186

Bioinformatics of the Brain

modalities, aka Multimodal Neuroimaging (MN), is another approach that al-

lows to examination of brain structural and functional changes [73]. However,

due to practical use and high temporal resolution, EEG is the prominent

modality for neural decoding. Neuropsychiatric diseases present substantial

public health challenges as they play a major role in the global burden of

disease and significantly influence the social and economic welfare of popu-

lations. Although the large majority of EEG studies in the literature have

been investigated various biomarkers specific to the neuropsychiatric diseases

([74] for bipolar disorder (BD); [75] for SCZ, AD, and BD; [76] for AD, PD),

evaluating the results and identification of biomarkers, particularly in the con-

text of distinguishing these diseases, are challenging processes, and requires

expertise; therefore, although some challenges (such as sample size) remain,

there is a growing interest in the prevalence of EEG studies employing AI

techniques as a prognostic or computer-aided diagnostic tool that decodes the

brain activities [73].

Decoding is an essential phase that predicts the course of diseases using

brain activities as well as the connections among structures. To achieve this,

various AI models (see Section 7.3.5) are used to investigate significant distinct

EEG patterns associated with numerous illnesses including neurodegenerative

diseases (e.g., AD, PD), mood disorders (e.g., BD, Major depressive disorder

(MDD)) and different mental disorders such as Schizophrenia (SCZ) spectrum.

AD is a progressive neurodegenerative illness characterized by progressive

cognitive decline, memory loss, and impaired daily functioning. AD typically

progresses slowly mainly in three stages which are (1) The preclinical phase

characterized by the absence of clinical symptoms, although neuropathological

changes have initiated. (2) The Mild Cognitive Impairment (MCI) stage where

the individuals do not meet the criteria for AD, but there is notable mem-

ory impairment, particularly in the area of episodic memory, when compared

to individuals without cognitive issues. (3) Dementia stage that is marked

by substantial memory loss, along with observable impairments in various

cognitive domains, including language [77]. BCI studies have held paramount

significance in impeding the advancement of dementia in AD by paving the

way for early detection. A large majority of AI models were utilized by novel

researches to enhance diagnostic precision, facilitate early detection, subtype

recognition, predictive modeling, and personalized treatment planning for the

AD continuum [78]. Among different ML techniques, the Support Vector Ma-

chine (SVM) with different kernels is the most used model, followed by K-

nearest neighbor (KNN) and Linear Discriminant Analysis (LDA). One of the

most recent studies that distinguish AD and healthy control (HC) was pub-

lished by Nour et al. [79]. They employed Deep Ensemble Learning (DEL)

without applying any feature extraction after cleaning from noise and arti-

facts and reached an average accuracy of 97.9%. Another study attempted to

differentiate MCI from HC, applying a comparative deep-learning analysis of

resting-state EEG time series [80].