Table 4.
Age groups | FT-G1 | FT-I-G7 | FT-I-G2 | FT-SaBN-G7 | FT-SaBN-G2 |
---|---|---|---|---|---|
B | 74.9 ± 6.8 | 74.1 ± 6.6 | 74.8 ± 6.2 | 72.2 ± 7.7 | 72.6 ± 7.7 |
C | 75.0 ± 9.8 | 74.9 ± 9.2 | 75.9 ± 9.1 | 74.8 ± 8.9 | 75.6 ± 10.1 |
A | 82.7 ± 13.7 | 80.0 ± 14.6 | 82.8 ± 13.6 | 82.3 ± 13.7 | 82.0 ± 14.0 |
YA | 80.8 ± 11.5 | 80.6 ± 11.6 | 80.6 ± 11.6 | 80.3 ± 11.9 | 79.9 ± 11.9 |
MA | 80.4 ± 7.8 | 79.90 ± 8.0 | 79.8 ± 8.2 | 79.6 ± 8.0 | 79.4 ± 8.3 |
E | 75.7 ± 10.1 | 74.2 ± 10.7 | 74.9 ± 10.2 | 74.5 ± 10.6 | 73.9 ± 10.9 |
OE | 75.2 ± 11.7 | 73.9 ± 11.0 | 74.9 ± 11.3 | 73.8 ± 11.7 | 74.0 ± 11.3 |
avg | 77.9 ± 10.7 | 77.0 ± 10.8 | 77.6 ± 10.7 | 76.9 ± 11.0 | 76.8 ± 11.1 |
Performance of U-Sleep-v1 on a single model fine-tuned on all the training set of the seven BSDB groups (FT-G1); on seven/two models fine-tuned on the independent training set of each group with G = 7 (FT-I-G7) and G = 2 (FT-I-G2) respectively; and on a single model fine-tuned on all the training set of the seven/two BSDB groups conditioned by G = 7 (FT-SaBN-G7) and by G = 2 (FT-SaBN-G2) groups respectively. All the fine-tuned models are evaluated on the associated test set of each group (data split in Supplementary Table 8). We report the F1-score (%F1), specifically the mean value and the standard deviation (μ ± σ) computed across the recordings. B Babies (0–3 years), C Children (4–12 years), A Adolescents (13–18 years), YA Young Adults (19–39 years), MA Middle-aged adults (40–59 years), E Elderly (60–69 years), OE Old Elderly (≥ 70 years). When G = 2 we have the following two groups G1 = {B ∪ C}, G2 = {A ∪ YA ∪ MA ∪ E ∪ OE}, further details in Supplementary notes: Age analysis.