Cheyne-Stokes Respiration Perception via Machine Learning Algorithms

With the development of science and technology, transparent, non-invasive general computing is gradually applied to disease diagnosis and medical detection. Universal software radio peripherals (USRP) enable non-contact awareness based on radio frequency signals. Cheyne-Stokes respiration has been reported as a common symptom in patients with heart failure. Compared with the disadvantages of traditional detection equipment, a microwave sensing method based on channel state information (CSI) is proposed to qualitatively detect the normal breathing and Cheyne-Stokes breathing of patients with heart failure in a non-contact manner. Firstly, USRP is used to collect subjects’ respiratory signals in real time. Then the CSI waveform is filtered, smoothed and normalized, and the relevant features are defined and extracted from the signal. Finally, the machine learning classification algorithm is used to establish a recognition model to detect the Cheyne-Stokes respiration of patients with heart failure. The results show that the system accuracy of support vector machine (SVM) is 97%, which can assist medical workers to identify Cheyne-Stokes respiration symptoms of patients with heart failure.