Automated sleep staging using wearable sensors
A method and system for automated sleep staging are disclosed. The method comprises determining at least one physiological signal during a predetermined time period, extracting at least one feature from the at least one physiological signal, and classifying the at least one feature using a machine learning classifier to output at least one sleep stage. The system includes a sensor to determine at least one physiological signal during a predetermined time period, a processor coupled to the sensor, and a memory device coupled to the processor, wherein the memory device includes an application that, when executed by the processor, causes the processor to extract at least one feature from the at least one physiological signal and to classify the at least one feature using a machine learning classifier unit to output at least one sleep stage.
1. A wireless sensor device for automated sleep staging, the wireless sensor device comprising:
a sensor to determine at least one physiological signal;
a processor coupled to the sensor; and
a memory device coupled to the processor, wherein the memory device includes an application that, when executed by the processor, causes the processor to:
extract at least one feature from each of a plurality of epochs of the at least one physiological signal;
scale the at least one feature for each epoch;
determine a feature vector for each epoch using the at least one scaled feature; and
classify each epoch into one of a plurality of sleep stages by inputting each feature vector into a machine learning classifier unit, wherein the at least one scaled feature enables the machine learning classifier unit to classify epochs with features of varying magnitudes.
2. The wireless sensor device of claim 1 , wherein the application, when executed by the processor, further causes the processor to:
adjust the at least one outputted sleep stage based on temporal dynamics.
3. The wireless sensor device of claim 1 , wherein the at least one physiological signal includes any of an electrocardiogram (ECG) signal, a microelectromechanical system (MEMS) accelerometer signal, a skin temperature signal, and an electroencephalography (EEG) signal.
4. The wireless sensor device of claim 1 , wherein the predetermined time period is every four minutes over eight continuous hours and wherein the at least one sleep stage is any of W, N 1 , N 2 , N 3 , and REM sleep stages.
5. The wireless sensor device of claim 1 , wherein the machine learning classifier unit is located on any of the processor of the wireless sensor device and a remote server, further wherein the machine learning classifier unit includes any of a support vector machine (SVM) unit, a Naïve Bayes classifer (NBC) unit, and an artificial neural network (ANN) unit.
6. The wireless sensor device of claim 1 , wherein the processor includes both the machine learning classifier unit and an algorithm unit, further wherein the machine learning classifier unit is a support vector machine (SVM) unit.
7. The wireless sensor device of claim 6 , wherein the application, when executed by the processor, further causes the processor to:
output a plurality of class probabilities via the support vector machine (SVM) unit; and
input the plurality of outputted class probabilities into the algorithm unit that adjusts the plurality of outputted class probabilities based on temporal dynamics.
8. The wireless sensor device of claim 6 , wherein the algorithm unit utilizes a Hidden Markov Model (HMM) algorithm that considers temporal dynamics of stage transitions to smooth the plurality of outputted class probabilities.
9. The wireless sensor device of claim 1 , wherein the application, when executed by the processor, further causes the processor to:
train the machine learning classifier unit using a data set of sleep stages derived from a polysomnography (PSG) study.