Non-contact dual pulse Doppler system based respiratory and heart rates estimation for CHF patients

Abstract

Long term continuous patient monitoring is required in many health systems for monitoring and analytical diagnosing purposes. Most of monitoring systems had shortcomings related to their functionality or patient comfortably. Non-contact continuous monitoring systems have been developed to address some of these shortcomings. One of such systems is non-contact physiological vital signs assessments for chronic heart failure (CHF) patients. This paper presents a novel automated estimation algorithm for the non-contact physiological vital signs assessments for CHF patients based on a patented novel non-contact biomotion sensor. A database consists of twenty CHF patients with New York Heart Association (NYHA) heart failure Classification Class II & III, whose underwent full Polysomnography (PSG) analysis for the diagnosis of sleep apnea, disordered sleep, or both, were selected for the study. The patients mean age is 68.89 years, with mean body weight of 86.87 kg, mean BMI of 28.83 (obesity) and mean recorded sleep duration of 7.78 hours. The propose algorithm analyze the non-contact biomotion signals and estimate the patients' respiratory and heart rates. The outputs of the algorithm are compared with gold-standard PSG recordings. Across all twenty patients' recordings, the respiratory rate estimation median accuracy achieved 92.4689% with median error of ± 1.2398 breaths per minute. The heart rate estimation median accuracy achieved 88.0654% with median error of ± 7.9338 beats per minute. Due to the good performance of the propose novel automated estimation algorithm, the patented novel non-contact biomotion sensor can be an excellent tool for long term continuous sleep monitoring for CHF patients in the home environment in an ultra-convenient fashion.

Doppler Radar-Based Non-Contact Health Monitoring for Obstructive Sleep Apnea Diagnosis: A Comprehensive Review

“…Studies have reported some commendable achievements, such as error of less than 0.5 breath per minute for respiratory rate and one beat per minute for heart rate [26,27,61]. In addition, the achieved accuracy of Doppler radar in measuring physiological parameters, such as respiratory rate is 92% [39], heart rate is 88% [39,59,81] to 91% [40], and with the addition of harmonics interference, the average error can be reduced further to 3.2% [84]. However, conventional FFT may not always be able to reliably separate the rich sinusoidal components due to smearing and leakage problems, particularly from the limited data samples [74].…”

“…Besides clutters, other issues such as DC offsets, phase-nulling contributions, motions artefacts and electromagnetic interferences are also posing as challenges to the signal processing of Doppler radar signals [13,57]. Another major challenge in sleep monitoring application is the noise associated with unpredictable body movements, body orientations, changes in sleeping posture, multi-subjects cancellation, undesired harmonics and intermodulation [14,15,39].…”

“…The ranges of the reported filter bandwidth frequencies for respiratory rate are from 0.1-0.5 Hz which is equivalent to 6-30 breaths per minute. The heart rate frequencies are from 0.8-2.0 Hz which is equivalent to 48-120 beats per minute [13,26,32,39,40,59,61,63,65,69].…”

“…The use of Wavelet Transform (WT) [87], Wavelet Filter [88], Wavelet Packet Decomposition (WPD) [39], Discrete Wavelet Transform (DWT), and Complex Wavelet Transform (CWT) with Morlet mother wavelet [29] have also been proposed in literature for respiratory and heart rate estimations. The advantage of Wavelet analysis over conventional methods, such as FFT, is that Wavelet is designed to operate on non-stationary signals and retains time and frequency information, providing good frequency resolution at lower frequencies.…”

Non-contact Doppler radar based prediction of nocturnal body orientations using deep neural network for chronic heart failure patients

“…The mother wavelet for WPD is Symlet wavelet with 4 th order filter. The reason for choosing the Symlet wavelet is for the performance and accuracy as detailed in the authors' previous work [5].…”