Table 1.
Applications of power spectrum analyses
| References | Data type | Subjects | Method | Disease/state | Application | Effect evaluation |
|---|---|---|---|---|---|---|
| Ogilvie et al. [20] | EEG | 18–25 years (male = 1, females = 8) | FFT | Sleep stages | Used the FFT method to reflect the energy changes during the onset and phase transition of sleep | During the transition from stage 2 to REM sleep, observable systematic changes in EEG power density were reported across four standard frequency bands |
| Hadjiyannakis et al. [21] | EEG | Normal (n = 9) | FFT | Sleep stages | Conducted an in-depth study of sleep 20 years ago using the FFT method, which reflected energy changes during the onset and staging transitions of sleep | Sleep onset was identified in this study by the cessation of responses coupled with sharp increases in EEG synchronization |
| Sun et al. [22] | EEG | SZ (males = 36, females = 18) | FFT | SZ | Adopted the FFT method to explore the EEG features of schizophrenic patients | An average accuracy of 96.34% was attained using FFT in this study |
| Behnam et al. [23] | EEG | Autism disorders (n = 10) | FFT | Autism disorders | Utilized the FFT method to analyze EEG differences in autistic patients with different electrodes and achieved significant performance at the FP1, F3 position | STFT-BW demonstrated an 82.4% discrimination rate between normal and autistic subjects using the Mahalanobis distance, whereas FFT and STFT did not yield significant results |
| Djamal et al. [24] | EEG |
Stroke (n = 25) Normal (n = 25) |
FFT | Stroke | Improved the recognition accuracy of a one-dimensional (1D) CNN for the EEG signals of stroke patients | Utilizing FFT for identification was suggested to enhance accuracy by 45–80% compared to relying solely on 1D CNN, according to the findings in this article |
| Farihah et al. [25] | EEG | Dyslexic (boys = 4) | FFT | Dyslexic | Applied the FFT method to explore the EEG characteristics of dyslexic patients | Distinct differences in hemispheric activation during the construction of sentences and nonsense sentences were observed between poor and capable dyslexic subjects, as reported in this study |
| Melinda et al. [26] | EEG | From King Abdul Aziz University (KUA) | FFT | ASD | Analyzed EEG differences in epileptic patients through the FFT approach | The article highlighted alterations in the PSD values, noting an increase in the alpha and beta sub-bands for normal EEG signals and a decrease for autistic EEG signals |
| Bian et al. [27] | EEG |
MCI (n = 16) Normal (n = 12) |
Welch | Amnestic MCI in diabetes | Achieved early identification of mild cognitive dysfunction using Welch’s method | Proposed indices derived from resting-state EEG recordings were proposed to serve as tools for monitoring cognitive function in diabetic patients and aiding in diagnosis, according to the claims in this article |
| Yuan et al. [28] | EEG | Normal rats (males = 5) | Welch | Transcranial ultrasound stimulation | Studied the changes in the brain function of animals responding to different intensities of transcranial ultrasound stimulation through Welch’s method | The article suggested that power spectrum analysis holds significant reference value for brain stimulation, providing estimates of the extent of stimulation or inhibition of excitement |
| Wang et al. [29] | EEG | ASD (males = 14, females = 4) | Welch | ASD | Investigated the impacts of neurofeedback training on the cognitive function of autistic children based on the changes in frequency band energy | Neuro-feedback was presented as an effective method for altering EEG characteristics associated with ASD in this article |
| Göker [30] | EEG |
Migraine (males = 5, females = 13) Normal (males = 9, females = 12) |
Welch | Migraine | Adopted the Welch method to improve the accuracy of automatic migraine detection | The highest performance, with a 95.99% accuracy, was reported for the BiLSTM deep learning algorithm using 128 channels in this article |
| Hu et al. [31] | EEG | Normal (n = 3) | Welch | Hypoxia | Employed Welch’s method in combination with BP and SVM classifiers for hypoxic EEG classification, achieving an accuracy of 94.2% (BP classifier) and 92.5% (SVM classifier) | Distinguishing hypoxic EEG from normal EEG in individuals was demonstrated in this study |
| Wijaya et al. [32] | EEG | Stroke (n = 10) | Welch | Stroke | Used the Welch method for the classification of patients with acute ischemic stroke, similar to CT scan results | All BSI calculations exceeded those of healthy subjects (0.042 ± 0.005), indicating acute ischemic stroke in all subjects, as presented in this article |
| Cornelissen et al. [33] | EEG | Infants (n = 36) | Multitaper | General anesthesia in infants | Applied this method to study the effects of drugs under anesthesia on neonatal brain function | The article emphasized the necessity of age-adjusted analytical approaches for developing neurophysiology-based strategies in pediatric anesthetic state monitoring |
| Yang et al. [34] | EEG | Normal (n = 35) | Multitaper | SSVEP | Improved the accuracy of 40-class SSVEP using the multitaper method | The proposed method was asserted to effectively enhance the performance of a training-free SSVEP-based BCI system and balance recognition accuracy across different stimulation frequencies |
| Oliva et al. [35] | EEG | Bonn dataset | Multitaper | Epilepsy | Adopted the multitaper method for epilepsy detection with the assistance of different classifiers | This article reported achieving the highest accuracy for both binary (100%) and multiclass (98%) classification problems |
| Oliveira et al. [36] | EEG | DREAMs dataset | Multitaper | Sleep | Used the multitaper method for the automatic detection of KC waveforms in a sleep EEG and obtained favorable results | The method for automatic KC detection was asserted to improve detection metrics, particularly F1 and F2 scores, according to the claims in this article |
| Mohammadi et al. [37] | EEG | Normal (males = 6, females = 4) | AR | Person Identification | Achieved personal identification using the AR model | Classification scores ranging from 80 to 100% were achieved, revealing the potential of the approach for personal classification/identification |
| Perumalsamy et al. [38] | EEG | Normal (n = 5) | AR | Sleep spindles detection | Extracted sleep feature waves through the AR model | The algorithm's effectiveness in detecting sleep spindles and revealing alpha and beta band activities in EEG was demonstrated in this article |
| Saidatul et al. [39] | EEG | Normal (males = 5) | AR | Relaxation and mental stress condition | Applied AR modeling techniques to analyze EEG differences between relaxed and stressed states | A maximum classification accuracy of 91.17% was reported in this study |
| Lawhern et al. [40] | EEG | Normal (n = 7) | AR | Artifacts detection | Adopted the AR model to remove artifacts in EEG signals | AR modeling was suggested as a useful tool for discriminating artifact signals within and across individuals, according to the claims in this article |
| Mousavi et al. [41] | EEG | Bonn dataset | AR | Epilepsy | Used the AR model to automatically detect epileptic events in EEG signals | Correct classification scores in the range of 91% to 96% for epilepsy detection were reported in this study |
AR autoregressive, ASD autism spectrum disorder, BCI brain-computer interface, BiLSTM bidirectional long short-term memory, BP backpropagation, BSI brain symmetry index, CNN convolutional neural network, CT computed tomography, DREAM dialogue-based reading comprehension examination, EEG electroencephalography, FFT fast Fourier transform, KC k-complex, MCI mild cognitive impairment, PSD power spectral density, REM rapid eye movement, SSVEP steady state visually evoked potential, SVM support vector machine, SZ schizophrenia, STFT-BW short time Fourier transform at bandwidth of total spectrum