Table 5.
Comparison with previous studies for automated depression detection
| Reference/year | Subjects normal (N) vs. depressed (D) | Processing method | Number of channels | Classifier | CV | Performance (%) | ||
|---|---|---|---|---|---|---|---|---|
| ACC | SEN | SPE | ||||||
| [26], Subha et al. 2012 | 15 N vs. 15 D | Rhythm separation by DWT + RWE + entropy features | 2 | ANN | No | 98.11 | 98.73 | 97.50 |
| [6], Ahmadlou et al. 2012 | 12 N vs. 12 D | Rhythm separation by DWT + HFD and KFD | 7 | EPNN | No | 91.30 | – | – |
| [27], Ahmadlou et al. 2012 | “11 male and 11 female depressed subjects” | Rhythm separation by DWT + STARC | 19 | – | – | They founded that there is a significant difference between gender with depression. | ||
| [24],Hosseinifard et al. 2013 | 45 N vs. 45 D | Rhythm separation by butterworth filter + DFA, HFD, Correlation dimension and Lyapunov exponent | 19 | LR | LOO | 90.05 | – | – |
| [25], Faust et al. 2014 | 15 N vs. 15 D | WPD + ApEn, SampEn, REN and Ph | 2 | PNN | No |
Left: 98.20 Right: 99.50 |
Left:97.10 Right: 99.20 |
Left: 99.40 Right: 99.70 |
| [28], Acharya et al. 2015 | 15 N vs. 15 D | Nonlinear features | 2 | SVM | No | 98.00 | 97.00 | 98.50 |
| [29], Mumtaz et al. 2017 | 30 N vs. 34 D | Synchronization likelihood | 19 | SVM | 10-Fold | 98 | 99.90 | 95 |
| [30], Liao et al. 2017 | 20 N vs. 20 D | Kernel Eigen-Filter-Bank Common Spatial Patterns + principal component analysis | 8 | SVM | LOO | 81.23 | – | – |
| [31], Bairy et al. 2017 | 15 N vs. 15 D | LPC | 2 | BT | LOO | 94.30 | 91.46 | 97.45 |
| [32], Knott et al. 2001 | 70 N vs. 23 D | Rhythm separation by Fourier Transform + rhythms power + frequency + asymmetry + coherence | 19 | JR | No | 91.3 | – | – |
| [57], Bachmann et al. 2018 | 13 N vs. 13 D | three linear and three nonlinear features: Spectral asymmetry index, alpha band power variability, relative gamma power, HFD, DFA, and Lempel–Ziv complexity | 2 | LR | LOO | 92 | – | – |
| [33], Acharya et al. 2018 | 15 N vs. 15 D | Deep learning + 13 layer convolution neural network | 2 | CNN | 10-Fold |
Left: 93.54 Right: 95.96 |
Left: 91.89 Right: 94.99 |
Left: 95.18 Right: 96.00 |
| [34], Sharma et al. 2018 | 15 N vs. 15 D | BDL TCWFB + logarithm of L2 norm | 2 | SVM | 10-Fold |
Left: 99.02 Right: 99.54 |
Left: 98.66 Right: 98.66 |
Left: 99.38 Right: 99.38 |
|
Present study 2020 |
22 N vs. 22 D | Rhythm separation by EWT + CC | 2 | SVM | 10-Fold |
Left: 98.33 Right: 98.76 |
Left: 98.02 Right: 98.47 |
Left: 98.65 Right: 99.05 |
| KNN | 10-Fold |
Left: 97.63 Right: 98.37 |
Left: 96.39 Right: 97.39 |
Left: 98.87 Right: 99.36 |
||||
The bold notations indicate the results obtained by our proposed strategy