A Method for Sleep Quality Analysis Based on CNN Ensemble With Implementation in a Portable Wireless Device

Abstract

The quality of sleep can be affected by the occurrence of a sleep related disorder and, among these disorders, obstructive sleep apnea is commonly undiagnosed. Polysomnography is considered to be the gold standard for sleep analysis. However, it is an expensive and labor-intensive exam that is unavailable to a large group of the world population. To address these issues, the main goal of this work was to develop an automatic scoring algorithm to analyze the single-lead electrocardiogram signal, performing a minute-byminute and an overall estimation of both quality of sleep and obstructive sleep apnea. The method employs a cross-spectral coherence technique which produces a spectrographic image that fed three one-dimensional convolutional neural networks for the classification ensemble. The predicted quality of sleep was based on the electroencephalogram cyclic alternating pattern rate, a sleep stability metric. Two methods were developed to indirectly evaluate this metric, creating two sleep quality predictions that were combined with the sleep apnea diagnosis to achieve the final global sleep quality estimation. It was verified that the quality of sleep of the nineteen tested subjects was correctly identified by the proposed model, advocating the significance of clinical analysis. The model was implemented in a non-invasive and simple to self-assemble device, producing a tool that can estimate the quality of sleep and diagnose the obstructive sleep apnea at the patient's home without requiring the attendance of a specialized technician. Therefore, increasing the accessibility of the population to sleep analysis.

Heuristic Optimization of Deep and Shallow Classifiers: An Application for Electroencephalogram Cyclic Alternating Pattern Detection

“…For the CNN classifier, the model with one dimension was selected since it can identify relevant patterns from challenging one-dimensional biomedical signals, using a small number of neurons and hidden layers [16,[38][39][40]. The small networks are easier to train and implement, requiring less computational resources to develop the algorithm [14].…”

“…For the FFNN optimization, the number of neurons employed for the hidden layer was varied from 100 to 400, in steps of 100. On the other hand, the Heuristic Oriented Search Algorithm (HOSA) employed in this work follows the concepts presented by Mendonça et al [40] and Mostafa et al [45], for the LSTM or 1D-CNN optimization to assess the most relevant architecture for the classifiers by considering a heuristic search for the parameters considered to be the most relevant for the examined models.…”

“…The algorithm started without overlapping, and the duration of overlapping was increased in steps of four seconds up to a maximum window, O max , of 35 s (the upper limit was empirically found to be above the saturation point for the best A phase AUC). The searching procedure was improved by using the group of layers concept (GofLayer) [40], where each group was composed of one convolution layer, followed by one subsampling layer, and a 10% dropout was applied at the output of the group. A downsample of factor two was applied in the subsampling layers with the chosen stride and filter size of two.…”

L-Tetrolet Pattern-Based Sleep Stage Classification Model Using Balanced EEG Datasets

“…The CAP Sleep Database on PhysioNet [ 24 ] is widely used in scientific work on sleep staging, and most of the published studies use the CAP database to establish the sleep phase [ 25 , 26 , 27 , 28 , 29 , 30 , 31 , 32 , 33 ]. Table 1 summarizes selected studies on sleep stage detection using different datasets.…”