Cross-Sampling Rate Transfer Learning for Enhanced Raw EEG Deep Learning Classifier Performance in Major Depressive Disorder Diagnosis
Ellis, C. A.; Miller, R. L.; Calhoun, V. D.
Show abstract
Transfer learning offers a route for developing robust deep learning models on small raw electroencephalography (EEG) datasets. Nevertheless, the utility of applying representations learned from large datasets with a lower sampling rate to smaller datasets with higher sampling rates remains relatively unexplored. In this study, we transfer representations learned by a convolutional neural network on a large, publicly available sleep dataset with a 100 Hertz sampling rate to a major depressive disorder (MDD) diagnosis task at a sampling rate of 200 Hertz. Importantly, we find that the early convolutional layers contain representations that are generalizable across tasks. Moreover, our approach significantly increases mean model accuracy from 82.33% to 86.99%, increases the models use of lower frequencies, ({theta}-band), and increases its robustness to channel loss. We expect this analysis to provide useful guidance and enable more widespread use of transfer learning in EEG deep learning studies.
Matching journals
The top 8 journals account for 50% of the predicted probability mass.
Similar papers in this journal
- Somnotate: A probabilistic sleep stage classifier for studying vigilance state transitions 93%
- Circadian distribution of epileptiform discharges in epilepsy: candidate mechanisms of variability 92%
- A hidden Markov model reliably characterizes ketamine-induced spectral dynamics in macaque LFP and human EEG 92%
Similar papers in this journal
- Off-body Sleep Analysis for Predicting Adverse Behavior in Individuals with Autism Spectrum Disorder 92%
- Time-adaptive Unsupervised Auditory Attention Decoding Using EEG-based Stimulus Reconstruction 92%
- Hyperbolic graph embedding of MEG brain networks to study brain alterations in individuals with subjective cognitive decline 91%
Similar papers in this journal
- On the development of sleep states in the first weeks of life 94%
- Discrimination of sleep and wake periods from a hip-worn raw acceleration sensor using recurrent neural networks 93%
- Rett syndrome severity estimation with the BioStamp nPoint using interactions between heart rate variability and body movement 93%
"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.