A Machine Learning Approach to Automatic Classification of Eight Sleep Disorders

Abstract

This research focuses on automatically classifying common sleep disorders. We attempt to answer the following basic research questions: Is a machine learning model able to classify all types of sleep disorders with high accuracy? Among the different modalities of sleep disorder signals, are some more important than others? Do raw signals improve the performance of a deep learning model when they are used as inputs? Prior research showed that most sleep disorders belong to eight categories (for instance, the PhysioBank dataset). To study the performance of machine learning models in classifying polysomnography recordings into the eight categories of sleep pathologies, we selected the Cyclic Alternating Pattern (CAP) Sleep Database, a collection of 108 polysomnographic recordings. We developed a multi-channel Deep Learning (DL) model where a set of Convolutional Neural Networks (CNNs) were applied to six channels of raw signals of different modalities, including three channels of EEG (lectroencephalogram) signals and one channel each of EMG (electromyography), ECG (electrocardiogram), and EOG (electrooculogram) signals. To compare the performance of the DL model with other models, we designed a model that took spectral features, instead of raw signals, as its inputs. We named the former the DL-R model and the latter the DL-F model. For comparison, we also designed two conventional classifiers, Random Forest (RF) and Support Vector Machines (SVM), which also took spectral features as inputs. We first studied the "importance" issue of signal modalities using the RF algorithm. We found that ECG contributed most to the important features and EMG second, among the four signal modalities. We then studied the accuracy performance of the proposed machine learning models. We verified that the multi-channel DL-R model, which took raw signals as its inputs, outperformed all other models, with its sensitivity and specificity scores both being above 95% across all the eight sleep disorders. This accuracy performance is on a par with those published results which dealt with fewer types of sleep disorders. To study the explanability of the DL-R model, we adopted two popular heatmap-generating techniques, with which we confirmed that the DL model's superior performance was owing to the CNN network's ability to extract potent features from raw signals. Its heatmaps produced frequency ranges and peak frequencies of various sleep disorders consistent with the results from clinic studies. We hope that the proposed approach is one step closer to more able and trustworthy machine learning techniques that one day will be adopted by practitioners.

Simultaneous Sleep Stage and Sleep Disorder Detection from Multimodal Sensors Using Deep Learning

“…Note that [ 19 ] classified five sleep stage categories, whereas our approach used six categories. Similarly, in sleep disorder classification, our approach outperformed the best results of [ 20 ] by 4% (MML-DMS1 and MML-DMS2).…”

“…While there is a relatively large body of research on sleep stage detection, research into sleep disorder classification has resulted in a smaller number of publications. Zhuang and Ibrahim [ 20 ] developed a multi-channel Deep Learning (DL-AR) model where a set of CNNs was applied to six channels of raw signals of different modalities, including three channels of EEG (electroencephalogram) signals and one channel each of EMG (electromyogram), ECG (electrocardiogram), and EOG (electrooculogram) signals. The model was tested on the PhysioNet CAP Sleep database [ 18 , 21 ], yielding specificity and sensitivity scores of around 95% for eight sleep disorders.…”