TABLE 1.
Other results from the literature of age estimation using EEG. These figures were obtained using different methods, age ranges and datasets. EO and EC, respectively, denote eyes open and eyes closed. The results are compared with ours for the best channel in age space, for NY, in log space and using bias correction (see Figure S5).
| Work | Individuals | Range | MAE | R 2/R | Details |
|---|---|---|---|---|---|
| Khayretdinova et al. (2022) | 1335 | [5, 88] | 6 ± 0.33 yr | R 2 = 0.81 | (EO and EC) Neural networks |
| Vandenbosch et al. (2019) | 702 | [5, 18] | — | 0.53 < r < 0.74 | Random Forest |
| Dimitriadis and Salis (2017) | 94 | [19, 67] | — | R 2 = 0.60/R 2 = 0.48 | EO/EC SV Regression + Clasifier |
| Al Zoubi et al. (2018) | 468 | [18, 58] | 6.68 ± 0.69 | R 2 = 0.37 | nested‐cross‐validation (NCV) and stack‐ensemble |
| Sun et al. (2019) I | 2535 | [18, 80] | 7.4 yr | R = 0.83 | (all electrodes) |
| Sun et al. (2019) II | 1974 | [40, 80] | Same I | Same I | Same I |
| Engemann et al. (2022) | 2500 | [25, 75] | Depends on a data set | R 2 in the range 0.60–0.74 (best) | 3 EEG data sets: (LEMON, CHBP, TUAB) |
| HarMNqEEG data set (present work) | 1926 | [5, 97] | 10.78 yr./ 5.24 yr. (NY best) | R 2 = 0.44 (best in age space) R2 = 0.59 (NY best) | R 2 = 0.53 (best in log space) R 2 = 0.73 (best with Bias correction) |