WO2022136909A1

WO2022136909A1 - Wearable cough sensor system for patient monitoring

Info

Publication number: WO2022136909A1
Application number: PCT/IB2020/062449
Authority: WO
Inventors: Neda SHAFIABADI HASSANI; Shiva GHARAEI; Abdolrasool ESMAILZADEH; Gholamreza AKBARZADEH
Original assignee: Shafiabadi Hassani Neda
Priority date: 2020-12-24
Filing date: 2020-12-24
Publication date: 2022-06-30

Abstract

This disclosure aims to introduce a low-cost portable monitoring system that can continuously measure the relevant metrics and detect cough for respiratory illnesses such as COVID-19 infection, including: Peripheral capillary oxygen saturation (SpO2), Cough occurrences, Heart rate, and Body temperature. As for the cough detection, the proposed design combines signals from an accelerometer and an audio sensor to automatically detect cough. Cost and accuracy are two competing factors in the development of wireless monitoring systems, with accurate products having a high cost affecting their potential for large scale usage. One distinguishing feature of our approach is minimizing costs by using simple hardware while maintaining high accuracy, partly by developing a powerful signal processing system based outside of the wearable part.

Description

wearable cough sensor system for patient monitoring

Technical Field

[0001] The invention relates to medical monitoring and more specifically monitoring/detecting coughs.

Background Art

[0002] Peripheral capillary oxygen saturation (SpO2), respiratory rate (RR), body temperature and heart rate are used to assess a patient's health. Changes in these metrics are used by healthcare professionals as indicators of alteration in the clinical situation of the patients, and if not caught in time may lead to egregious consequences. With some exceptions (e.g. in an ICU), these metrics are not measured continuously in a healthcare facility nor at home. Consequently, any patient in need of continuous monitoring of the vital signs should currently be admitted to a hospital, urging the health care system to select only those at a more critical situation for admission, essentially depriving the other patients from necessary monitoring. This is especially relevant to the COVID-19 pandemic, where rapidly increasing numbers of patients and long hospitalization durations are imposing an extreme workload on the healthcare system, and where the clinical course of the patients is not predictable from the outset and is oftentimes associated with rapid deterioration.

[0003] Function of the respiratory system is usually monitored using RR and SpO2. In infectious diseases such as COVID-19 infection, body temperature is another important sign used in medicine for evaluating the patient’s condition. Likewise, cough is a major debilitating factor that affects the patients’ quality of life during the illness period. It is present in % of COVID-19 patients at presentation and in most of those with severe disease. An effective monitoring of these 4 indices will allow the health care system to remotely, but continuously monitor the clinical status of the patients.

[0004] There are many software applications on smartphones and smart watches claimed to monitor SpO2, heart rate and respiratory rate. However, almost none has been approved for medical use due to lack of measurement validity and reliability. Having the ability to accurately measure important vital metrics in a small, low-cost package, this product will be a unique one in the market in terms of design and functionality with definite potential for medical degree usage. In addition to COVID-19 cases, thousands of other respiratory patients with a need for chronic care can be continuously, accurately, and cheaply monitored using this device. The capability to monitor the coughing occurrences is quite novel.

[0005] Most available portable monitoring systems in the market rely on an embedded microcontroller to carry out calculations. Such an approach not only makes the device heavy, but also prevents using advanced and highly accurate algorithms due to relatively low memory, battery life, and mathematical capacities of the hardware. Therefore, in the proposed design, most of the computational operations will be carried out on a nearby base device to improve accuracy and affordability. There is almost unlimited capacity for calculation and processing in the base device, with no added cost for the hardware. Therefore, the proposed device will be much more accurate, much smaller and easier-to-wear, and much less costly. In spite of the presence of numerous articles on using different sensors for detecting coughs, there is a lack of a commercial cough sensor in the market. As cough severity is a relevant factor in clinical decision making for COVID-19 and many other respiratory pathologies, using an accelerometer in addition to an audio sensor for monitoring cough occurrences will be of clinical advantage. Moreover, using modern wearable patches which offer a secure attachment to the body, the amount of noise signals detected by an accelerometer is much lower than those detected by an audio sensor, and the accelerometer data can be used to remove the noise detected by the audio sensor. Also, with using advanced machine-learning techniques, the accelerometer noises are amenable to accurate differentiation from the body motions caused by a cough.

[0006] There have been several methods for detection/monitoring of cough. The most accurate method of detecting cough is by one or more trained observers who are either present in person with the index patient, or are listening and observing a video tape recording of the patient. As such, this method is often used as a reference ("gold standard") for validation of an automated cough detection device. Automated cough detection has been attempted and the following is a brief description of the available prior art. These methods may be divided into those that use the cough sound alone and those that use the cough sound in combination with other signals. Cough detection by loud-sound detection and recording on a sound-activated tape (or digital) recorder have been used in research by Mori et Al. [1] . The timing of the tape activation may also be recorded with each loud sound. The total recording time, or the number of recorder activations are used as a cough activity index. This method is sensitive, but lacks specificity since other loud noises in the environment may trigger the recording as well. Subsequent auditory screening of the recorded sounds by a trained observer, or an automated algorithm may improve the specificity of this method.

[0007] Earis and co-workers [2] described digital signal processing methods including spectral analysis and voice analysis methods to evaluate loud sounds to differentiate between cough sounds and other sounds such as vocalization. These studies as well as studies by other researchers analyzed ambient sounds recorded with a single microphone.

[0008] Other methods to detect coughs use two or more signals. Munyard et al [3], described a method by which a microphone output is used in combination with the signals from Electromyograph (EMG) electrodes placed on the abdominal muscles. These muscles contract in order to generate the elevated expiratory intrathoracic pressure needed for an effective cough. Gavriely N. described in US patents 6,261 ,238 and 6,168,568 a method in which the loud output of a microphone alongside a simultaneous sudden motion of the chest detected by a chest motion detector such as an electrical impedance plethysmograph are used as the first phase of cough detection algorithm.

Summary of Invention

[0009] This disclosure aims to introduce a low-cost portable monitoring system that can continuously measure the relevant metrics and detect cough for respiratory illnesses such as COVID-19 infection, including: Peripheral capillary oxygen saturation (SpO2), Cough occurrences, Heart rate, and Body temperature. As for the cough detection, the proposed design combines signals from an accelerometer and an audio sensor to automatically detect cough.

[0010] Cost and accuracy are two competing factors in the development of wireless monitoring systems, with accurate products having a high cost affecting their potential for large scale usage. One distinguishing feature of our approach is minimizing costs by using simple hardware while maintaining high accuracy, partly by developing a powerful signal processing system based outside of the wearable part.

[0011 ] To overcome the shortcomings of the health-care system during the pandemics, we have designed a remote sensor system for monitoring the nonhospitalized patients. Real-time signals on vital signs and cough occurrences will be captured and transmitted to a smartphone for processing and then to the care providers. This will offer inexpensive but accurate monitoring of an unlimited number of COVID- and other respiratory patients for unlimited durations. Its unique features include the ability to detect coughs and to process the signals out of the sensor package, the latter of which will provide high accuracy and user- friendliness, high battery life and low cost.

Technical Problem

[0012] There is no wearable device that could be comfortably used by the patients with COVID-19 disease or any other respiratory disease in need of continuous monitoring.

[0013] Coughs are common symptoms of many respiratory diseases yet difficult to analyse. Automatic detection of cough sound requires solving at least four major problems:

[0014] (1 ) ambient noise filtering;

[0015] (2) distinguishing from patient sounds, for instance, laughter, speech, and sneezing;

[0016] (3) the variability of cough acoustics both within and between individuals;

[0017] (4) differentiation of dry and wet coughs. [4]

Solution to Problem [0018] To remove ambient noise and distinguishing cough from other patient-sounds more effectively, besides the microphone an accelerometer is also utilized. The accelerometer can detect the chest abrupt movements caused by coughing and then, the system uses this information to process voice signals.

[0019] The strategy to carry out computations mostly on the mobile set or tablet enhances the computational abilities of the device, and meanwhile, reduces the cost and power consumption of the portable section.

[0020] Example 1

[0021] Our device can be utilized for monitoring patients who suffer from respiratory problems, for example, those who have developed COVID-19 and are quarantined at home. It can monitor cough events and other vital signs continuously and let the health center have access to the data.

[0022] Example 2

[0023] Another potential application for the device is for people who suffer from seasonal allergies triggered by airborne pollen. The decision about effective medicine sometimes becomes too challenging, and it differs from case to case. Since, to evaluate the effectiveness of any prescribed drugs, we need to gather data for at least several days, the device can track the effect of a recipe uninterruptedly and provide the physician with enough statistics to reach a conclusion.

Advantageous Effects of Invention

[0024] Most available portable monitoring systems in the market rely on the embedded microcontroller to carry out calculations. Such an approach not only makes the device very costly but also prevents using developed and high- accuracy algorithms mainly due to the relatively low memory, battery, and mathematical capacities of the hardware. To afford the computational costs required by such advanced methods on one hand, and reduce the device price, on the other hand, we developed a set-up consisting of two parts. The first part, which is a portable adhesive board, records required signals. Since we expect this part to carry out simple preprocessing steps on the data, its cost would be very affordable. The acquired data, then, will be sent to a nearby device such as a mobile, tablet, or laptop to carry out advanced signal processing methods for cough monitoring.

[0025] Since our product enjoys two main advantages of low cost and high accuracy, it interests many people with and without serious health issues. The product also can be utilized to send crucial data to the physician.

Brief Description of Drawings

[0026] Fig. 1 : Hardware Description

[0027] Fig. 2: Signal Processing Flow Chart

[0028] Fig. 3: Illustration of the wearable monitoring device and its location on patient’s body

Description of Embodiments

[0029] Figure 1 : depicts the schematic of the proposed system. The wearable patch will be attached to the chest somewhere next to the left serratus anterior muscle. The signals collected by the portable device will be transmitted via Bluetooth signals to the base devices (mobile, tablet, or laptop) to be processed, using advanced signal processing methods and machine learning techniques. We will develop proper software applications for various operating systems such as IOS, Android, and windows to carry out estimation procedures based on the acquired data.

[0030] Figure 2: indicates various steps to detect and classify cough signals. After applying appropriate signal preparation steps, we will consider time-domain, frequency-domain, or time-frequency domain representations to extract features from cough sound signals using machine learning techniques.

[0031 ] The device is able to monitor cough events automatically all the time. It is because of its ability to distinguish cough sounds from environmental noise properly.

Industrial Applicability

[0032] The novel cough detection technology is integrated with other portable monitoring devices to add more facilities like blood pressure, respiratory rate, heart rate, and body temperature measurements. This device will offer continuous monitoring of countless COVID-19 cases, and to chronically ill patients who presently have little access to healthcare at home, especially with the growing senior population.

[0033] Decreasing inpatient volume during the pandemic prevents virus exposure and minimizes the need for PPE, reducing environmental pollution. It will improve access to crucial resources by increasing our limited monitoring capacity and ease tensions for ICUs with potential to save many lives. It will be of particular benefit in our remote and underserved populations such as indigenous communities, decreasing disparities in quality of care and patient outcomes.

Reference Signs List

[0034] [1 ] Mori A, Nishino C, Enoki N, Tawata S (1987). Antibacterial activity and mode of action of plant flavonoids against Proteus vulgaris and Staphylococcus aureus. Phytochemistry, 26(8): 2231 -2234.

[0035] [2] Smith JA, Ashurst HL, Jack S, Woodcock AA, Earis JE. The description of cough sounds by healthcare professionals. Cough. 2006 Jan 25;2:1

[0036] [3] Munyard P, Busst C, Logan-Sinclair R, Bush A. A new device for ambulatory cough recording. Pediatr Pulmonol. 1994 Sep;18(3):178-86

[0037] [4] Shi, Y„ Liu, H„ Wang, Y„ Cai, M. and Xu, W„ 2018. Theory and application of audio-based assessment of cough. Journal of Sensors, 2018.

Claims

8 Claims

[Claim 1 ] wearable cough sensor system for patient monitoring comprising: a. Wearable Part b. Smartphone

[Claim 2] According to claim 1 , the wearable patch will be attached to the chest somewhere next to the left serratus anterior muscle and the signals like Peripheral capillary oxygen saturation (spo2), Cough occurrences, Heart rate, and Body temperature and for cough detection, collected by the portable device will be transmitted via Bluetooth signals to the base devices to be processed, using advanced signal processing methods and machine learning techniques.

[Claim 3] According to claim 1 and 2, after applying appropriate signal preparation steps, time-domain, frequency-domain will be considered, or timefrequency domain representations to extract features from cough sound signals using machine learning techniques.

[Claim 4] According to claim 1 to 3, to remove ambient noise and distinguishing cough from other patient-sounds more effectively, in the wearable part besides the microphone an accelerometer is also utilized that the accelerometer can detect the chest abrupt movements caused by coughing and then, the system uses this information to process voice signals.

[Claim 5] According to claim 1 to 4, decision trees, dimension reduction, and clustering algorithms will be applied to extracted features and the potential of artificial neural networks, including back-propagation, radial basis function networks, and convolutional neural networks will be investigated as efficient approaches for the characterization of cough sounds.