EEG based dementia diagnosis using multi-class support vector machine with motor speed cognitive test

Highlights

• An experimental protocol is introduced for clinical setup for dementia diagnosis.

• An efficient set of EEG features classify dementia in different cognitive states.

• Higher frequencies were prominent in the dementia group during cognitive events.

• Motor speed test (MST) outperform over resting state and relaxing state EEG.

• Classifier model achieved maximum 92.36% accuracy of dementia diagnosis with MST.

Abstract

Dementia is the most burdensome disorder in elders. The Dementia diagnosis is the challenging task at the earliest stages of a neurodegenerative disease when cognitive decline does not interfere with daily life activities. This work focuses on the detection of mild cognitive impairment (MCI) by classifying dementia, MCI and age-matched normal subjects. The classification is based on a different set of EEG features. The multi-class support vector machine (SVM) used to classify EEG features during resting-state, relaxing-state, and motor speed test (MST) events. This work investigated the efficient set of EEG features to calculate maximum classification accuracy for each cognitive event. The motor speed of subjects evaluated and correlated the difference between the dominant and non-dominant hand reactivity with ageing in MST event. The proposed work achieved the highest overall accuracy of 87.59% of MST event after 85.09% in relaxing-state and 80.10% in resting-state. The diagnostic accuracy of MCI group achieved 87.22% in resting-state, 89.72% in relaxing state, and 91.23% in MST. Similarly, dementia achieved 88.72% accuracy in the resting state, 90.23% in relaxing state, and 92.36% in MST event. The normal group achieved 94.66% accuracy in the resting state, 90.23% in relaxing state, and 91.60% in MST event. These findings are comparatively higher than the latest research in this area, and MST findings are novel using multi-class SVM. Thus, MST is the most reliable tool for dementia diagnosis in the clinical setting.

Introduction

Mild cognitive impairment (MCI) defined as the substantial decrease in one or two or more than two cognitive domains which does not interfere with daily activities but requires a more significant effort from an individual's previous level of performance on an activity [1]. Main cognitive domains are memory and learning, complex attention, executive function, language, perceptual-motor and social cognition. A person with MCI is at a higher risk of developing Alzheimer's disease (AD) and other types of dementia like vascular dementia (VD), frontotemporal dementia. The AD and VD are the most prominent causes of dementia in elders worldwide [2]. The estimations of the dementia prevalence are abruptly increasing, especially in Asia continent. In 2015, 22.9 million people had dementia out of 46.8 million worldwide population, and their number doubles every 20 years all over the globe as elders population is projected to jump nearly 17% by 2050 [3]. It also reported that 15–20% of elders aged >65 years have MCI [4]. The differential diagnosis between normal cognition and MCI is very challenging because of inherent arbitrary boundaries. For example, the decline in executive functions may go unnoticed with ageing, which could be a prodromal phase of any neurodegenerative disorder.

Traditionally, cognitive testing alone is the standard evaluation criteria for MCI in which person's performance compared with appropriate norms of age group, educational and cultural background. Cognitive testing cannot be available in all circumstances, and its scores are sensitive to particular tests and norms as well as test environments, sensory inputs, and intercurrent illness. There is not a single cognitive test enough to diagnose MCI accurately on the mass population with educational and cultural versatility [5], [1].

Recently, dementia and related neurocognitive disorder are potential field of medical science. In this, EEG signal processing could be used for effective diagnosis and prognosis of the disease [6]. Some studies suggest that certain EEG features could be used to predict the early sign of cognitive impairments using nonlinear analytical tools and machine learning techniques [7], [8]. Some recent studies have reported the excellent classification accuracy of dementia group using classifiers such as support vector machine (SVM) [9], the artificial neural network (ANN) [10] and linear discriminate [11] with various EEG features. Most of these studies have been done using binary classification [12], [13], [14], [15].

Among many classifiers, SVM has widely used classifier for neurodegenerative disorders and neurological disease, such as AD [16], [17], [18] and epilepsy [8]. Because it performs well with small data-set depending upon the feature input. In our previously reported work [16], Sharma et al. (2019) have diagnosed dementia, MCI and age match normal in four cognitive events with good accuracy ranging between 73% to 89.8%. The main limitation of work was a hierarchy based binary SVM model, which does not support the practical application of a diagnostic tool. Besides this, normal and MCI group could not be differentiated entirely by cognitive parameter or EEG features alone. Therefore, it is essential to search for efficient EEG features for multi-classification. The objective of the current work is to classify dementia, MCI and normal subjects with reasonable accuracies using six EEG features with multi-class SVM in three different cognitive states; resting-state, relaxing-state and motor speed test (MST). This work investigates the efficiency of Six different EEG features, namely spectral kurtosis (SK), skewness, spectral entropy (SE), fractal dimension (FD) and power spectral density (PSD) of beta and gamma. This approach made it original work with different features and classifier than our previously reported work [16]. In this experiment, motor speed considered as an age-dependent cognitive parameter. Because, it devoid the significant bias, i.e., educational status and language of the subjects participating in the study which acts as an essential confounder in the memory and comprehension-based study tools. Our EEG findings have the potential to distinguish MCI with good accuracy, sensitivity and specificity. The main concern was undefined differences between healthy ageing and MCI, and diagnose these using a multi-class classifier to assist automation in future. The signs of neurocognitive impairment are loss of memory function, language function and psychomotor function [19]. Motor speed has been a vital parameter in the diagnosis of epilepsy [20], stroke, Parkinsonism and traumatic brain injury [21]. The main contributions of the proposed work summarise as

• In this study, an experimental protocol is introduced for clinical setup which is independent of dialect and education of the subjects.

• An efficient set of EEG features is utilised to classify dementia, MCI and normal subject in different cognitive states.

• Model a classifier to differentiate the subject groups which has performance validation based on accuracy, precision, specificity, sensitivity, MCC and F-measure.

• We have achieved novel findings of the MST with multi-classification for dementia diagnosis.