Deep transfer learning for improving single-EEG arousal detection
This incremental improvement helps sleep science researchers with small datasets by enabling the use of pre-trained deep learning models.
The paper tackled the problem of channel mismatch in EEG arousal detection by applying deep transfer learning to adapt models from multivariate to single-channel data, achieving an F1 score of 0.682, similar to the baseline's 0.694 and better than a single-channel model.
Datasets in sleep science present challenges for machine learning algorithms due to differences in recording setups across clinics. We investigate two deep transfer learning strategies for overcoming the channel mismatch problem for cases where two datasets do not contain exactly the same setup leading to degraded performance in single-EEG models. Specifically, we train a baseline model on multivariate polysomnography data and subsequently replace the first two layers to prepare the architecture for single-channel electroencephalography data. Using a fine-tuning strategy, our model yields similar performance to the baseline model (F1=0.682 and F1=0.694, respectively), and was significantly better than a comparable single-channel model. Our results are promising for researchers working with small databases who wish to use deep learning models pre-trained on larger databases.