Table 9 Comparison of our results with recent DL studies on same databases.

From: Automated depression detection via cloud based EEG analysis with transfer learning and synchrosqueezed wavelet transform

Study (year)

Database

Processing methods

Evaluation

Accuracy

13 (2020)

I

Frequency-dependent multilayer brain, multilayer deep CNN

10-fold CV

97.27%

22 (2023)

I

DSNet

LOSO CV

91.69%

20 (2023)

I

Multi-head self-attention, connectivity measures (TE, PLV, Coh), CNN

LOSO CV

91.06%

10 (2022)

I

10 selected EEG data, Inception time model

10-fold CV

90.10%

4 (2023)

I

Paired asymmetry and DFA (temporal, parietal), SVM

Hold out CV

89.24%

53 (2022)

I

WCOH, 2D CNN

10-fold CV

98.10%

Ours (2024)

I, II

SSWT (parietal lobe), data augmentation, ResNet-18

LSO CV

98% (Database I), 91% (Database II)

  1. Articles on database II are almost based on classification of multiclass of depression and could not compare with this study.