Table 3.
Study characteristics of notable research relevant to SZ diagnosis utilizing EEG (patients—with SZ).
| Work | Dataset | Number of Samples | Data Preparation | Software for Data Preparation | Feature Extraction Method | Approach | Result (%) |
|---|---|---|---|---|---|---|---|
| Lei Zhang [69]—2020 | Kaggle Basic Sensory Task data | Patients = 49, controls = 32 | Baseline selection, min–max normalization | - | Temporal, spatial, demographic & time–frequency features | Artificial neural network | Accuracy = 98.5 |
| Siuly et al. [70]—2023 | Patients = 49, controls = 32 | Average filtering | - | Deep ResNet | Softmax Layer and deep features with SVM | Accuracy = 99.23 | |
| Buettner et al. [72]—2019 | RepOD | Patients = 14, controls = 14 | ICA, normalization | - | Fourier transformation | Random Forest | Accuracy = 100 |
| Krishnan et al. [73]—2020 | Patients = 14, controls = 14 | - | - | Extraction using MEMD and entropy measures | SVM-RBF | Accuracy = 93, precision = 92, recall = 94 | |
| Shoeibi et al. [74]—2024 | Patients = 14, controls = 14 | Filtering, normalization, segmentation into time windows | - | 1D transformer architecture | Softmax classifier, 10-fold cross-validation | Accuracy = 97.62 | |
| Sara et al. [75]—2022 | RepOD-IBIB PAN | Patients = 14, controls = 14 | - | - | Connectivity matrix, TE | CNN-LSTM, 10-fold cross validation | Accuracy = 99.9 |
| Febles [79]—2022 | Clinical | Patients = 54, controls = 54 | Filtering, baseline correction, artifact rejection | - | Features related to peak-to-peak measurements and signal characteristics, Boruta algorithm | Multiple kernel learning | Accuracy = 86 |
| Aslan et al. [1]—2022 | Mental Health Research Center, Institute of Psychiatry & Neurology in Warsaw | Patients = 45 healthy = 39 Patients = 14, controls = 14 |
- | - | Time–frequency features | CNN | Accuracy = 98, precision = 98, recall = 98 Accuracy = 99.5, precision = 99, recall = 99 |
| Saadatinia et al. [80]—2024 | Patients = 45, controls = 39 Patients = 14, controls = 14 |
- | - | - | CNN, WGAN-GP and VAE | Accuracy = 99 | |
| C. Phang et al. [81]—2019 | MSU | Patients = 45, controls = 39 | - | - | DC-CN features | DNN-DBN | Accuracy = 95 |
| Rajesh et al. [82]—2021 | Patients = 45, controls = 39 | - | - | SLBP-based histogram features | LogitBoost Classifier | Accuracy = 91.66 | |
| Sobahi et al. [83]—2022 | Patients = 45, controls = 39 | - | - | Time–frequency features | ELM-based AE | Accuracy = 97.7 |