Sleep studies are important for diagnosing sleep disorders such as insomnia,narcolepsy or sleep apnea. They rely on manual scoring of sleep stages from rawpolisomnography signals, which is a tedious visual task requiring the workloadof highly trained professionals. Consequently, research efforts to purse for anautomatic stage scoring based on machine learning techniques have been carriedout over the last years. In this work, we resort to multitaper spectralanalysis to create visually interpretable images of sleep patterns from EEGsignals as inputs to a deep convolutional network trained to solve visualrecognition tasks. As a working example of transfer learning, a system able toaccurately classify sleep stages in new unseen patients is presented.Evaluations in a widely-used publicly available dataset favourably compare tostate-of-the-art results, while providing a framework for visual interpretationof outcomes.
展开▼
