Among the large commercial offer of different models of spirometers, handheld spirometry devices are those that are most suitable for telemonitoring applications. Modern models are connected to a gateway, usually a smartphone. The main drawbacks of these devices are that only spot assessments (e.g., once per day) are possible; furthermore, these devices require patient collaboration and analyze specific maneuvers, not spontaneous breathing. In general, assessment of breathing through a flowmeter is not feasible for continuous monitoring as it requires the use of a mouthpiece, a mask or a tube, therefore it is uncomfortable to wear and not usable for a prolonged time outside of clinical or laboratory settings.4

Forced oscillation technique (FOT) is a noninvasive method to assess lung function during normal breathing by using a small pressure oscillation (equivalent to the one exerted by 2cm of water) imposed on regular quiet breathing to “force” a small quantity of air to move in and out the respiratory system. Thanks to this, the mechanical properties of the lung can be measured without supervision and also in home contexts; from resistance and reactance measurements, information on the breathing pattern can be obtained.5 A product that can be used in home settings is Resmon Pro Diary by Restech, which was used in a clinical trial on COPD patients.6

Another approach for measuring breathing consists in deriving breathing parameters from body surface (namely, chest wall) motion detection, i.e. by inferring changes in thoracic volume from geometrical changes at discrete locations on the torso. Systems for continuous monitoring of breathing can use either wearables (e.g., respiratory inductance plethysmography, resistance-based sensors, capacitance-based sensors, inertial measurement units, fiber optic sensors…) or non-wearable devices (e.g. mechanical ventilators, long-term oxygen therapy, polysomnography…).

• a

  Systems using wearable devices

Respiratory Inductance Plethysmography (RIP). RIP involves the use of two transducer bands placed around the subject to monitor excursions of the chest and abdomen over time. As the subject breathes, the volumes of the two compartments change, and these changes are reflected in alterations in the self-inductance of the coils. One drawback is that slippage of bands can be a problem in RIP, leading to inaccurate readings. This technique also requires direct contact with the subject during placement and positioning of the bands.7 The LifeShirt® by Vivometric® was an example of device that included RIP to measure ventilatory parameters,8 but nowadays is no longer produced. Hexoskin®, on the other hand, is currently available on the market and bases its respiratory rate and minute ventilation measurements on the use of RIP. This device was validated in multiple studies that involved tests in different postures and with incremental levels of exercise; Hexoskin® volume measurements have low variability and good agreement and consistency.9

Resistance-based wearable sensors. Several wearable sensors are based on the principle that the resistance measured varies with the movement of the torso. An example is L.I.F.E.’s medical compression garment composed of 12 electrodes for ECG monitoring, 5 respiratory strains sensors and 1 accelerometer. The respiratory sensors are based on the variation of the resistance and are positioned on the anterior external surface of the garment as follows: two are thoracic, one is central and two are abdominal.10 Another wearable respiratory monitor is AirGo™, which measures the thoracic circumference changes with a stretchable knitted matrix of nylon and spandex with a built-in silver coated yarn. The system takes the form of an electrically conductive chest band that encircles the lower chest and ribs of a patient or individual. The on-board microprocessor is also equipped with a three-axial accelerometer for motion detection.11 There are also some recent projects with a quantitative estimation of volumes from wearable sensors. In one study, measurements of respiratory rate and tidal volume using a disposable Band-Aid®-like sensor have been evaluated. This sensor allows us to derive the two aforementioned parameters by simply measuring the local strain of the ribcage and abdomen during breathing.12 While devices like L.I.F.E.’s garment and AirGo™ are both placed around the torso, this type of sensor has the advantage of being a patch, hence less invasive and obtrusive.

Capacitance-based wearable sensors. Another field of research and development regards sensors based on variations of capacitance. In a study, the variations in electrical capacity between two conductive wires integrated in an elastic garment was employed to obtain the respiratory pattern.

Approaches to the design of portable devices for non-contact monitoring of respiratory rate by capacitive sensing have also been developed: this technology is based on the measurement of the capacitance existing between two electrodes between which the thoracic tissue acts as a dielectric material. The mechanical changes produced by breathing cause variations in the capacitance, which can be correlated with the respiratory rate.13 Respiratory sensors with a single electrode have been studied, but are not yet validated and must be considered as future developments: in this case, the capacitance is constructed between one electrode attached on the skin and another constituted by the electrolytes inside, i.e., the conductive body fluid. At exhalations, the body volume decreases and the skin contrasts increasing its thickness, therefore the capacitance changes; the same principle can be applied to analyse inhalations.14

Inertial measurement units. The processed information captured by an accelerometer can be used to derive the respiratory rate from the movements of the rib cage; several studies with this type of solution can be found.15 An example of this type of device is RespirHò, composed of three IMU-sensor units communicating via Bluetooth with a smartphone. Two units record chest-wall respiratory-related movements and are placed on the thorax and on the abdomen; the third sensor unit is placed on a body area that is not involved in respiratory movement.16,17

Fiber optic sensors. Recently, smart textiles based on fiber optic sensors have shown promising results in the field of respiratory monitoring. Especially fiber Bragg grating (FBG) sensors have been used to monitor the respiratory rate by integration of the sensing elements in patches or garments.

The deformations of the thorax and the abdomen can indirectly estimate the contribution of the thoracic and abdominal compartments on the tidal volume, because the reflected wavelengths sensed by the FBGs are affected by strain and temperature. As an example, a research group in Rome has recently developed a garment based on 12 FBGs placed in specific body landmarks.18

• a

  Systems using non-wearable devices

Typically, continuous monitoring of breathing is performed with unobtrusive wearable devices, however there are specific cases in which measurements can be performed with more cumbersome techniques. In the case of Long-Term Oxygen Therapy (LTOT), there are commercial devices that calculate the respiratory rate via sensors that detect pressure changes in the oxygen line. Another case is polysomnography (PSG), which consists of a simultaneous recording of multiple physiologic parameters related to sleep and wakefulness. In standard PSG, at least seven signals are required: electroencephalography (EEG), electrooculography (EOG), chin electromyography (EMG), ECG, an airflow measurement using thermistors and nasal pressure transducers, pulse oximetry and respiratory effort (thoracic and abdominal).19 In some studies on sleep disorders, portable testing can be limited to ECG or heart rate, oxygen saturation and at least two channels of respiratory movement or respiratory movement combined with airflow.20 Finally, non-invasive mechanical ventilation (NIV) can be performed with Home Care Ventilators (HCVs), where a monitoring system can be embedded. In ventilators, directly measurable variables are the pressure at the airways opening and the flow; subsequently, other parameters are derived from the analysis of these signals. The accuracy of these measurements is reduced by the presence of leaks, which may be intentional or unintentional, and may cause discomfort and a decrease in compliance. Some ventilators allow continuous pulse oximetry monitoring and even end-tidal carbon dioxide (ETCO2) monitoring. In particular, ETCO2 is referred to as capnography and has been clinically demonstrated to provide an early indicator of respiratory distress.21

Some models of HCV are equipped with built-in software to download and analyse data over the previous weeks, providing compliance, estimation of leaks, tidal volume, minute ventilation, respiratory rate, apnoea and apnoea-hypopnea indexes and percentage of inspirations triggered by the patient (or by the ventilator).22

• a

  Contactless methods

A promising field of study is the development of telemonitoring techniques that do not require contact, even though a main drawback is that these devices are usually in fixed positions, as is further illustrated in the section dedicated to activity recognition, and therefore their range of utilization is limited. For instance, Doppler bio-radars have the ability to monitor physiological signals that depend on the Doppler effect. Electromagnetic waves can travel to the surface of the chest cavity and the reflected wave contains information on chest displacement, mainly caused by breathing.23

Linear accelerations and angular/rotational velocities can be detected via micro-electro-mechanical systems (MEMS), which measure either capacity changes or the deflection of magnetically excited comb structures.

Accelerometers. At present, movement registration with body-fixed motion sensors offers the best alternative for physical activity assessment. Their use is based on demonstrated relationships between accelerometer output and energy expenditure in studies on gait analysis and ergonomics. Miniaturized accelerometers are able not only to quantify activities like walking, running, cycling and exercising, but also to recognise the time spent under sedentary conditions and distinguish different postures through the detection of gravity acceleration.30 There are several activity monitors that are solely based on accelerometers and commercially available, for instance the ones produced by Philips Respironics, ActiGraph LLC (current models are CentrePoint Insight Watch, Actigraph GT9X Link, WGT3X-BT31; other monitors combine accelerometers with other type of sensors such as heat flux, galvanic skin response and skin temperature, like the SenseWear Armand by Bodymedia.32

Gyroscopes. In several studies, researchers combined accelerometers with gyroscopes to perform fall detection, gait analysis and gesture recognition. In most circumstances, the accelerometer acts as the lead sensor while the gyroscope functions as the supplementary sensor. Activities such as standing and sitting are better recognized from accelerometer data, while walking upstairs and downstairs are activities better recognized from gyroscopic data. For walking, biking and jogging activities accelerometer data performs slightly better than gyroscope data.33

Magnetometers. While accelerometers and gyroscopes are the most common types of sensors used in HAR, the possibility of using magnetometers has been studied. Starting from the information of the orientation of the magnetic field, the angular velocity in each direction is derived.

The main drawback of this technology is that there are magnetic disturbances caused by electrical appliances and metallic objects in the environment.34

Barometric pressure sensors. Barometric pressure sensors can be useful in activity recognition, for instance to detect falls due to a sudden pressure increase or to recognize if a subject is going up or down the stairs. In fact, although current activity monitors based on inertial sensors are accurate in recognizing dynamic activities (walking and running), their abilities to classify static postures (standing vs. sitting or sitting vs. lying) remain limited as there is no estimation of the elevation, which on the other hand can be obtained from barometric pressure.35

Video-based recognition systems are a useful tool for activity monitoring. Common in-camera types include RGB video, depth video and RGB-D video; their software is capable of recognizing complex gait activities. However, such strategies restrict the movement of the user within a specific range, depending on how many 2D/3D cameras are employed. Furthermore, outputs are easily influenced by lighting variability and other exogenous factors.

Radio-based recognition systems employ radiofrequencies to detect positions and activities; the most common technologies are RFID, Wi-Fi, ZigBee and infrared light. There are recent studies on other strategies, like the Doppler radar system, which can also be used to detect information on chest displacement.23 In terms of radio-based HAR systems, the base stations must be prearranged and the tags are often attached to a person’s wrist, ankle, head, or other parts. By observing that different activities cause different wireless communication patterns between the attached tags and the base station, activities can be recognized. However, these technologies only work within a specific range of distances from the base stations.33

According to the World Health Organization, air pollution exposure is responsible for 7 million premature deaths every year. Thanks to wearable devices that monitor air quality in real time, environmental data can be transmitted to the Internet to create maps of air quality using a GPS. There are devices that can be clipped to the clothes, for instance TZOA, AirBeam or the Lapka Personal Environment Monitor. A recent trend is to create specific clothes with embedded environmental sensors, such as the WAIR scarf, also equipped with a filter that blocks particulates (PM>0.1), pollen, gases and bacteria.2

Household air pollution is ranked as the 9th largest Global Burden of Disease risk and derives from a range of sources. The most important ones are combustion for heating, tobacco smoke, cooking, building materials, paints and varnishes. Indoor air pollution using biomass fuel in low-income countries has also been found to contribute to the prevalence of COPD.36 Using modern sensor technology, it is possible to measure online not only the climatic parameters but also the concentrations of polluting substances.37 There are several commercially available devices that measure environmental parameters of interest in respiratory research; examples are uHoo38 and Foobot,39 among others.

Several environmental parameters have been found to be particularly significant in the case of respiratory patients. Concentrations of CO2 higher than 800ppm cause a broad range of diseases and symptoms, related to the upper and lower respiratory tract, eyes and skin and include headache, fatigue and difficulty in concentrating. With rising values of CO2 in the atmosphere, symptoms worsen up to unbearable dyspnoea, vomiting, disorientation, hypertension and loss of consciousness for concentrations above 100,000ppm.40 Carbon monoxide (CO) reacts with haemoglobin to form carboxyhaemoglobin: reversible binding occurs on the same heme site where oxygen binds, thus preventing the transport of oxygen to the tissues.41 Among other causes, it has been found that CO concentration increases if there is no ambient ventilation after smoking.42 Chemical and microbiological volatile organic compounds (VOCs) influence the air quality in an indoor space and are primarily generated by humans. Internal sources are derived from human activities such as cooking or smoking and from building materials, as well as metabolic and biochemical processes.43–46 Evidence has been found that nitrogen dioxide (NO2) has a causal role in mortality and in the development of chronic respiratory diseases. Some epidemiological studies suggest associations of long-term NO2 exposures with respiratory and cardiovascular mortality and with children’s respiratory symptoms and lung function.47 Ozone (O3) is a strong oxidant agent with a variety of effects including lung inflammation,48 alveolar epithelial damage and changes in chemical composition of lung lavage fluids, as well as reduction in lung function.49,50 Fine particles (PM2.5) have small diameters but large surface areas and may therefore carry toxic agents, because, through the nose filtration, they reach the end of the respiratory tract with airflow and accumulate there by diffusion, then enter the circulation.51 PM2.5 causes asthma and respiratory inflammation, jeopardizes lung functions and even promotes cancers.52 In scientific literature, excessive cold temperatures have been linked to increases in the mortality and morbidity of COPD patients.48 The interactions between temperature and humidity have also been studied: humidity itself is not associated with patients’ symptoms, but there is an interactive effect between temperature and humidity.53

CO2 and O2 are the gases exchanged between the blood and the inhaled air during respiration, with CO2 diffusing approximately 20 times faster than O2. A high level of PaCO2 (hypercapnia) causes symptoms such as mild headaches, lethargy, and confusion, or severe ones with a hypnotic effect and acidosis with subsequent organ dysfunction, which can lead to coma and death. PaCO2 can be estimated with transcutaneous CO2 measurements (tcpCO2) sensors, which are non-invasive, continuous and cause no sleep disturbance. However, this measurement is nowadays performed by electrochemical sensors requiring a temperature increase, which causes burns in long-term monitoring. Another drawback is the approximate 2-min lag time for the PaCO2 changes to be reflected in the tcpCO2. The problem of burns caused by electrochemical sensors may be overcome by using optical sensors able to determine CO2 by measuring its optical absorption in the evanescent wave of a waveguide integrated in their surface.54 Various devices offer the possibility to simultaneously monitor the level of oxygenation and CO2, combining either SpO2 with tcpCO2 (e.g., SenTec™, Therwil, Switzerland, and TCM Tosca™, Radiometer, Neuilly Plaisance, France) or transcutaneous tcpO2 with tcpCO2 (e.g., TCM Combi MTM, Radiometer, Neuilly Plaisance, France).24

Holter monitors are traditionally used for ECG acquisitions in remote settings, but they are being surpassed by wireless bipotential patch devices. The patch ECG (PECG) sensors are unobtrusive, wire free and can record from weeks to months, which helps detection of long-term malignant arrythmias. Examples are the Savvy monitor, ZIO® XT Patch by iRhythm Technologies, Inc. and SEEQ™ MCT patch device by Medtronic, Inc.24 The previously cited L.I.F.E. device includes a wearable ECG, as it features 12 electrodes located on the internal surface of the garment.10 Even more recently, very short-term (seconds to minutes) handheld smartphone-enabled systems (AliveCor®, ECG Check) have been developed. They require attachment of an electrode-embedded module to a smartphone that detects electrical impulses from the user’s fingertips and transmits signals to the mobile device to generate continuous single-channel ECG for the duration of the contact between the fingers and the sensor.55

The arterial blood pressure can be noninvasively measured by the auscultatory method with an inflatable cuff, a manometer measuring the pressure in the cuff and a stethoscope placed over the brachial artery in the elbow. With the oscillometric method, on the other hand, the cuff is inflated to a point above systolic blood pressure and slowly deflated while oscillations are detected by a pressure sensor. Oscillations begin approximately at systole and continue below diastole with maximal oscillations occurring at mean arterial pressure.56 There are also continuous, noninvasive arterial pressure-monitoring devices that measure in real time without cuffs, but their precision is yet to be proved.24

Normal body function depends on a relatively constant body temperature. Body temperature increases in infections, malignancy and many inflammatory conditions; a value above 38.5° can indicate a severe exacerbation in COPD patients. The most common sites used to obtain body temperature readings with medical thermometers are the anus, the mouth, under the arm and the ear; approaches for continuous monitoring of the temperature have been reported in the literature.24

The main sensor-manager link technologies57 are shown in Fig. 2.

Fig. 2.

Main characteristics of common mHealth sensor-manager link technologies. The wireless range, the data-transfer rate and the battery life are represented.

(0.14MB).

Bluetooth and Bluetooth Low Energy (BTLE). The Bluetooth technology operates at unlicensed ISM (Industrial, Scientific, and Medical) 2.4GHz frequency band, which is currently being used by Home RF, Wi-Fi networks, cordless telephones, baby monitors, walkie-talkies, garage door operators and other applications, so it is subject to interference. Bluetooth transmits both data signals and voice signals over a short distance of up to 100m.58 In contrast to Bluetooth, Bluetooth Low Energy (BTLE) has a longer battery life at the price of reducing the wireless range and the data transfer rate, therefore it is an appropriate substitute for applications in which the amount of data sent is low but there is the need to reduce costs and improve the autonomy of the device.

ANT and ANT+. ANT is a protocol for ultra-low-power, short-range wireless communications in sensor networks. It ensures low power consumption by using low data rate, short delay cycles, and deep-sleep mode and can operate for longer periods of time (more than a year). It operates in the 2.4GHz ISM band and can communicate with multiple nodes over a single 1MHz channel, switching channels if any interference occurs. ANT can be distinguished from other wireless protocols by its unique feature in which it acts as a master for one channel and slave for another channel. Like ZigBee, ANT supports multiple network topologies and also ensures coexistence with neighbouring ANT nodes. The maximum data rate achieved by ANT systems ranges from 20 to 60 kbps and there is a trade-off between data rate and low power consumption. A recent advancement in the ANT protocol, ANT+, uses application-specific device profiles to communicate between two devices, i.e. the set of network rules, parameters, data formats specific to a particular application. Furthermore, the ANT+protocol has the advantage of being interoperable with other ANT+devices having the same device profile.59

Z-Wave. Z-Wave is a wireless communication standard designed for remotely controlled application. Its data rate is 40 kbps at a transmission frequency of 915MHz (900MHz ISM band) and it can reach up to 30m in the open air.60 This technology is particularly optimized for low-overhead commands such as on-off (as in a light switch or an appliance) and raise-lower (as in a thermostat or volume control). As it does not interfer with other devices usually present in home environments, Z-Wave allows for a standardized low bandwidth control medium that can be reliable.61

ZigBee. One of the most widely used wireless technologies in WBANs is ZigBee (IEEE 802.15.4) because it is targeted at applications that require a low data rate and long battery life. In particular, this protocol is advantageous in real-time monitoring because of the low power consumption, large range (10−1500m), limited data rate (250 kbit/s), low cost and flexibility in supporting multiple network topologies.62 The working frequencies of ZigBee are 868MHz (Europe), 915MHz (Americas/Australia) and 2.4GHz (worldwide) and transmits data over one, ten and sixteen channels respectively. In the last band, it is subject to interference with other devices sharing this band.3 While Bluetooth and BLE only support peer-to-peer (P2P) and star topologies, ZigBee devices can be connected using P2P, star, tree and mesh network topologies. Prior to transmitting a packet, the ZigBee protocol assesses the communication link by using CSMA/CA (Carrier Sense Multiple Access with Collision Avoidance) protocol or by sending beacons to other nodes.59

Wi-Fi. Wi-Fi is a family of radio technologies that is commonly used for wireless local area networking (WLAN), which is based on the IEEE 802.11 family of standards, and designed to interwork with its wired sister protocol Ethernet. Devices that can use Wi-Fi technologies include desktops, laptops, smartphones, tablet, smart TVs, printers, digital audio players, digital cameras, cars and drones. Compatible devices can connect to each other via Wi-Fi through a wireless access point as well as to connected Ethernet devices. Such an access point (or hotspot) has a range of about 20m indoors.

Wi-Fi uses the 2.4GHz and 5GHz ISM radio bands, which are subdivided into multiple channels, each time-shared by multiple networks.

Several sensing devices have embedded Wi-Fi modules and therefore can be connected to the Internet directly. This is advantageous when the sensors are in fixed positions, for instance in the case of home environmental sensors.

4G. The 4G (Fourth Generation) technology is a cellular technology which relies on an all-Internet Protocol (IP) based communication, IP telephony.

With the ultra-broadband internet service, 4G networks are projected to provide speeds of 100 Mbps while moving and 1 Gbps for stationary devices.

The mobile 4G network uses several frequencies, depending on the provider and the country, in the ranges of MHz and GHz. Several actions are possible thanks to this technology, such as broadband access in remote locations, seamless connectivity, interoperability, access to music and data and photo sharing.

5G. The 5G (Fifth Generation) technology provides real time success device-to-device communication (D2D). The frequencies covered are in the bands of 700MHz, 3.5GHz and 26GHz and a speed of 10 Gbps can be reached. The main advantages of this technology are that the data are processed close to the final user thanks to the utilization of core Multi-Edge Computing (MEC) networks positioned close to the access network, which grants a low latency. Other differentiating characteristics are the high band capacity and number of connected devices. The 5G technology is based on a Multiple Input Multiple Output (MIMO) technology and multiple access is obtained with Orthogonal Frequency division: these traits are distinctive compared to previous technologies.63 In fact, Massive MIMO utilizes extensive service antennas by spatial multiplexing. Overall, the 5G technology helps in concentrating energy into smaller areas, therefore high throughput and efficiency can be readily achieved on a large scale and with great momentum.64 Nowadays, the 5G technology is mostly used only in wide networks, while local networks usually use sensor-manager link technologies.