Sleep stage scoring on MASS (subject-wise)
87.5AccuracyTinySleepNet
Evaluation results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| TinySleepNetYear=2020, Technique=CNN & RNN, LP=LFS, Signals=EEG2022.07 | 87.5 | 83.2 | 0.82 | — | — | — | — | — | |
| Classical ML PipelineYear=2022, Technique=Catboost, LP=LFS, Signals=EEG + EOG + EMG2022.07 | 86.7 | 81.7 | 0.803 | — | — | — | — | — | |
| IITNetArchitecture=CNN + RNN, Channel=F4-EOG (left), Input=Raw Signal, Approach=Many-to-One, Sequence Length=10 (9 past), Subjects=622019.02 | 86.3 | 80.5 | 0.79 | 85.4 | 54.1 | 91.3 | 86.8 | 84.8 | |
| Classical ML PipelineYear=2022, Technique=Catboost, LP=LFS, Signals=EEG + EOG2022.07 | 86.3 | 80.9 | 0.797 | — | — | — | — | — | |
| IITNetYear=2019, Technique=CNN & RNN, LP=LFS, Signals=EEG2022.07 | 86.3 | 80.5 | 0.79 | — | — | — | — | — | |
| DeepSleepNetEEG Channel=F4-EOG (Left), Test Epochs=586002017.03 | 86.2 | 81.7 | 0.8 | 87.3 | 59.8 | 90.3 | 81.5 | 89.3 | |
| IITNetArchitecture=CNN + RNN, Channel=F4-EOG (left), Input=Raw Signal, Approach=Many-to-One, Sequence Length=4 (3 past), Subjects=622019.02 | 86.2 | 80 | 0.79 | 85.2 | 51.8 | 91.4 | 86.9 | 84.5 | |
| SupratakArchitecture=CNN + RNN, Channel=F4-EOG (left), Input=Raw Signal, Approach=Many-to-One, Sequence Length=Whole previous night epochs, Subjects=622019.02 | 86.2 | 81.7 | 0.8 | 87.3 | 59.8 | 90.3 | 81.5 | 89.3 | |
| DeepSleepNetYear=2017, Technique=CNN & RNN, LP=FT, Signals=EEG2022.07 | 86.2 | 81.7 | 0.8 | — | — | — | — | — | |
| Ref. [13]EEG Channel=F4-EOG (Left), Test Epochs=590662017.03 | 85.9 | 80.5 | — | 84.6 | 56.3 | 90.7 | 84.8 | 86.1 | |
| DongArchitecture=Mixed RNN, Channel=F4-EOG (left), Input=Handcrafted Features, Approach=Many-to-One, Sequence Length=5 (4 past), Subjects=622019.02 | 85.9 | 80.5 | 0.79 | 84.6 | 56.3 | 90.7 | 84.8 | 86.1 | |
| Classical ML PipelineYear=2022, Technique=Logistic regr., LP=LFS, Signals=EEG + EOG + EMG2022.07 | 85.3 | 80.7 | 0.786 | — | — | — | — | — | |
| IITNetArchitecture=CNN + RNN, Channel=F4-EOG (left), Input=Raw Signal, Approach=One-to-One, Sequence Length=1, Subjects=622019.02 | 84.5 | 76.6 | 0.77 | 86.9 | 36.8 | 90.5 | 88.1 | 81 | |
| Classical ML PipelineYear=2022, Technique=Logistic regr., LP=LFS, Signals=EEG + EOG2022.07 | 84.5 | 79.4 | 0.775 | — | — | — | — | — | |
| Multitask 1-max CNNInput channel=C4-A1 + ROC-LOC + CHIN1-CHIN2, Feature type=learned, Subjects=200, Independent testing=yes2018.05 | 83.6 | — | — | — | — | — | — | — | |
| Multitask 1-max CNNInput channel=C4-A1 + ROC-LOC, Feature type=learned, Subjects=200, Independent testing=yes2018.05 | 82.5 | — | — | — | — | — | — | — | |
| DeepSleepNet (Deep CNN)Input channel=C4-A1 + ROC-LOC + CHIN1-CHIN2, Feature type=learned, Subjects=200, Independent testing=yes2018.05 | 80.7 | — | — | — | — | — | — | — | |
| Deep CNNInput channel=C4-A1 + ROC-LOC + CHIN1-CHIN2, Feature type=learned, Subjects=200, Independent testing=yes2018.05 | 79.9 | — | — | — | — | — | — | — | |
| ResNetInput channel=C4-A1 + ROC-LOC + CHIN1-CHIN2, Feature type=learned, Subjects=200, Independent testing=yes2018.05 | 79.4 | — | — | — | — | — | — | — | |
| Multitask 1-max CNNInput channel=C4-A1, Feature type=learned, Subjects=200, Independent testing=yes2018.05 | 78.6 | — | — | — | — | — | — | — | |
| PhanArchitecture=Multitask 1-max CNN, Channel=C4-A1, Input=Spectrogram, Approach=One-to-Many, Sequence Length=1, Subjects=2002019.02 | 78.6 | 70.6 | 0.7 | — | — | — | — | — | |
| Deep CNN (Tsinalis et al.)Input channel=C4-A1 + ROC-LOC + CHIN1-CHIN2, Feature type=learned, Subjects=200, Independent testing=yes2018.05 | 77.9 | — | — | — | — | — | — | — | |
| RobustSleepNetYear=2021, Technique=RNN, LP=FT, Signals=EEG + EOG2022.07 | — | 84 | — | — | — | — | — | — | |
| RobustSleepNetYear=2021, Technique=RNN, LP=LFS, Signals=EEG + EOG2022.07 | — | 82.2 | — | — | — | — | — | — | |
| RobustSleepNetYear=2021, Technique=RNN, LP=DT, Signals=EEG + EOG2022.07 | — | 80.8 | — | — | — | — | — | — | |
| SimpleSleepNetYear=2020, Technique=RNN, LP=LFS, Signals=EEG + EOG2022.07 | — | 84.7 | — | — | — | — | — | — | |
| U-SleepYear=2021, Technique=CNN, LP=DT, Signals=EEG + EOG2022.07 | — | 80 | — | — | — | — | — | — |