From: The applied principles of EEG analysis methods in neuroscience and clinical neurology
References | Data type | Subjects | Method | Disease/state | Application | Effect evaluation |
---|---|---|---|---|---|---|
Samiee et al. [49] | EEG | Bonn dataset | STFT | Epilepsy | Utilized the STFT method to extract EEG data during seizure periods and non-seizure periods | The proposed method in this study was asserted to surpass competing techniques in classification accuracy while offering a compact representation of EEG time series |
Beeraka et al. [50] | EEG | Bonn dataset | STFT | Epilepsy | Detected epileptic seizures in patients using the STFT method | Using CNN and BiLSTM models, this article reported an average classification accuracy of 93.9% and 97.2%, respectively |
Bajaj et al. [51] | EEG | Normal (n = 120) | STFT | Alcohol | Employed the STFT algorithm to classify EEG features in patients with alcoholic encephalopathy | Experimental outcomes and comparisons with state-of-the-art algorithms led the article to claim the superior performance of the proposed method over competing algorithms |
Sheikhani et al. [52] | EEG | Autism disorder (males = 9, female = 1) | STFT | Autism disorder | Adopted the STFT algorithm to analyze the brain activity of autistic patients | The beta band (14–34 Hz) was highlighted in this article for demonstrating an 82.4% discrimination rate between the two groups |
Krishnan et al. [53] | EEG | DROZY database | STFT | Drowsiness | Utilized the STFT algorithm to detect and categorize two sleep states, “drowsy” and “alert” | Employing the KNN classifier, the article achieved classification accuracies of 96.1% (dataset 1) and 95.5% (dataset 2) |
Bajaj et al. [54] | EEG | Normal (n = 8) | WVD | Sleep stages | Applied the WVD to extract the characteristics of sleep stages | Effectiveness in classifying sleep stages from EEG signals was demonstrated in this article through the proposed method |
Yan et al. [55] | iEEG | Epilepsy (n = 21) | WVD | Epilepsy | Detected seizures in patients using the WVD | This article showcased a satisfactory sensitivity of 94.26%, a specificity of 96.34%, and a very short delay time of 0.56 s |
Ebrahimzadeh et al. [56] | EEG | Normal (n = 32) | WVD | Deception | Developed a P300-based deception detection method using the WVD | The method presented in this article claimed to detect deception with an accuracy of 89.73%, surpassing the performance of previously reported methods |
Khare et al. [57] | EEG | SZ (n = 49) Normal (n = 32) | WVD | SZ | Used the WVD approach to automatically detect EEG signals in schizophrenic patients | An accuracy of 93.36% was achieved in this article using the smoothed pseudo-WVD-based time–frequency representation and CNN model |
Faust et al. [58] | EEG | Bonn dataset | WT | Epilepsy | Adopted the WT algorithm to develop an assisted seizure detection system | The article asserted that the method presented is the most effective for the automated EEG-based diagnosis of epilepsy |
Gandhi et al. [59] | EEG | Epilepsy Normal | WT | Epilepsy | Developed an expert model for epileptic activity detection using the WT method | Detection accuracy of 99.33%, along with a sensitivity of 99.6% and specificity of 99%, was claimed in this article |
Anuragi et al. [60] | EEG | Alcoholic and non-alcoholic (n = 122) | WT | Alcohol | Applied the WT algorithm for the automatic detection of alcoholism | The LS-SVM using a polynomial kernel was reported as the best performer in this article, achieving an accuracy of 99.17%, sensitivity of 99.17%, and specificity of 99.44% with a tenfold cross-validation technique |
Murugappan et al. [61] | EEG | Normal (n = 20) | WT | Emotion | Reported the application of the WT algorithm for emotional recognition | Maximum average classification rates of 83.26% using KNN and 75.21% using LDA were achieved in this article, outperforming conventional features |
Adeli et al. [62] | EEG | Epilepsy (n = 2) | WT | Epilepsy | Analyzed the EEG signals of epileptic patients with the WT algorithm | Wavelet analyses of EEGs from a patient population were suggested in this article to potentially indicate physiological processes during epilepsy onset |
Hamad et al. [63] | EEG | Bonn dataset | WT | Epilepsy | Utilized the WT algorithm to extract multiple epileptic features | Extracting 10 features from EEG signals based on discrete WT, this article aimed to improve accuracy in classifying EEG signals for epilepsy detection |
Li et al. [64] | EEG | Bonn dataset Epilepsy (n = 6) | EMD | Epilepsy | Analyzed automated seizure detection using the EMD algorithm | This article reported sensitivities of 97.00% and 98.00%, and specificities of 96.25% and 99.40% for interictal and ictal EEGs and normal and ictal EEGs, respectively, on Bonn datasets |
Siuly et al. [65] | EEG | SZ (n = 49) Normal (n = 32) | EMD | SZ | Applied the EMD algorithm for the automated detection of SZ | The ensemble bagged tree was identified as the best classifier in this article, achieving a 93.21% correct classification rate for SZ |
Chen et al. [66] | EEG | Awareness (samples = 558) Anesthesia (samples = 558) | EMD | Anesthesia | Adopted the EMD algorithm to classify anesthetized and awake states | The article claimed that the IMF performed the best, with an AUC of 0.993 for FFT (or 0.989 for Hilbert transform) |
Babiker et al. [67] | EEG | Normal (males = 38, females = 5) | EMD | Situational interest | Detected the situational interest of students in the learning process with the application of the EMD algorithm | While SVM achieved high accuracies of 93.3% and 87.5% for two datasets using features from four EEG channels, the KNN classifier achieved comparable accuracies of 87.5% and 86.7% for the same datasets using a single EEG channel, as reported in this article |
Priya et al. [68] | EEG | Alcoholic and non-alcoholic (n = 122) | EMD | Alcohol | Employed an EMD algorithm to differentiate between alcoholic EEG and normal EEG | Accuracy results for an LS-SVM classifier with polynomial and RBF kernels were stated as 96.67% and 97.92%, respectively, in this article |