanisms underlying the onset of dementia. The purpose of this review was to discuss the current ML application in the field of neuropsychology and electrophysiology, exploring its results in both prediction and diagnosis for different forms of dementia, such as Alzheimer's disease (AD), Vascular Dementia (VaD), Dementia with Lewy bodies (DLB), and Frontotemporal Dementia (FTD).

Methods: Main ML-based papers focusing on neuropsychological assessments and electroencephalogram (EEG) studies were analyzed for each type of dementia.

Results: An accuracy ranging between 70 % and 90 % or even more was observed in all neurophysiological and electrophysiological results trained by ML. Among all forms of dementia, the most significant findings were observed for AD. Relevant results were mostly related to diagnosis rather than prediction, because of the lack of longitudinal studies with appropriate follow-up duration. However, it remains unclear which ML algorithm performs better in diagnosing or predicting dementia.

Conclusions: Neuropsychological and electrophysiological measurements, together with ML analysis, may be considered as reliable instruments for early detection of dementia.

Keywords: Cognitive neurorehabilitation; Dementia; Electroencephalogram; Machine learning; Mild cognitive impairment; Neuropsychology.