Table 3.
Applications of correlation analyses
| References | Data type | Subjects | Method | Disease/state | Application | Effect evaluation |
|---|---|---|---|---|---|---|
| Kang et al. [73] | EEG | SHHS dataset | CORR | Sleep stages | Analyzed the severity of symptoms in patients with OSA using the CORR method | Variations in microstructures were identified between the PSG-derived sleep EEG of non-OSA participants and those with varying severities of OSA in this study |
| Dauwels et al. [74] | EEG |
MCI (n = 25) Normal (n = 56) |
CORR | AD | Employed CORR to assess early symptoms of AD | Stochastic event synchrony was proposed as a feature to differentiate MCI patients from age-matched controls, achieving a leave-one-out classification rate of 83%, as reported in this article |
| Yasuhara [75] | EEG | Autistic children (n = 1014) | CORR | ASD | Used CORR to analyze the relationship between EEG abnormalities and ASD | The article suggested a correlation between ASD and dysfunction in the mirror neuron system |
| Islam et al. [76] | EEG | Normal (males = 16, females = 16) | CORR | Emotion | Integrated the CORR method with a CNN to identify emotions | Maximum accuracies of 78.22% on valence and 74.92% on arousal were attained using the internationally authorized DEAP dataset in this study |
| Sheorajpanday et al. [77] | EEG | Stroke (n = 110) | CORR | Stroke | Investigated the correlation between the EEG symmetry index and the Rankin scale and determined the prognostic value of EEG signals in the diagnosis of stroke | Prognostic value for disability, dependency, and death after 6 months in the subacute setting of ischemic stroke was attributed to EEG, according to this article |
| Alba et al. [78] | EEG |
ADHD (n = 10) Normal (n = 12) |
COH | ADHD | Adopted COH to analyze the functional connectivity of EEG in patients with ADHD under different resting states | Global connectivity of each region and its temporal variability were posited as better reflections of the underlying neural dysfunctions producing ADHD in this article |
| Carrasco-Gomez et al. [79] | EEG | Postanoxic coma (n = 594) | COH | Postanoxic coma | Assessed EEG functional connectivity in the context of post-anoxic coma through COH | The best non-coupling-based model, using data at 12 h and 48 h, achieved a sensitivity of 32% at 100% specificity, as claimed in this article |
| Barry et al. [80] | EEG | Normal (boys = 40, girls = 40) | COH | Developmental trends in normal children | Employed COH to analyze brain development in normal children of different ages and genders | This article asserted that EEG coherences in normal children aged 8 to 12 systematically develop with age |
| Locatelli et al. [81] | EEG |
AD (n = 10) Normal (n = 10) |
COH | AD | Analyzed the EEG signal characteristics of AD | Alpha coherence decrease was linked to alterations in cortico-cortical connections, while delta coherence increase was associated with the lack of influence of subcortical cholinergic structures on cortical electrical activity, as claimed in this article |
| Coben et al. [82] | EEG |
ASD (n = 20) Normal (n = 20) |
COH | ASD | Found neural underconnectivity in patients with ASD through COH, which is consistent with the results of other methods | Dysfunctional integration of frontal and posterior brain regions, along with a pattern of neural underconnectivity, was suggested in autistic subjects, as reported in this article |
| Catarino et al. [83] | EEG |
ASC (n = 15) Normal (n = 15) |
WTC | ASC | Probed task-related functional connections in the setting of the autism spectrum using the WTC algorithm | Impairment in task differentiation in individuals with ASC relative to typically developing individuals was reflected in this article |
| Omidvarnia et al. [84] | EEG | Epilepsy (n = 7) | WTC | Epilepsy | Discussed whether there was a direct correlation between EEG and regional hemodynamic brain connectivity changes in focal epilepsy | A strong time-varying relationship between local fMRI connectivity and interictal EEG power in focal epilepsy was claimed in this article |
| Khan et al. [85] | EEG |
MDD (n = 30) Normal (n = 30) |
WTC | MDD | Studied the diagnosis of depression using the WTC approach | An accuracy of 98.1%, sensitivity of 98.0%, and specificity of 98.2% were achieved in this article, with another method yielding 100% accuracy, sensitivity, and specificity |
| Sankari et al. [86] | EEG | AD (n = 20) | WTC | AD | Utilized the WTC method to explore the diagnosis of AD | WTC was proposed as a powerful tool to differentiate between healthy older individuals and probable AD patients in this article |
| Briels et al. [87] | EEG |
SCD (n = 399) AD (n = 410) |
PLV/PLI | AD | Analyzed the reproducibility of EEG functional connections in AD using PLV/PLI | In alpha/beta bands and PLI and wPLI in the theta band were highlighted for providing valid insights into disease-associated changes, correlating with disease severity, as indicated in this study |
| Olejarczyk et al. [88] | EEG |
SZ (males = 7, females = 7) Normal (males = 7, females = 7) |
PLV/PLI | SZ | Assessed brain connectivity in patients with SZ using PLV/PLI | Comparing different connectivity measures using graph-based indices for each frequency band separately was suggested as a useful tool in the study of connectivity disorders, such as SZ |
| Wang et al. [89] | EEG |
DEAP dataset Normal (males = 7, females = 8) |
PLV/PLI | Emotion | Explored the dynamics of rich-club structures in the brain during emotional changes, utilizing dynamic PLV brain networks and ReliefF algorithm to derive emotionally relevant features for accurate emotion recognition | Rich-club composition with subtle temporal variations was revealed, emphasizing the importance of small-scale structure connections in distinguishing emotions, achieving high accuracy (86.11% and 87.92%) in valence dimension validation on DEAP and SEED datasets |
| Huang et al. [90] | EEG | CAP dataset | PLV/PLI | Sleep stages | Highlighted the importance of exploring global information exchange between brain regions for improved sleep evaluation and disease diagnosis | High classification accuracy (96.91% intra-subject, 96.14% inter-subject) in sleep stage classification surpassed the performance of decision-level and hybrid fusion methods in this study |
| Zuchowicz et al. [91] | EEG |
MDD (n = 8) BP (n = 10) |
PLV/PLI | MDD | Explored the impact of repeated transcranial magnetic stimulation on patients with depression through the PLV/PLI approach | PLV analysis was indicated as a potential indicator of the response to depression treatment, enhancing therapy effectiveness in this research |
| Chen et al. [92] | EEG |
ADHD (girls = 9) Normal (n = 51) |
MI | ADHD | Adopted MI to extract the brain network of children with ADHD | A convincing performance with an accuracy of 94.67% regarding the test data was achieved in this article |
| Aydin et al. [93] | EEG | Normal | MI | Sleep stages | Analyzed the EEG of insomnia patients using MI | The level of cortical hemispheric connectivity was claimed to be strongly associated with sleep disorders in this article |
| Piho et al. [94] | EEG |
DEAP dataset MAHNOB dataset |
MI | Emotion | Determined emotion recognition features through MI | Significant improvement in emotion recognition accuracy was demonstrated in experimental results on publicly available datasets, as claimed in this article |
| Hassan et al. [95] | EEG |
Bonn dataset CHB-MIT dataset |
MI | Epilepsy | Applied MI to identify individual features for epileptic seizure detection | Significant performance improvement compared to recent state-of-the-art methods was reported in this article |
| Yin et al. [96] | EEG |
SZ (positive = 14, negative = 14) Normal (n = 14) |
MI | SZ | Analyzed brain functional connectivity in patients with SZ using the MI approach | Information interactions in SZ patients were claimed to be fewer than in normal controls, with positive SZ exhibiting more interactions than negative SZ, along with slower and less efficient information transfer between brain regions, according to this article |
| Sanz-García et al. [97] | EEG | SAH (n = 21) | GC | Subarachnoid hemorrhage | Used the GC algorithm to determine the causal relationship between EEG activity and changes in ICP in neurocritical care patients | A significant GC statistic from EEG activity to ICP was found during 37.88% of the analyzed time, with typical lags of 25–50 s between them, as reported in this article |
| de Tommaso et al. [98] | EEG | Migraine (males = 3, females = 28) | GC | Migraine | Adopted the GC algorithm to explore the functional connectivity of EEG signals in migraine patients responding to laser stimulation | Brain network analysis was suggested to aid in understanding subtle changes in pain processing under laser stimuli in migraine patients in this article |
| Nicolaou et al. [99] | EEG | Normal (males = 21) | GC | Anesthetized | Utilized the GC algorithm to distinguish between “awake” and “anesthetized” states | Features derived from GC estimates resulted in the classification of awake and "anesthetized" states in 21 patients with maximum average accuracies of 0.98 and 0.95, respectively, according to this article |
| Nicolaou et al. [100] | EEG | Normal (males = 21) | GC | Anesthetized | Utilized the GC algorithm to distinguish between “awake” and “anesthetized” states | The methodology of GC analysis of EEG data was claimed to carry implications for integrated information and causal density theories of consciousness in this article |
| Barrett et al. [101] | EEG | Normal (n = 7) | GC | Anesthetized | Investigated propofol-induced anesthesia using the GC algorithm | Significant increases in bidirectional GC during loss of consciousness, especially in the beta and gamma frequency ranges, were claimed in this article |
| Coben et al. [102] | EEG | Epilepsy (n = 2) | GC | Seizure location | Analyzed brain functional connectivity in epilepsy through the GC algorithm | Hypercoupling near the seizure foci and low causality across nearby and associated neuronal pathways were suggested in this article |
| Chen et al. [103] | EEG |
MCI (n = 46) AD (n = 43) |
MCI and AD | CFC | Analyzed resting state EEG in patients with MCI and AD using CFC | Alterations in theta-gamma coupling in the temporal lobe were claimed to become progressively obvious during disease progression, serving as a valuable indicator of MCI and AD pathology, as suggested in this article |
| Lynn et al. [104] | EEG | Not reported | SZ | CFC | Analyzed the working memory of schizophrenic patients using CFC | Formal testing of theta-gamma interaction was proposed as imperative in this article |
| Papadaniil et al. [105] | EEG | Normal (males = 14) | Auditory Perception | CFC | Used CFC to study auditory perception tasks | Stronger coupling in the delta band, closely linked to sensory processing, was observed and claimed in this article |
| Park et al. [106] | EEG | Normal | visual memories | CFD | Used CFD to study the formation of visual memory | Encoding of visual information reflecting a state determined by the interaction between alpha and gamma activity was asserted in this article |
AD Alzheimer’s disease, ADHD attention deficit hyperactivity disorder, ASC autism spectrum condition, ASD autism spectrum disorder, CFC cross-frequency coupling, CFD cross-frequency directionality, CHB-MIT Children’s Hospital Boston and the Massachusetts Institute of Technology, COH coherence, CORR correlation, DEAP database for emotion analysis using physiological signals, EEG electroencephalography, fMRI functional magnetic resonance imaging, GC granger causality, ICP intracranial pressure, MAHNOB Multimodal Database for Affect Recognition and Implicit Tagging, MCI mild cognitive impairment, MDD major depressive disorder, MI mutual information, OSA obstructive sleep apnea, PLV/PLI phase lag value/index, PSG polysomnography, SEED Shanghai Jiao Tong University emotion EEG dataset, SHHS sleep heart health study, SZ schizophrenia, wPLI weighted phase lag index, WTC wavelet coherence