Abstract
DeepSleep 2.0 is a compact version of DeepSleep, a state-of-the-art, U-Net-inspired, fully convolutional deep neural network, which achieved the highest unofficial score in the 2018 PhysioNet Computing Challenge. The proposed network architecture has a compact encoder/decoder structure containing only 740,551 trainable parameters. The input to the network is a full-length multichannel polysomnographic recording signal. The network has been designed and optimized to efficiently predict nonapnea sleep arousals on held-out test data at a 5 ms resolution level, while not compromising the prediction accuracy. When compared to DeepSleep, the obtained experimental results in terms of gross area under the precision–recall curve (AUPRC) and gross area under the receiver operating characteristic curve (AUROC) suggest a lightweight architecture, which can achieve similar prediction performance at a lower computational cost, is realizable.
Highlights
The main aim of this paper is to realize an efficient and automatic nonapnea sleep arousal segmentation method based on deep learning methods
Ŷi corresponds to a predicted sleep arousal probability, i.e., ŷi ∈ [0, 1], or to a sleep arousal state, i.e., ŷi ∈ {0, 1}, at time instance ti = i∆t, and is associated with xi ∈ RC, a slice from the overall recording signal x = [ x1, . . . , xi, . . . , xS ]
DeepSleep is able to perform automatic sleep arousal segmentation with a less-complex architecture, fewer parameters, and achieving close to similar area under the precision–recall curve (AUPRC) and area under the receiver operating characteristic curve (AUROC) scores on the held-out test data considered in this work
Summary
T ∈ Z+ stands for the total length of the recorded signal in seconds and ∆t > 0 is the per-second sampling resolution. The aim is to find a model f ( x, θ) which maps the input signal x into the prediction space ŷ = [ŷ1 , . Ŷi corresponds to a predicted sleep arousal probability, i.e., ŷi ∈ [0, 1], or to a sleep arousal state, i.e., ŷi ∈ {0, 1}, at time instance ti = i∆t, and is associated with xi ∈ RC , a slice from the overall recording signal x = [ x1 , . I ∈ R is a discrete time instance and R = {1, . S} represents a set of all discrete time instances of the record x
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
