[go: up one dir, main page]
More Web Proxy on the site http://driver.im/
Next Article in Journal
3D Reconstruction Based on Iterative Optimization of Moving Least-Squares Function
Next Article in Special Issue
Artificial Intelligence-Based System for Retinal Disease Diagnosis
Previous Article in Journal
EAND-LPRM: Enhanced Attention Network and Decoding for Efficient License Plate Recognition under Complex Conditions
Previous Article in Special Issue
Prediction of Hippocampal Signals in Mice Using a Deep Learning Approach for Neurohybrid Technology Applications
You seem to have javascript disabled. Please note that many of the page functionalities won't work as expected without javascript enabled.
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Algorithms
Volume 17
Issue 6
10.3390/a17060261
algorithms-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles thumb_up
...
Endorse textsms
...
Comment
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Systematic Review

Artificial Intelligence-Based Algorithms and Healthcare Applications of Respiratory Inductance Plethysmography: A Systematic Review

by
Md. Shahidur Rahman
,
Sowrav Chowdhury
,
Mirza Rasheduzzaman
and
A. B. M. S. U. Doulah
*
Department of Electrical and Electronic Engineering, University of Liberal Arts Bangladesh, Dhaka 1209, Bangladesh
*
Author to whom correspondence should be addressed.
Algorithms 2024, 17(6), 261; https://doi.org/10.3390/a17060261
Submission received: 6 May 2024 / Revised: 1 June 2024 / Accepted: 11 June 2024 / Published: 14 June 2024
(This article belongs to the Special Issue Artificial Intelligence-based Algorithms with Potential Applications in Healthcare and Prediction of Disease Evolution)
Browse Figures
Figure 1
<p>Functional block diagram of RIP.</p> "> Figure 2
<p>Flow chart of the screening and selection process for the studies.</p> "> Figure 3
<p>Distribution of number of subjects per study.</p> "> Figure 4
<p>RIP belt coil design with three different sine wave patterns with step sizes of (<b>a.</b>) 1 cm, (<b>b.</b>) 1.5 cm, and (<b>c.</b>) 3 cm [<a href="#B44-algorithms-17-00261" class="html-bibr">44</a>].</p> "> Figure 5
<p>(<b>a</b>) Patient asleep while using a CPAP device. (<b>b</b>) OSA patient with tissue blocking the airway.</p> ">
Review Reports Versions Notes

Abstract

:
Respiratory Inductance Plethysmography (RIP) is a non-invasive method for the measurement of respiratory rates and lung volumes. Accurate detection of respiratory rates and volumes is crucial for the diagnosis and monitoring of prognosis of lung diseases, for which spirometry is classically used in clinical applications. RIP has been studied as an alternative to spirometry and shown promising results. Moreover, RIP data can be analyzed through machine learning (ML)-based approaches for some other purposes, i.e., detection of apneas, work of breathing (WoB) measurement, and recognition of human activity based on breathing patterns. The goal of this study is to provide an in-depth systematic review of the scope of usage of RIP and current RIP device developments, as well as to evaluate the performance, usability, and reliability of ML-based data analysis techniques within its designated scope while adhering to the PRISMA guidelines. This work also identifies research gaps in the field and highlights the potential scope for future work. The IEEE Explore, Springer, PLoS One, Science Direct, and Google Scholar databases were examined, and 40 publications were included in this work through a structured screening and quality assessment procedure. Studies with conclusive experimentation on RIP published between 2012 and 2023 were included, while unvalidated studies were excluded. The findings indicate that RIP is an effective method to a certain extent for testing and monitoring respiratory functions, though its accuracy is lacking in some settings. However, RIP possesses some advantages over spirometry due to its non-invasive nature and functionality for both stationary and ambulatory uses. RIP also demonstrates its capabilities in ML-based applications, such as detection of breathing asynchrony, classification of apnea, identification of sleep stage, and human activity recognition (HAR). It is our conclusion that, though RIP is not yet ready to replace spirometry and other established methods, it can provide crucial insights into subjects’ condition associated to respiratory illnesses. The implementation of artificial intelligence (AI) could play a potential role in improving the overall effectiveness of RIP, as suggested in some of the selected studies.

1. Introduction

Respiratory disorders are prevalent worldwide, with an estimated 454.6 million cases of chronic respiratory illnesses, which have seen a 39.8% increase during the last 30 years [1]. Climate change and increasing air pollution are negatively affecting respiratory health globally, and there is an increasing death toll due to respiratory illnesses [2]. Non-communicable chronic respiratory diseases, i.e., asthma, chronic obstructive pulmonary disease (COPD), cystic fibrosis, and mesothelioma, have caused the deaths of an estimated 4.1 million people worldwide in recent years [3,4], while infectious lung diseases have killed about 2.5 million people in the same time period [5]. Acute respiratory infections often need hospitalization of the patient along with administration of antibiotics and medical care. Chronic respiratory conditions affect the life quality of patients in the long term, as they often require medication and monitoring of prognoses. Respiratory function assessment is crucial for the early identification of potential respiratory disorders, playing a critical role in guiding interventions and management strategies.
Over recent decades, there has been a significant body of research dedicated to enhancing respiratory monitoring systems and refining pulmonary function assessment through advanced sensor technologies. Pulmonary function tests (PFTs) are instrumental in examining key parameters in individuals with respiratory conditions, aiding in treatment during both the diagnostic and recovery phases. Among these tests, spirometry stands out as the conventional method for the measurement of respiratory functioning, which is typically conducted using a pneumotachometer (PnT). This approach requires the individual to use a mouthpiece for the direct measurement of changes in internal ventilation. Another time-intensive PFT method, full-body plethysmography, demands significant patient cooperation [6,7].
Substantial resources have been mobilized in recent years, particularly due to the SARS-CoV-2 pandemic, to enhance medical diagnostics, develop wearable sensors, and create technologies for identifying various pulmonary diseases. The adoption of wearable sensor technology for health monitoring has experienced a significant uptick, especially within the domain of respiratory health. Numerous studies have explored the potential of breath monitoring devices and the application of sensors in evaluating pulmonary functions [8]. These pulmonary function tests (PFTs) are indispensable in assessing the respiratory health status of individuals with confirmed or potential respiratory conditions, providing crucial insights into their health status.
Respiratory Inductive Plethysmography (RIP) is a non-invasive measure for monitoring key respiratory parameters, i.e., respiratory rates and volumes, which has been thoroughly studied as an alternative to spirometry due to its convenience of use and wide range of applications. RIP can be used in both stationary and ambulatory subjects and is generally suitable for continuous long-term monitoring, which offers an advantage over traditional spirometry in some specific cases [9,10,11]. Moreover, analysis of breathing patterns from RIP data can be used in diagnostic procedures in some illnesses associated with respiratory functioning [12,13]. With the advancement in computation and signal processing techniques, RIP has huge potential to be used as an alternative alongside some other established methods in clinical practices. Numerous studies have examined the validity of RIP and implemented this technique in their research methodologies. This review work carefully examines the current literature to synthesize data and provide a conclusive statement on the feasibility of RIP in healthcare applications.

Research Question and Objectives

This study evaluated the scope of use of RIP by conducting a systematic review in accordance with the PRISMA guidelines to answer the following question:
How does the RIP method contribute to the healthcare sector in assessing a person’s respiratory health status through the measurement of respiratory parameters, and how can the application of machine learning improve the outcomes of RIP’s applications?
A comprehensive quality assessment of the studies was conducted to minimize the risk of bias. Furthermore, the following research objectives were set to answer the research question:
  • Conduct a comprehensive systematic review of the existing literature to explore the application and effectiveness of Respiratory Inductive Plethysmography (RIP) for various healthcare purposes, including monitoring respiratory activity and diagnosis.
  • Assess the reliability and medical applicability of RIP as an alternative or supplement to established methods within its designated scope of use, considering factors such as performance and usability.
  • Investigate the influence of different calibration methods on the performance of Respiratory Inductive Plethysmography (RIP) in terms of measurement accuracy and reliability.
  • Explore the evolution and recent technological advancements of Respiratory Inductance Plethysmography (RIP) devices, focusing on wearability and mobility.
  • Evaluate the applications of machine learning (ML) techniques in the processing of RIP data and assess the practicality of ML-based approaches.
  • Identify areas for future research to advance RIP implementation through both conventional techniques and machine learning-based approaches.

2. A Brief Overview of the Related Concepts

2.1. Pulmonary Function Test (PFT)

Spirometry is considered the “gold standard” for the measurement of respiratory volumes as well as diagnosing both obstructive and restrictive lung diseases [9,14,15]. However, the surge in technology has led to the development of a variety of methods for monitoring PFT and other respiratory measures, such as chest wall motion, airflow and respiratory volumes, respiratory rate, resultant transcutaneous oxygen saturation (SpO2), and respiratory carbon dioxide (CO2) elimination. Notably, the authors of [16] developed a wearable patch-type strain sensor that facilitates real-time lung function measurements, such as forced volume capacity (FVC) and forced expiratory volume (FEV), alongside breath monitoring, which is particularly pertinent in the context of the COVID-19 pandemic. The authors of [17] introduced a novel wearable sensor device employing a multi-modal wireless sensor for monitoring pulmonary health status, showcasing a new approach for remote primary diagnostics. A comprehensive wearable sensor system utilizing a multi-sensor strategy for lung function monitoring was presented in [18], demonstrating the potential for non-invasive estimation of tidal lung volume from electromyogram (EMG) measurements. Furthermore, a machine learning-assisted system with wireless wearable sensors for the real-time evaluation of respiratory behaviors was proposed in [19], marking a significant advancement in tracking the progression of respiratory disorders and diseases. Also, the authors of [20] reported on a wearable miniaturized non-dispersive infrared sensor for continuous monitoring of transcutaneous CO2, which could revolutionize the remote assessment of pulmonary gas exchange efficiency. These innovations underscore the rapid advancements in wearable sensor technologies, offering new avenues for respiratory health monitoring and disease management. The equipment and methods for assessing respiration are classified in Table 1.

2.2. Respiratory Inductive Plethysmography (RIP)

In addition to the conventional methods of measuring respiratory parameters, Respiratory Inductive Plethysmography (RIP) is a non-invasive method of measuring lung ventilation. Tidal volume measured with RIP shows a strong correlation with spirometry in healthy subjects [21,22].
It generally involves the movement of the chest during respiration to measure spontaneous tidal breathing, though labored breathing has also been studied with RIP. Two elastic belts are placed around the patient, one at the abdomen around the umbilicus and the other at the ribcage around the nipple line. Generally, RIP can be performed on subjects in around 5–10 min depending on the condition and calibration [23]. RIP devices usually use a transducer with coils connected to an analogue circuit that functions as a variable-frequency LC oscillator. The inductance may be calculated using the frequency and amplitude of the coils due to their inductive nature. The belts’ length changes with each breath, causing the self-inductance of the coils to also vary. Changes in the ribcage’s and the belly’s circumference or cross-sectional area can be identified by observing alterations in the coils’ self-inductance [24]. Figure 1 illustrates the basic operation of RIP.
Although most current technologies show potential, the major advantage of RIP is that it allows for the examination of tidal breathing and breathing pattern changes in patients who are unable to cooperate due to their age or disease severity [25]. Moreover, the RIP system can also be implemented for individuals in an ambulatory environment [26,27,28]. The ease of use of this system, the device, and the method are coming to be of great interest to researchers and clinicians. This work intends to provide a complete account of the usability of RIP in a wide range of healthcare applications and to provide insight into its future research directions.
The usage of RIP is not limited to adults only; its utility in pediatric care is underscored in [29], in which the authors provide age-related ranges of reference values for infants and children, emphasizing its non-invasive nature and utility in monitoring respiratory patterns in younger populations without requiring active cooperation. Technological advancements have led to the development of a low-cost, portable bio-impedance system capable of detecting both respiratory and heart activities, potentially improving patient care outside clinical settings [30]. The feasibility of RIP for long-distance medical consultations through a mobile application is particularly relevant in areas with limited access to medical facilities, offering a method for pre-diagnosing lung conditions remotely [31].
Innovative applications, such as the automated detection of swallowing events using RIP, offer potential in remote monitoring of patients with swallowing disorders, providing a non-invasive method to assess and aid in managing these conditions [32]. Additionally, the authors of [33] introduced a method for detecting respiratory motions of the chest and abdomen without constraints using a flexible tactile sensor array, enhancing the comfort and feasibility of respiratory monitoring, especially in sleep studies. These studies illustrate the continuous evolution of RIP technology and its expanding applications across different fields, from pediatric care to remote health monitoring and sleep studies. The ongoing research and development in this area promise to further integrate RIP into diverse healthcare and research contexts, improving non-invasive respiratory assessment techniques and enhancing patient care.

3. Methods

3.1. Search Strategy and Data Collection Process

This work adhered to the PRISMA guidelines for systematic reviews. The goal of the primary search was focused on locating studies on RIP or the implementation of RIP in research methodologies published from 2012 to 2023. A final search was performed in March 2024 for the identification of relevant studies. Additional searches were performed to identify earlier works on RIP to comprehend the evolution of the RIP technique. The scope of use of RIP was identified first to determine the necessary keywords for the search procedure. Five databases (Springer, IEEE Explore, ScienceDirect, Google Scholar, and PLoS ONE) were then searched using the selected keywords. A total of 395 studies were primarily selected for further evaluation. Table 2 lists the search keywords along with the numbers of search results. It is noteworthy that the abbreviation “RIP” in Table 2 was used for “Respiratory Inductive Plethysmography” along with quotation marks during the literature search procedure. Moreover, compound keywords, e.g., “Respiratory Inductive Plethysmography” and “machine learning applications”, were combined with the AND operator in corresponding searches to identify relevant studies. The included studies were thoroughly investigated by two reviewers working independently for the extraction of necessary data. Study designs, experimental protocols, and statistical analyses of the results were carefully examined during the data extraction procedure.

3.2. Inclusion and Exclusion Criteria

Forty studies were selected from the original selection based on specific inclusion and exclusion criteria outlined in Table 3. Two reviewers working individually screened the primarily selected studies following the inclusion and exclusion criteria. Figure 2 illustrates the entire screening method.

3.3. Characteristics of the Selected Studies

Table 3 presents the detailed search procedure, and Figure 2 depicts a PRISMA flow chart for the complete screening process for the selection of studies for complete review. From the 2955 studies identified in the databases, 623 duplicates in the search results were removed.
The studies in the primary selection were checked for eligibility by inspecting their titles and abstracts. Along with the 40 studies included in this review, further cross-checking was conducted for better comprehension and analysis following the primary selection. A synthesis of results was performed, considering the test system design and setup; the number of participants along with their ages, gender, and health status; the gold standard for the test procedure; and the data analysis method used. Two reviewers working independently collected data from each study. Based on the characteristics of the study, an overview of the selected studies and the results of these studies with relevant data figures were synthesized in Section 4 and Section 5.

3.3.1. System Design

Among the studies included in this review, 12 studies implemented commercially available complete systems incorporating RIP along with other biometric sensors. Refs. [9,34,35,36] analyzed data obtained using the Hexoskin Smart Shirt system, refs. [37,38,39] analyzed data obtained using the VivoMetrics LifeShirt system, refs. [40,41] used the Respitrace system, and refs. [10,23,42] implemented the pneuRIP™ system in their experiments. The remaining 24 studies implemented commercially available generic RIP belts, though these were used along with some other biometric sensors in several cases. All of the studies used the dual RIP belt setup (chest and abdomen), except for [27,43], which used a single thoracic belt. All the included studies implemented RIP either exclusively or along with other sensor(s) monitoring vital information of the subject. Moreover, all of the included studies relied on RIP data for their final results.

3.3.2. Scope of Use

The studies were conducted for various applications, including measurement of respiratory volumes [9,27,34,35,37,44,45,46,47,48,49,50], monitoring of respiratory activity in infants [11,40,50,51], sleep studies in adults [12,13,52,53,54,55,56], and tracking of selected human activities [36] and smoking [57,58]. The studies [37,38,48] exclusively focused on the calibration of RIP devices, and the studies [11,23,55] analyzed the performance of RIP exclusively in children. The study [59] utilized RIP to detect stiffening of the aorta by measuring pulse wave velocity (PWV). In [39,43], experiments were performed to determine the validity of RIP for high-BMI subjects. The studies [41,42,60,61,62,63] solely focused on machine learning techniques for automated detection or prediction of non-ideal conditions. Studies on all applications of RIP identified in our search were included in this work having met the inclusion and exclusion criteria and passed the quality assessment.

3.3.3. Number of Subjects

The number of subjects varied in the included studies due to differences in their system designs, measurement protocols, target age ranges, and scope. Figure 3 is a frequency histogram of the number of subjects in the studies assessed in this work.

3.4. Quality Assessment of Included Studies

A qualitative assessment of the studies included in this review was conducted based on the MMAT 2018 criteria for quantitative descriptive studies by two reviewers working independently [64]. The initial screening questions (S1 and S2) were as follows:
S1.
Does a research question (RQ) exist?
S2.
Do the data collected for analysis address that RQ?
Studies that did not satisfy the criteria in S1 and S2 did not qualify for inclusion. For all the studies (n = 40) included in this work, “Yes” was the answer to questions S1 and S2, and further assessment was performed according to the following questionnaire:
Q1.
Does the sampling methodology relevantly address the research question?
Q2.
Does the sample properly represent the target population?
Q3.
Have the measurements been taken appropriately?
Q4.
Does the risk of non-response bias remain at a low level?
Q5.
Does the statistical analysis appropriately answer the RQ?
The answers to these questions were classified into 3 categories, “Yes”, “No”, and “Not specified”. A further breakdown of the assessment for Q1–Q5 is elaborated as follows:
  • The sampling methodology had to be relevant to the research question, i.e., if the subjects’ status was healthy, that must align with the research question (Q1).
  • Subjects’ health status must be clarified within the paper, i.e., non-smoking and no history of respiratory illnesses in the case of healthy subjects. Subjects with illnesses must not have suffered from any other conditions that may have produced faulty results (Q2).
  • The authors must describe a complete and comprehensive experimentation method (Q3).
  • The authors ensured that the dropped-out subjects and faulty data were excluded from the results (Q4).
  • If the results did not properly answer the research question, the paper was excluded (Q5).
The quality assessment of the selected studies is presented in Table 4.

4. Results

4.1. Evolution of RIP

RIP’s design and performance have been carefully tested throughout the years. In a 1983 study [28], the authors studied the validation of RIP measurements by comparing RIP signals with spirometry results, where 100% and 96% of the validation values remained within a ±10% error range for ambulatory and ventilator-supported subjects, respectively. Further development of RIP and study of its effectiveness was observed in the literature; in [68], the authors found a very high correlation coefficient between RIP and spirometer readings (r = 0.982 and r = 0.978 for volume and airflow, respectively), with <8% error for tidal volume. In [69], the authors studied and verified RIP in healthy and simulated obstructed breathing, achieving a regression coefficient of 0.95 ± 0.056 (mean ± SD), which demonstrates RIP’s potential to identify respiratory changes in obstructive lung diseases. In [7], the authors studied RIP in children of 3–5 years of age to observe various respiratory parameters to assess the feasibility of RIP in such a challenging age group. The scope of use of RIP has broadened over time, and numerous studies have tested and validated RIP in a range of practical case scenarios. RIP was studied to detect post-operative apnea [70,71] in infants and achieved a promising <2% error figure in measuring tidal volume [71]. RIP was implemented by integrating RIP belts in wearable garments in a 2007 study to assess the effect of adherence of RIP belts to the human body [72]. RIP has demonstrated its ability to be used for long-term monitoring purposes, i.e., in patients with neuromuscular disease [73].

4.1.1. Technological Advancements

Due to the dependency of the RIP technique on wearable belts, the design and specifications of the wire woven or sewn along the belt appears to affect the sensor properties of the RIP. The work in [44] investigated the influence of the step size of the sine wave pattern on the characteristics of RIP sensors by building three RIP sensors with the same wire pattern (sine wave), height (2.2 cm), and length (30 cm) but variable step sizes, as shown in Figure 4. The inductance of each of these belts was measured using an Agilent 4285A Precision LCR Meter. It was observed by the authors that the smaller the step size, the greater the sensitivity. There was a tiny variance in linearity between the 1 cm and 1.5 cm step sizes. However, the linearity significantly reduced with a step size of 3 cm. The authors’ conclusion directs that it is thus prudent to keep the step size modest in order to maintain linearity.
In [10], the researchers conducted benchtop testing of PneuRIPTM to determine the circuitry and instrument design, reliability, responsiveness, and repeatability. The RIP bands were mounted on a laboratory-grade rocker/waver platform to mimic different breathing situations, and the tilt angles and speed were changed. Over two days, the testing ran without taking any pauses. The benchtop testing was performed along with testing of the Respitrace system (Sensormedics), an RIP system from 1991 [71]. With typical breathing frequencies and phase angles, both the PneuRIPTM and Respitrace systems produced similar results, but the phase angle variance was substantially reduced using PneuRIPTM for low and high phase angles, as well as low and high breathing frequencies. PneuRIPTM is more precise and reliable at greater and lower phase angles because the mechanical fluctuation is less. Children with respiratory distress frequently have large phase angles, which PneuRIPTM is intended to detect. According to one study [44], the step size should be reduced to attain adequate linearity.
Over time, the RIP device has evolved depending on the applications. One important aspect of the RIP device is that it is a piece of oscillator-based hardware that requires calibration over some time. Additionally, it also requires several inductive sensors to obtain mutual inductance between RIP sensors. RIP calibration was studied in both healthy subjects and COPD patients in [74], and it was found that calibrating the RIP device in each posture produced less error in tidal volume measurements. Research investigations [45,69] suggest adjusting the frequencies of the thoracic and abdominal band oscillators to minimize modulation and prevent frequency locking. Creating a dependable oscillator for each inductive sensor in a respiratory monitoring system while maintaining frequency separation can be challenging. Energy consumption is crucial for battery-powered portable devices; hence, a compact printed circuit board size is necessary for mobility. Thus, using several inductive sensors in a typical RIP architecture with an LC oscillator might lead to an increase in chip count and complexity in power management. To mitigate this problem, the study [68] devised an RIP system that uses traditional Electrical Impedance Plethysmography (EIP) to quantify the coils’ self-inductance. Since a steady current travels through the belts, this approach does not require an LC oscillator. Sinton’s method incorporates a continuous AC signal which resolves the frequency locking but requires high power, which makes the system non-portable. Despite the use of a variable-frequency oscillator, the authors of [45] used a Pulse Amplitude Modulation (PAM) approach for indirect inductance measurement [68,69,72]. During the pulse excitation phase, the RIP sensor is periodically and regularly delivered a high current from a constant-current source, and its inductance is measured. The feeding current does not consistently flow through the inductive sensor, unlike the method used by Sinton. The PAM technique considerably reduces power consumption while maintaining sensitivity due to the huge feeding current. Contemporary developments have completely diminished the use of an LC oscillatory circuit by using a high-performance, low-power embedded microprocessor that instantly digitizes the inductance and therefore requires no calibration and allows the storage of data locally and simultaneously, transmitting it wirelessly by computer and cloud using PneuRIP™ [10]. As a result, it is recommended that users account for advancements in RIP devices prior to applying them for the intended function.

4.1.2. Wearability and Mobility

Continuous monitoring of respiration is crucial for detecting and anticipating high-risk situations, and appropriate monitoring techniques may prove vital for the life-saving management of conditions. The implementation of RIP in a wearable form allows PFT and breathing pattern (BP) monitoring of the user in a completely unrestrained environment, allowing the measurement of data while the individual carries on with their daily activities. The mobility of RIP devices is also advantageous in ambulatory situations, where PFT and BP are critical measurements. Having such a system embedded in an ambulance might help diagnose patients and provide critical reports for medics. A remote monitoring system to analyze and determine breathing patterns was proposed in [31] for consultation purposes. The system can identify the differences between regular breathing, somewhat irregular breathing due to smoking, and irregular breathing caused by cough and cold.
Researchers have also contributed to investigating the best implementation of RIP sensors. The authors of [27] scrutinized three different attachment options for RIP devices: (i) an RIP belt secured using tension only; (ii) an RIP belt held in place by adhesive and tension; (iii) an RIP belt sewn into a T-shirt. From the findings over 24 h, the method with the RIP belt stitched inside a T-shirt demonstrated the highest performance. This approach has been proven to be more resistant to abnormalities induced by body movement. The RIP device has only to be removed during showering, and participants in the trial were instructed to go about their everyday routines. The calibration coefficients of the system based merely on the belt or the belt + adhesive became unreliable over time. This happened as a result of their being unreliable and inconvenient when the participants were performing day-to-day activities. It was observed that the lowest measurement error occurred with the belt + shirt combination (RMS error <0.1 L), while the belt-only and the belt + adhesive configurations produced noticeable errors in all three subjects (RMS errors ranging from 0.039 L to 0.495 L).
Carre Technologies in Canada offers wearable shirts with integrated sensors designed to monitor vital signs, such as blood pressure. The Hexoskin Smart Shirt (HX) was designed with activity sensors specifically for use in sports and medical treatment environments [9,34,35,36]. This sort of wearable shirt can also be highly beneficial in other disciplines of research, such as cigarette smoking monitoring [57,58]. Some other commercially available wearable RIP systems are available, and a few selected systems have been addressed and analyzed in the results synthesis in this work.

4.2. Overview of RIP Applications and Implemented Methodologies

The application of RIP in different contexts needs different system design protocols customized for the intended purposes. Table 5 details the synthetic overview of the selected studies for general applications of RIP.

4.3. General Applications of RIP

Applications of RIP are entirely dependent on the analysis of the signals derived from thoracic and abdominal belts due to the expansion and contraction of the thoracic cage and abdomen to sustain a normal respiratory pattern. Interruption in respiration can thus be detected with RIP as anomalous movement of the ribcage and abdomen, which is clearly visible in the signals from the belts. Both automated and manual statistical analyses of RIP data are performed to measure respiratory rates and volumes, the phase angle between the thorax and abdomen, as well as other relevant information, e.g., breathing pauses, changes in breathing patterns, etc. The analyzed data reveal information on the aforementioned respiratory parameters along with probable reasons behind any abnormal breathing pattern based on experimental conditions.

4.3.1. Sleep Studies and Detection of Apnea

Obstructive sleep apnea (OSA) is a medical condition interlinked with obesity and hypertension that has become more prevalent in recent decades in both rich and developing nations [75]. OSA causes decreased quality of life and mortality in patients in some severe cases [76]. Apneas are defined by temporary cessation of respiration during sleep for ≥10 s, with ≥5 such events per hour, which corresponds to a reduction of ≥4% saturation in oxygenated hemoglobin. However, respiration does not completely cease in hypopnea, which is characterized by a reduction in respiratory airflow by 30–50% [77]. The severity of OSA is classified with the apnea–hypopnea index (AHI), with scores ≥5 for mild, ≥15 for moderate, and ≥30 for severe forms [53]. The American Academy of Sleep Medicine (AASM) recommends using an oronasal thermistor and a nasal cannula for AHI scoring in sleep studies [78], but these sensors might be uncomfortable for some people and can even interrupt their sleep. Patients can move them about or even remove them during sleep studies, making the sleep study data incorrect and unusable. Continuous positive airway pressure (CPAP), a prevalent treatment for obstructive sleep apnea, is depicted in Figure 5a, while Figure 5b depicts the airway obstruction caused by apnea. The authors of [12] proposed that the combination of tracheal sounds and RIP could be used as a reliable method for detecting and characterizing respiratory events without the need to place sensors on patients’ faces and that it could be used as an alternative to or in addition to the recommended AASM sensors. RIP can monitor breathing asynchrony between the thorax and abdomen occurring in OSA and can aid in diagnosis of OSA, though asynchronous breathing is not exclusive to OSA [79]. Detection of sleep stages has also been studied with RIP, along with apnea detection, in [56], where a strong correlation was found between RIP and PSG (r = 0.91) for AHI estimation, and accuracy ranges between 57.1% and 75% were determined for the detection of sleep stages as well as wakefulness.

4.3.2. Post-Operative Apnea in Infants

Post-operative apnea (POA) is the term for the condition specified as apnea which is caused by anesthesia in infants during the post-operative recovery phase [70,76,80]. Preterm infants are most at risk of apnea, as 25% of surgical cases in infants with a post-birth age of <44 weeks suffer from POA [81]. Both central and obstructive apnea have been reported, along with mixed events [76,81]. Because of the blockage in the pharyngeal airway, apneas create respiratory effort with diminished to ceased inspiratory airflow. Both central and obstructive apnea are potentially dangerous, as death may occur in severe cases. The majority of neonates require medical intervention to restart normal breathing [82]. In [11], the authors developed an Automated Unsupervised Respiratory Event Analysis (AUREA) to identify breathing patterns recorded using twin-belt RIP. The findings of this study suggest that AUREA can efficiently detect the respiratory rate in RIP signals on a sample-by-sample basis. The AUREA pattern classifications have been utilized to analyze the breathing patterns of premature newborns following the removal of endotracheal tubes (ETTs). A predictor of extubation readiness in extremely premature infants is planned to be developed using both AUREA patterns and AUREA measurements. Evaluating the patterns and measures produced by AUREA can help identify respiratory patterns and indicate vulnerable children at risk of POA.

4.3.3. Neuromuscular Functioning Associated with Respiration

Neuromuscular illnesses encompass a wide range of conditions that impact muscle function, either by affecting the voluntary muscles directly or by impacting the peripheral nervous system or neuromuscular junctions indirectly. Respiratory failure in neuromuscular illnesses might manifest as an abrupt occurrence or in a chronic, progressive manner [8]. Monitoring of respiratory activity in neuromuscular diseases can be expensive, time-consuming, and upsetting for patients. Better evaluation methods, such as in-home assessments, are being investigated. Assessment with PneuRIPTM might be a more efficient way to assess a patient’s lung health. In [10], an outpatient clinic environment was set up and the efficacy of PneuRIPTM in adolescents with neuromuscular illnesses was evaluated and their RIP parameters were compared with those of age-matched healthy individuals. The PneuRIPTM technique has been used to successfully evaluate work of breathing (WOB) indices and respiratory patterns and has proven to be effective in recognizing abnormalities in WOB indices and breathing patterns in patients with neuromuscular illnesses. RIP can detect swallowing in both healthy and neurologically affected subjects, which is crucial for identification of choking and subsequent respiratory arrest [32,83]. Difficulty in swallowing associated with neuromuscular pathology has been shown to be detectable with RIP with 95% sensitivity and 99% specificity [83]. Increased intercoastal muscle activity in patients with spinal cord injury at cervical level has been successfully monitored with RIP in comparison with normal subjects, which activity has been undetectable with electromyography (EMG) [84].

4.3.4. Other Restrictive and Obstructive Lung Diseases

The authors of [85] studied the efficiency of RIP in detecting pulmonary airway obstructions and found RIP to be useful for monitoring bronchoconstriction. Bronchoconstriction is characterized by contraction of the smooth muscles surrounding the air exchange pathway in the lungs and subsequent reduced gaseous exchange between the blood and the air. In medical practice, body plethysmography can deduce both restrictive and obstructive lung diseases, such as asbestosis, pulmonary fibrosis, sarcoidosis, scoliosis, asthma, bronchiectasis, and emphysema. Along with these, some other cardiovascular parameters can also be monitored via PFTs using RIP. Table 6 presents the results synthesis for the selected studies for general applications of RIP.

5. Artificial Intelligence-Based Applications of RIP

The integration of machine learning (ML) and artificial intelligence (AI) with RIP marks a significant advancement in the field of respiratory medicine, where feature extraction from RIP data can reveal hidden information to identify a variety of both physiological and pathological conditions. The utilization of advanced technologies allows for a more detailed and thorough examination of breathing patterns. An important development in this field involves the study and application of non-linear computational models to enhance the calculation of respiratory volume based on thoracoabdominal motions with the use of RIP. The shift towards employing machine learning (ML) models, such as artificial neural networks (ANNs) and adaptive neuro-fuzzy inference systems (ANFISs), is a divergence from the linear calibration approaches formerly used. These models have shown an impressive ability to greatly improve the accuracy and dependability of respiratory volume estimations. In comparative research [86], advanced non-linear approaches were compared to standard linear models, demonstrating the improved efficacy of machine learning algorithms in accurately representing the intricate dynamics present in respiratory patterns. This study highlighted the capacity of machine learning (ML) to offer a more dependable and accurate assessment of respiratory function, which is essential for the efficient monitoring and control of respiratory well-being.
Moreover, the extensive applicability of machine learning in respiratory medicine is demonstrated by the creation of sophisticated diagnostic instruments for respiratory illnesses. In [87], the authors highlighted the crucial significance of artificial intelligence (AI) in examining lung images for the purpose of diagnosing disorders like asthma and chronic obstructive pulmonary disease (COPD). This potential advancement in automating and precisely analyzing pulmonary function tests (PFTs) and RIP data signifies a big stride forward. Integrating RIP and machine learning (ML) could improve the diagnostic process and enable personalized treatment plans, leading to optimized patient care and better outcomes. In [38], the authors presented a revolutionary approach for calibrating Respiratory Inductive Plethysmography (RIP) using machine learning (ML) and artificial intelligence (AI) in the field of respiratory medicine without requiring ground truth, i.e., spirometry. This method allows for the integration of anthropometric data without the requirement for collecting RIP and flowmeter data simultaneously. However, this technique is still in the experimental phase and requires further improvement.
In [88], the authors studied the correlation between data from a smart mask and a specially designed RIP sensor, in which the RIP coils are knitted into a wearable garment. The resultant graphs showed good agreement between predicted CO2 levels and CO2 data from the smart mask. Such integration of RIP sensors into wearable garments is anticipated to improve convenience and user comfort to a greater degree. In [89,90], the authors studied machine learning-based sleep stage classification. Data packet loss in RIP signals was taken into account in [89], and 76% accuracy with a Cohen’s kappa of 0.54 was achieved when sleep stages were classified into three categories. On the other hand, ref. [90] utilized ECG and body movement data alongside abdominal RIP signals and achieved 84.3% accuracy and a 0.8 F1 score. Table 7 details the synthetic overview, whereas Table 8 presents the results synthesis of the selected studies for machine learning based applications of RIP.

6. Discussion

This systematic review identified the most crucial implementations of RIP studied between 2012 and 2023. Thirty-four studies were selected through searching relevant databases for articles published within the selected timeline. Implementations of machine learning (ML)-based approaches used to analyze RIP data were extensively analyzed in the included studies for a comprehensive evaluation of the capabilities and potential applications of AI within the scope of the uses of RIP. RIP serves as a non-invasive tool that can be used as an alternative to spirometry to aid diagnosis and monitoring of respiratory activities, i.e., tidal volume, respiratory rate, minute ventilation, and phase angle of breathing between thorax and abdomen. However, usage of RIP is not restricted to the conventional scope of spirometry and extends to the monitoring of some cardiovascular activities as well. Through carefully pre-processing raw RIP data, statistical analyses and ML-based approaches have been used to extract relevant information in studies. All of the included studies validated their results and outcomes by implementing parallel methods, often gold standards within the corresponding fields. Several applications of RIP have been identified in these studies, e.g., the validation of RIP in subjects in both stationary and ambulatory cases, monitoring of infants’ breathing in non-ideal conditions, diagnosis of sleep apnea, classification of sleep stages, calibration procedures for RIP, and recognition of human activity through conventional statistical analysis and the application of machine learning where applicable.
Studies [9,34,35] validated the Hexoskin smart shirt through the measurement of correlations with spirometry and percentages of measurement error, which demonstrated its potential for use in practical conditions. Study [10] compared pneuRIPTM with the Respitrace system, and the percentage error demonstrated the superiority of the pneuRIPTM system in the experiment. Studies [23,51,65] expressed the magnitude of breathing asynchrony in phase-angle degrees, where a 180° phase angle indicates complete asynchrony between thoracic and abdominal breathing. Studies [12,53] expressed the detection of apnea and hypopnea in terms of specificity and sensitivity figures, which indicate false-positive and false-negative results, respectively. Studies [44,45] expressed their results in percentage error ranges, and the measurements showed an error range between 5% and 20% in most cases. In [27], the errors were expressed in root-mean-square (RMS) values, which were much lower for the belt + shirt combination than for the other test conditions. Studies [43,46,47,48,50] expressed their results in terms of correlation coefficients between pneumotachometer and RIP measurements, which ranged between 0.8 and 0.95, which indicated a relatively high accuracy. Study [37] adopted a support vector regression (SVR) model to compare RIP calibration methods, where a least-squares approximation (LSQ) model proved to be the most accurate. Study [13] expressed AHI scoring errors measured with RIP in terms of scoring biases with manual scoring, which showed some errors, whereas [54] expressed AHI scoring accuracy through correlation coefficients, which indicated zero error in their experiment. Study [59] demonstrated a high correlation of RIP-measured pulse wave velocities (PWVs) with brachial PWV measurements, where p < 0.001, which indicates the potential feasibility of RIP in this setting. Study [55] demonstrated a moderate correlation coefficient between gold-standard and RIP data and emphasized the importance of analysis of low-frequency data for sleep stage classification. Study [40] measured tidal volumes, breathing asynchrony, and percentage ribcage contributions for breathing to identify the best positioning for premature infants, where high tidal volume, low breathing asynchrony, and a higher ribcage contribution indicate a better position. Study [38] demonstrated that data collected from standing still before exercise provide a better calibration, which is indicated by a higher limit of equivalence with a flowmeter. Study [67] monitored breathing asynchrony associated with muscle fatigue, and a higher degree of asynchrony was found in cases of muscle fatigue due to repeated work. In [39], the authors found the limits of agreement of RIP with spirometry to be too high (≥20%) to validate RIP in obese people. The studies [36,41,42,52,56,57,58,60,61,62,63,91] implemented machine learning and expressed their results in terms of accuracy, precision, sensitivity, specificity, and/or Cohen’s kappa figures. The resultant figures were self-explanatory in terms of statistical significance. All of these studies demonstrate the potential for the application of machine learning trained on RIP data based on their accuracy figures.
Respiratory rates and volumes detected with RIP show generally good agreement with spirometry results for both stationary and ambulatory scenarios, and light to moderate exercise does not seem to affect the usability of RIP, except in [35,49]. However, upon subject-wise inspection of RIP measurements, errors were greater for a few subjects in this regard, deviating from averaged aggregated results. For example, ≥93% and ≥99% of results remained within the ±10% and ±20% error limits, respectively, in [45]. The ~6% results lying within the ±10%–±20% error range is a concern for practical usability, which may result in unreliable outcomes in clinical practices. Moreover, results lying within the ±10% error range might be unusable for some clinical procedures, e.g., diagnosis of obstructive lung diseases, as FEV1, FVC, respiratory rate, and tidal volume measurements are required to be precise and accurate in such cases. Other selected studies [9,35,46,47,48,50,66] showed strong correlations between RIP and spirometry in this regard (0.80≥ r ≥1.00), except for [49]; however, some degree of measurement error existed in all of these studies. On the other hand, [34] achieved a flow-rate detection accuracy >99% using a custom algorithm developed by the authors. There is no conclusive evidence of an acceptable error range for measurement of VC, FEV1, FVC, and FEV1/FVC ratios in any of the selected studies for diagnostic purposes. Therefore, we do not recommend RIP as an alternative to spirometry for clinical and diagnostic purposes.
The variation in RIP measurement errors might occur due to the data processing techniques used and the inclusion or exclusion of selected data based on signal quality. Moreover, none of the selected studies specified the gender-wise performance of RIP; thus, the gender-based variability in RIP measurements remains unclear. Additionally, the RIP data collection and extraction procedure for the measurement of respiratory volumes may not be more suitable than spirometry in practical conditions due to the necessity of strict adherence to the study protocols. The study [27] was conducted on three subjects of 30 ± 2 years of age, which limits the acceptability of the results for a broader range of subject age groups. The studies [35,44] only measured respiratory rates and did not measure respiratory volumes. Subjects were found to be healthy in all of the studies except for [46], while [44,45] did not mention subject health status. RIP measurements were validated in high-BMI subjects (BMI ≥30) against a control group in [43], whereas the authors of [39] found low or insignificant agreement between RIP and spirometry when a control group and patients with obesity hypoventilation syndrome were compared. It is our opinion that further investigation is necessary to validate RIP measurements across the whole BMI range in both healthy and disease conditions.
Upon analyzing the results synthesis for the selected studies, RIP appears to perform better when the belts are sewn or embedded into wearable garments. The reasons lie in the better adherence of RIP belts to the surface of the body, as well as retained placement of the belts throughout the experimental procedure, as suggested by the evidence. Custom-sized belts in wearable garments should provide better subject-wise calibration, which may reduce errors to a great extent, as suggested in [27,46]. Thus, we prefer wearable RIP systems to the conventional dual RIP belt setup if the conditions are suitable and accuracy is crucial, as there might be an increased cost associated with such implementations of RIP.
RIP has been found to be effective in detecting sleep apnea, as apnea–hypopnea index (AHI) figures measured from RIP data have an acceptable agreement with the AHI figures derived from polysomnography. The breathing asynchrony produced by partial or complete cessation of breathing produces a relatively “flat” signal in RIPsum channels, which represents the summation of signals from thoracic and abdominal RIP belts. RIP appeared to be less sensitive for detection of apnea and hypopnea (AHI scoring) when compared to manual scoring in [12,13,54], which did not mention the BMIs of the subjects. However, RIP provided more sensitivity and specificity than sleep study results for AHI scoring in study [53], which was conducted exclusively on obese people. The reason for such differences in results may lie in the experimental and signal processing methodologies used, as well as subject BMIs, sample sizes and demographics, and criteria of manual scoring. Detection of sleep stages from RIP signals has also been studied in the literature, and results have been found to be moderately to strongly correlated with those of polysomnography. However, sleep stage classification with RIP is still in the experimental phase and requires further study. Frequency-based analysis focusing on lower frequencies (<1 Hz) might offer more insights in this regard, as suggested in [52,55]. Based on the analysis of the results from the selected studies, we recommend RIP for AHI scoring in sleep studies, with RIP devices being embedded in wearable garments in cases where patient comfort and convenience are a priority.
RIP is well-tolerated in infants for monitoring respiratory activities to identify post-operative apnea (POA) and respiratory distress, due its non-invasive nature. Phase shifting between thoracic and abdominal RIP channels is observed in respiratory distress in preterm infants, which indicates increased respiratory effort due to breathing insufficiency, which can be detected with RIP. Studies have been carried out on the validation of RIP measurements [48]; detection of POA [11]; observation of respiratory effort to adjust respiratory support [49], i.e., high-flow nasal cannula (HFNC) and continuous positive airway pressure (CPAP); and examination of infant positioning to reduce breathing effort [40] in infants. Though [50] demonstrated a strong agreement of RIP with esophageal manometry, signals with artifacts were excluded from the analysis. The classification of breathing in [11] takes movement of the infant into account but requires pre-processing of collected data prior to analysis, which negatively affects the feasibility of implementing such a system for real-time monitoring. In [51], 78% of the collected data were included in the final assessment, while the rest of the data were discarded due to poor signal quality. The necessity of including selected data indicates that the implementation of RIP could be challenging for continuous monitoring of breathing parameters due to movement and positioning of the infant. Removing movement-induced artifacts from raw data could resolve this issue to a certain extent, which might be addressed in future research.
RIP has been studied in children to measure work of breathing (WoB) in normal subjects [10] and patients with neuromuscular illnesses [23] with the PneuRIPTM system. Though RIP measurements were validated in [10], the study only experimented on healthy subjects and simulated breathing insufficiency by adding a resistive load to the respiratory pathway. This method does not fulfil the criterion of representing patients with obstructive lung diseases in our eyes, as the vital capacity remains unchanged in a healthy subject even if additional resistance is added to the respiratory pathway. However, integration with handheld smart devices is an advantage of the PneuRIPTM system in the latter study, which provides conventionality with respect to the target user group. On the other hand, increased WoB parameters were observed in patients with neuromuscular diseases, though significant overlaps were present between the control and the test subjects’ phase angles and labored breathing indices.
Breathing asynchrony in adults has been studied with RIP in the assessment of muscle fatigue due to repetitive work [67] and COPD [65] and in children with respiratory diseases [23], and significant increases in phase angles were found in all of these cases. In the case of COPD patients, breathing asynchrony became evident during a 6 min walk test (6MWT), which did not appear during the resting phase prior to the 6MWT. As the baseline respiratory parameters remained the same in both the healthy subjects and the COPD patients, the differences between the workloads required to produce breathing asynchrony appear to be indicators of subjects’ lung conditions. Increased phase angles were also found in children with respiratory diseases, which are also associated with increased WoB. Asynchronous breathing associated with increased WoB was observed in preterm infants with respiratory distress, for which RIP proved to be a great monitoring tool, being both convenient and non-invasive. However, calibration remains challenging in such uses of RIP due to the movement of subjects and change in positioning.
It appears to us that improper calibration might be the root cause of measurement errors in RIP implementations. Though [37] found that the least-squares approximation (LSQ) method performed the best, it still produced more errors than the gold standard. Calibration of RIP devices is challenging due to the difficulties associated with determining the gain parameter of RIP signals, which usually requires a ground truth as a reference [48]. This might be the reason behind the reference-free calibration technique studied in [38] having the poorest performance. Other studies included in this work suggest that calibration is required for each subject separately prior to measurement. Moreover, change in the subject’s posture and positioning may require the device to be recalibrated. The development of new calibration techniques is another area of exploration for future research regarding RIP. Faulty results are also caused by improper belt positioning and poor adherence of the belt to the body, as the systems implementing RIP in wearable garments performed the best in our results synthesis. One study [46] suggested that re-using RIP belts tends to cause a higher magnitude of measurement error, which might be due to the distortion in the inductive material embedded in the RIP belts after repeated use. Attention from researchers is required in this regard to prove the validity of such claims. Estimation of pulse wave velocity has been experimented on in one study [59]. The high correlation (r = 0.86, p < 0.001) between measurements from RIP data and ECG is promising and requires further study to establish conclusive evidence.
The application of machine learning algorithms for automated analysis of RIP data has been studied in the literature and shown great potential for some unique applications. For example, changes in breathing patterns were studied by analyzing RIP data through classifier models to identify human activities. Studies [57,58] have recruited regular smokers as subjects to identify smoking activity from RIP data. Though the ≥80% accuracy figure demonstrates great potential for practical uses, research could be conducted in the future to improve the overall precision and F1 scores. The study [36] achieved >95% accuracy in recognizing human activity; however, the experimental protocol was limited to 10 tasks. No alternative to artificial intelligence (AI) for human activity recognition (HAR) was identified in our investigation. The accuracy figures suggest that training with a larger dataset could improve the accuracy and precision of HAR in the near future. In one study [52], a bagging classifier achieved superior accuracy in sleep stage detection when a heuristic approach was applied (80.38 ± 8.32% vs. 77.85 ± 6.63), while the Cohen’s kappa score was improved simultaneously (0.65 ± 0.13 vs. 0.59 ± 0.11). Another ML-based approach achieved a sensitivity and a specificity of 0.70 and 0.66, respectively, for the same purpose [56]. It is noteworthy that these ML-based applications have been found in our research to be unique to RIP and cannot be implemented with traditional statistical analysis. A binary K-means classifier in [11] achieved an 80% overall accuracy with <26% overall category-wise confusion for the detection of neonatal apnea and breathing asynchrony.
Power output during exercise was studied in [60] with various predictive models, which showed moderate accuracy (R2 = 0.56, mean absolute percentage error: 0.20–0.24) and need improvement. However, the fact that the data were obtained from a single subject is the greatest limitation of this study. We suggest repeating the same procedure with multiple subjects, including both male and female subjects. Moreover, the inclusion of subjects with respiratory diseases along with healthy subjects would further validate the results. There was a dependency on heart rate in this study to achieve the results, which could be further studied solely using RIP data. Apnea detection through machine learning showed promising results in [61], which were comparable to results obtained with other techniques in [12,54,55,56]. However, the results from abdominal RIP data fell behind the results from nasal pressure data (accuracy: 90.3% vs. 78.5%, respectively). It is our recommendation to experiment with RIP data from both channels and the summation of both channels (RIPsum) for better results. Real-time detection of drowsiness in drivers through machine learning was studied in [62], which produced a high specificity and sensitivity (>96% and >90%, respectively). The implementation of such an approach could be implemented in real life along with vision-based drowsiness detection. Determination of the accuracy of a system implementing both methods simultaneously could be a topic for future research. However, the limitation of such implementations is the requirement of wearing an RIP belt while driving, which may be inconvenient and result in less compliance by drivers.
Detection of breathing asynchrony in children was studied in [42], which benefits from adding inverse cumulative percentage (ICP) features. The training dataset included data from healthy subjects who breathed through a resistive load to simulate abnormal breathing, which could have been replaced with actual data from patients with neuromuscular illnesses to represent actual scenarios. Moreover, this study could have been performed with a larger population size. Detection of COPD stages has been studied with machine learning in [63], which utilized instantaneous phase differences as a feature. However, the results suffered from some inaccuracies, which need to be improved for practical implementation. A larger dataset might have contributed to less error in the results. Extubation success prediction was performed in [41] in extreme preterm infants with promising results. As extubation fails in 25% of real-life cases, the >83% accuracy for the failure class results demonstrates the potential of such a system. However, the system produced some false-positive results in predicting extubation failure, so there is scope for improvement for real-life applications.
It is our observation that the capability of AI to reveal hidden information from RIP data could be stretched further to replace traditional statistical analysis methods and revolutionize the RIP technique in the future. Moreover, problems related to signal processing, i.e., removal of noise and artifacts, filtering, and discarding of faulty signals, could be solved with the use of AI, as observed in the studies in this review. Though the machine learning techniques illustrated in the studies demonstrate strong potential in their corresponding applications, the error rates must be acceptable with respect to clinical standards for their practical usability. However, analysis of RIP data through artificial intelligence methods might uncover previously unknown correlations between breathing patterns and other activities within the human body as well as human activity. Data pre-processing techniques might play a key role in the success of such applications of AI.
In our analysis, RIP has demonstrated the potential to be used in multiple diagnostic and healthcare settings where the degree of measurement error produced by RIP is acceptable. Though RIP might not be the non-invasive method to replace the gold standards in the corresponding fields yet, it has proved its ability to be used as a powerful tool to gain insights into subjects’ health statuses that are relevant to physicians. It is our recommendation to apply ML-based algorithms in parallel with traditional data analysis methods to reveal a clear picture of the capabilities of AI in this regard. As the non-invasive nature of RIP and the convenience of its use are the greatest strengths of RIP when compared to the traditional but gold-standard methods, addressing the issues of measurement errors along with the application of AI may establish RIP as an acceptable method for clinical use and diagnostic procedures.

7. Conclusions and Future Work

This systematic review work focused on identifying the scope of RIP, its usability, performance, and reliability, by examining data from existing research. Experimental data from 40 studies were extracted and analyzed in a thorough investigation to achieve this purpose. RIP can be used as a non-invasive alternative to spirometry and enables the measurement of respiratory parameters of ambulatory subjects while sacrificing some accuracy. However, its usability beyond the lab environment enables the collection and interpretation of data while subjects engage in normal daily activities. It can be concluded that RIP is an acceptable method in some specific applications, while the technique requires further improvement to be used as a diagnostic tool. RIP has been studied for monitoring of respiratory volumes, detection of apnea in infants and adults, recognition of human activity through machine learning, and it can identify any respiratory event associated with respiratory insufficiency. RIP has the potential to be used for remote monitoring and diagnosis, as well as the detection and classification of specific human activities based on respiratory patterns through the application of machine learning.
To our knowledge, this is the only paper that provides a systematically reviewed summary of the usage of RIP devices in the medical field. Collecting data from a large population of both healthy subjects and patients with illnesses affecting respiratory function could provide further insight into the performance and reliability of the RIP technique. There is scope for future research based on our observations, including but not limited to the development of RIP devices with better sensitivity and data processing techniques, self-calibrating methods that do not require ground truth, new methods for the integration of RIP devices into wearable garments that allow subjects to perform normal daily activities, portable RIP systems capable of storing respiratory function data and remote monitoring, and machine learning algorithms embedded in RIP systems capable of providing real-time diagnoses.

Author Contributions

M.S.R.: Investigation, methodology, formal analysis, writing—original draft. S.C.: Investigation, writing—original draft, formal analysis, visualization. M.R.: Conceptualization, methodology, resources, writing—review and editing, supervision. A.B.M.S.U.D.: Conceptualization, methodology, writing—review and editing, supervision. All authors have read and agreed to the published version of the manuscript.

Funding

This research did not receive any external funding.

Data Availability Statement

This work is a systematic review; thus, no data were generated throughout the study.

Conflicts of Interest

The authors declare no conflicts of interest.

References

  1. Momtazmanesh, S.; Moghaddam, S.S.; Ghamari, S.-H.; Rad, E.M.; Rezaei, N.; Shobeiri, P.; Aali, A.; Abbasi-Kangevari, M.; Abbasi-Kangevari, Z.; Abdelmasseh, M.; et al. Global burden of chronic respiratory diseases and risk factors, 1990–2019: An update from the Global Burden of Disease Study 2019. EClinicalMedicine 2023, 59, 101936. [Google Scholar] [CrossRef]
  2. Tran, H.M.; Tsai, F.-J.; Lee, Y.-L.; Chang, J.-H.; Chang, L.-T.; Chang, T.-Y.; Chung, K.F.; Kuo, H.-P.; Lee, K.-Y.; Chuang, K.-J.; et al. The impact of air pollution on respiratory diseases in an era of climate change: A review of the current evidence. Sci. Total Environ. 2023, 898, 166340. [Google Scholar] [CrossRef]
  3. World Health Organization World Health Statistics 2023: Monitoring Health for the SDGs, Sustainable Development Goals 2023. Available online: https://cdn.who.int/media/docs/default-source/gho-documents/world-health-statistic-reports/2023/world-health-statistics-2023_20230519_.pdf (accessed on 30 December 2023).
  4. Rose, S.; Paul, C.; Boyes, A.; Kelly, B.; Roach, D. Stigma-related experiences in non-communicable respiratory diseases: A systematic review. Chronic Respir. Dis. 2017, 14, 199–216. [Google Scholar] [CrossRef]
  5. icddr,b. “Pneumonia and Other Respiratory Diseases”. Available online: https://www.icddrb.org/news-and-events/press-corner/media-resources/pneumonia-and-other-respiratory-diseases (accessed on 30 March 2024).
  6. Rosenfeld, M.; Allen, J.; Arets, B.H.G.M.; Aurora, P.; Beydon, N.; Calogero, C.; Castile, R.G.; Davis, S.D.; Fuchs, S.; Gappa, M.; et al. An Official American Thoracic Society Workshop Report: Optimal Lung Function Tests for Monitoring Cystic Fibrosis, Bronchopulmonary Dysplasia, and Recurrent Wheezing in Children Less Than 6 Years of Age. Ann. Am. Thorac. Soc. 2013, 10, S1–S11. [Google Scholar] [CrossRef]
  7. Mayer, O.H.; Clayton, R.G.; Jawad, A.F.; McDonough, J.M.; Allen, J.L. Respiratory Inductance Plethysmography in Healthy 3- to 5-Year-Old Children *. Chest 2003, 124, 1812–1819. [Google Scholar] [CrossRef]
  8. Yuki, K.; Fujiogi, M.; Koutsogiannaki, S. COVID-19 pathophysiology: A review. Clin. Immunol. 2020, 215, 108427. [Google Scholar] [CrossRef]
  9. Villar, R.; Beltrame, T.; Hughson, R.L. Validation of the Hexoskin wearable vest during lying, sitting, standing, and walking activities. Appl. Physiol. Nutr. Metab. 2015, 40, 1019–1024. [Google Scholar] [CrossRef]
  10. Rahman, T.; Page, R.; Page, C.; Bonnefoy, J.-R.; Cox, T.; Shaffer, T.H. pneuRIPTM: A Novel Respiratory Inductance Plethysmography Monitor. J. Med. Devices 2017, 11, 011010. [Google Scholar] [CrossRef]
  11. Robles-Rubio, C.A.; Kearney, R.E.; Bertolizio, G.; Brown, K.A. Automatic unsupervised respiratory analysis of infant respiratory inductance plethysmography signals. PLoS ONE 2020, 15, e0238402. [Google Scholar] [CrossRef]
  12. Sabil, A.; Marien, C.; LeVaillant, M.; Baffet, G.; Meslier, N.; Gagnadoux, F. Diagnosis of sleep apnea without sensors on the patient’s face. J. Clin. Sleep Med. 2020, 16, 1161–1169. [Google Scholar] [CrossRef]
  13. Magalang, U.J.; Arnardottir, E.S.; Chen, N.-H.; Cistulli, P.A.; Gíslason, T.; Lim, D.; Penzel, T.; Schwab, R.; Tufik, S.; Pack, A.I.; et al. Agreement in the Scoring of Respiratory Events Among International Sleep Centers for Home Sleep Testing. J. Clin. Sleep Med. 2016, 12, 71–77. [Google Scholar] [CrossRef]
  14. Johns, D.P.; Walters, J.A.E.; Walters, E.H. Diagnosis and early detection of COPD using spirometry. J. Thorac. Dis. 2014, 6, 1557–1569. [Google Scholar] [CrossRef]
  15. Mehrparvar, A.H.; Sakhvidi, M.J.Z.; Mostaghaci, M.; Davari, M.H.; Hashemi, S.H.; Zare, Z. Spirometry values for detecting a restrictive pattern in occupational health settings. Tanaffos 2014, 13, 27–34. [Google Scholar]
  16. Kumar, A.; Rakesh Kumar, R.K.; Shaikh, M.O.; Lu, C.-H.; Yang, J.-Y.; Chang, H.-L.; Chuang, C.-H. Ultrasensitive Strain Sensor Utilizing a AgF–AgNW Hybrid Nanocomposite for Breath Monitoring and Pulmonary Function Analysis. ACS Appl. Mater. Interfaces 2022, 14, 55402–55413. [Google Scholar] [CrossRef]
  17. Moon, K.S.; Lee, S.Q. A Wearable Multimodal Wireless Sensing System for Respiratory Monitoring and Analysis. Sensors 2023, 23, 6790. [Google Scholar] [CrossRef]
  18. George, U.Z.; Moon, K.S.; Lee, S.Q. Extraction and Analysis of Respiratory Motion Using a Comprehensive Wearable Health Monitoring System. Sensors 2021, 21, 1393. [Google Scholar] [CrossRef]
  19. Chen, A.; Zhang, J.; Zhao, L.; Rhoades, R.D.; Kim, D.-Y.; Wu, N.; Liang, J.; Chae, J. Machine-learning enabled wireless wearable sensors to study individuality of respiratory behaviors. Biosens. Bioelectron. 2021, 173, 112799. [Google Scholar] [CrossRef]
  20. Tipparaju, V.V.; Mora, S.J.; Yu, J.; Tsow, F.; Xian, X. Wearable Transcutaneous CO₂ Monitor Based on Miniaturized Nondispersive Infrared Sensor. IEEE Sens. J. 2021, 21, 17327–17334. [Google Scholar] [CrossRef]
  21. Solà-Soler, J.; Pérez, D.R.; Balchin, L.; Serra, A.M.; Torné, M.L.; Koborzan, M.R.P.; Giraldo, B.F.G. Respiratory Pattern Analysis for Different Breathing Types and Recording Sensors in Healthy Subjects. In Proceedings of the 2023 45th Annual International Conference of the IEEE Engineering in Medicine & Biology Society (EMBC), Sydney, Australia, 24–27 July 2023; pp. 1–4. [Google Scholar]
  22. Hurtado, D.E.; Chavez, J.A.P.; Mansilla, R.; Lopez, R.; Abusleme, A. Respiratory Volume Monitoring: A Machine-Learning Approach to the Non-Invasive Prediction of Tidal Volume and Minute Ventilation. IEEE Access 2020, 8, 227936–227944. [Google Scholar] [CrossRef]
  23. Strang, A.; Ryan, L.; Rahman, T.; Balasubramanian, S.; Hossain, J.; Heinle, R.; Shaffer, T.H. Measures of respiratory inductance plethysmography (RIP) in children with neuromuscular disease. Pediatr. Pulmonol. 2018, 53, 1260–1268. [Google Scholar] [CrossRef]
  24. Martinot-Lagarde, P.; Sartene, R.; Mathieu, M.; Durand, G. What does inductance plethysmography really measure? J. Appl. Physiol. 1988, 64, 1749–1756. [Google Scholar] [CrossRef]
  25. Beydon, N.; Davis, S.D.; Lombardi, E.; Allen, J.L.; Arets, H.G.M.; Aurora, P.; Bisgaard, H.; Davis, G.M.; Ducharme, F.M.; Eigen, H.; et al. An Official American Thoracic Society/European Respiratory Society Statement: Pulmonary Function Testing in Preschool Children. Am. J. Respir. Crit. Care Med. 2007, 175, 1304–1345. [Google Scholar] [CrossRef]
  26. Hutchison, A.A.; Leclerc, F.; Nève, V.; Pillow, J.J.; Robinson, P.D. The Respiratory System. In Pediatric and Neonatal Mechanical Ventilation; Rimensberger, P.C., Ed.; Springer: Berlin/Heidelberg, Germany, 2015; pp. 55–112. ISBN 978-3-642-01218-1. [Google Scholar]
  27. Ramos-Garcia, R.I.; Imtiaz, M.H.; Sazonov, E.; Tiffany, S.T. Evaluation of RIP sensor calibration stability for daily estimation of lung volume. In Proceedings of the 2017 Eleventh International Conference on Sensing Technology (ICST), Sydney, Australia, 4–6 December 2017; pp. 1–5. [Google Scholar]
  28. Tobin, M.J.; Jenouri, G.; Lind, B.; Watson, H.; Schneider, A.; Sackner, M.A. Validation of Respiratory Inductive Plethysmography in Patients with Pulmonary Disease. Chest 1983, 83, 615–620. [Google Scholar] [CrossRef]
  29. Balasubramaniam, S.L.; Wang, Y.; Ryan, L.; Hossain, J.; Rahman, T.; Shaffer, T.H. Age-related ranges of respiratory inductance plethysmography (RIP) reference values for infants and children. Paediatr. Respir. Rev. 2019, 29, 60–67. [Google Scholar] [CrossRef]
  30. Piuzzi, E.; Pisa, S.; Pittella, E.; Podesta, L.; Sangiovanni, S. Low-Cost and Portable Impedance Plethysmography System for the Simultaneous Detection of Respiratory and Heart Activities. IEEE Sens. J. 2019, 19, 2735–2746. [Google Scholar] [CrossRef]
  31. Torres, J.L.; Yumang, A.; Hortinela, C.; Plameras, J.M.F.; Si, J.E.Z.; Uywan, J.A.O. Respiratory Inductance Plethysmography for Long-Distance Medical Consultation Using Mobile Application for Android Smartphones. In Proceedings of the 2019 IEEE International Conference on Consumer Electronics—Asia (ICCE-Asia), Bangkok, Thailand, 12–14 June 2019; pp. 187–189. [Google Scholar]
  32. Tatulli, E.; Fontecave-Jallon, J.; Calabrese, P.; Gumery, P.-Y. Respiratory Inductance Plethysmography for Automated Swallowing Detection. Int. J. E-Health Med. Commun. 2020, 11, 64–77. [Google Scholar] [CrossRef]
  33. Guo, S.; Zhao, X.; Matsuo, K.; Liu, J.; Mukai, T. Unconstrained Detection of the Respiratory Motions of Chest and Abdomen in Different Lying Positions Using a Flexible Tactile Sensor Array. IEEE Sens. J. 2019, 19, 10067–10076. [Google Scholar] [CrossRef]
  34. Harbour, E.; Lasshofer, M.; Genitrini, M.; Schwameder, H. Enhanced Breathing Pattern Detection during Running Using Wearable Sensors. Sensors 2021, 21, 5606. [Google Scholar] [CrossRef]
  35. Phillips, M.B.; Beach, J.; Cathey, M.; Lockert, J.; Satterfield, W. Reliability and Validity of the Hexoskin Wearable Body Metrics Telemetry Shirt. J. Sport Hum. Perform. 2017, 5. Available online: https://jhp-ojs-tamucc.tdl.org/jhp/index.php/JHP/article/view/108 (accessed on 30 December 2023).
  36. Chaari, M.; Abid, M.; Ouakrim, Y.; Lahami, M.; Mezghani, N. A Mobile Application for Physical Activity Recognition using Acceleration Data from Wearable Sensors for Cardiac Rehabilitation. In Proceedings of the 13th International Joint Conference on Biomedical Engineering Systems and Technologies, Valletta, Malta, 24–26 February 2020; SCITEPRESS—Science and Technology Publications: Valletta, Malta, 2020; pp. 625–632. [Google Scholar]
  37. Leutheuser, H.; Heyde, C.; Gollhofer, A.; Eskofier, B.M. Comparison of a priori calibration models for respiratory inductance plethysmography during running. In Proceedings of the 2014 36th Annual International Conference of the IEEE Engineering in Medicine and Biology Society, Chicago, IL, USA, 26–30 August 2014; pp. 6393–6396. [Google Scholar]
  38. Leutheuser, H.; Heyde, C.; Roecker, K.; Gollhofer, A.; Eskofier, B.M. Reference-Free Adjustment of Respiratory Inductance Plethysmography for Measurements during Physical Exercise. IEEE Trans. Biomed. Eng. 2017, 64, 2836–2846. [Google Scholar] [CrossRef]
  39. Hollier, C.A.; Harmer, A.R.; Maxwell, L.J.; Menadue, C.; Willson, G.N.; Black, D.A.; Piper, A.J. Validation of respiratory inductive plethysmography (LifeShirt) in obesity hypoventilation syndrome. Respir. Physiol. Neurobiol. 2014, 194, 15–22. [Google Scholar] [CrossRef] [PubMed]
  40. Gouna, G.; Rakza, T.; Kuissi, E.; Pennaforte, T.; Mur, S.; Storme, L. Positioning Effects on Lung Function and Breathing Pattern in Premature Newborns. J. Pediatr. 2013, 162, 1133–1137.e1. [Google Scholar] [CrossRef]
  41. Precup, D.; Robles-Rubio, C.A.; Brown, K.A.; Kanbar, L.; Kaczmarek, J.; Chawla, S.; Sant’Anna, G.M.; Kearney, R.E. Prediction of extubation readiness in extreme preterm infants based on measures of cardiorespiratory variability. In Proceedings of the 2012 Annual International Conference of the IEEE Engineering in Medicine and Biology Society, San Diego, CA, USA, 28 August –1 September 2012; pp. 5630–5633. [Google Scholar]
  42. Ratnagiri, M.V.; Ryan, L.; Strang, A.; Heinle, R.; Rahman, T.; Shaffer, T.H. Machine learning for automatic identification of thoracoabdominal asynchrony in children. Pediatr. Res. 2021, 89, 1232–1238. [Google Scholar] [CrossRef] [PubMed]
  43. Rétory, Y.; David, P.; de Picciotto, C.; Niedzialkowski, P.; Bonay, M.; Petitjean, M. Validity of thoracic respiratory inductive plethysmography in high body mass index subjects. Respir. Physiol. Neurobiol. 2017, 242, 52–58. [Google Scholar] [CrossRef] [PubMed]
  44. Ngo, H.T.; Nguyen, C.V.; Nguyen, T.M.H.; Van Vo, T. A Portable Respiratory Monitor Using Respiratory Inductive Plethysmography. In 4th International Conference on Biomedical Engineering in Vietnam; Toi, V.V., Toan, N.B., Dang Khoa, T.Q., Lien Phuong, T.H., Eds.; IFMBE Proceedings; Springer: Berlin/Heidelberg, Germany, 2013; Volume 49, pp. 222–225. ISBN 978-3-642-32182-5. [Google Scholar]
  45. Zhang, Z.; Zheng, J.; Wu, H.; Wang, W.; Wang, B.; Liu, H. Development of a Respiratory Inductive Plethysmography Module Supporting Multiple Sensors for Wearable Systems. Sensors 2012, 12, 13167–13184. [Google Scholar] [CrossRef] [PubMed]
  46. Montazeri, K.; Jonsson, S.A.; Agustsson, J.S.; Serwatko, M.; Gislason, T.; Arnardottir, E.S. The design of RIP belts impacts the reliability and quality of the measured respiratory signals. Sleep Breath. 2021, 25, 1535–1541. [Google Scholar] [CrossRef]
  47. Retory, Y.; Niedzialkowski, P.; de Picciotto, C.; Bonay, M.; Petitjean, M. New Respiratory Inductive Plethysmography (RIP) Method for Evaluating Ventilatory Adaptation during Mild Physical Activities. PLoS ONE 2016, 11, e0151983. [Google Scholar] [CrossRef]
  48. Lo, W.L.A.; Huang, D.F. Assessment of an alternative calibration technique to record breathing pattern and its variability with respiratory inductive plethysmography. J. Clin. Monit. Comput. 2017, 31, 755–764. [Google Scholar] [CrossRef]
  49. Cabiddu, R.; Pantoni, C.B.F.; Mendes, R.G.; Trimer, R.; Catai, A.M.; Borghi-Silva, A. Inductive plethysmography potential as a surrogate for ventilatory measurements during rest and moderate physical exercise. Braz. J. Phys. Ther. 2016, 20, 184–188. [Google Scholar] [CrossRef]
  50. Sivieri, E.M.; Wolfson, M.R.; Abbasi, S. Pulmonary mechanics measurements by respiratory inductive plethysmography and esophageal manometry: Methodology for infants on non-invasive respiratory support. J. Neonatal-Perinat. Med. 2019, 12, 149–159. [Google Scholar] [CrossRef]
  51. Kovatis, K.Z.; Locke, R.G.; Mackley, A.B.; Subedi, K.; Shaffer, T.H. Adjustment of high flow nasal cannula rates using real-time work of breathing indices in premature infants with respiratory insufficiency. J. Perinatol. 2021, 41, 1711–1717. [Google Scholar] [CrossRef]
  52. Tataraidze, A.; Anishchenko, L.; Korostovtseva, L.; Kooij, B.J.; Bochkarev, M.; Sviryaev, Y. Sleep stage classification based on respiratory signal. In Proceedings of the 2015 37th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), Milan, Italy, 25–29 August 2015; pp. 358–361. [Google Scholar]
  53. Kogan, D.; Jain, A.; Kimbro, S.; Gutierrez, G.; Jain, V. Respiratory Inductance Plethysmography Improved Diagnostic Sensitivity and Specificity of Obstructive Sleep Apnea. Respir. Care 2016, 61, 1033–1037. [Google Scholar] [CrossRef] [PubMed]
  54. Park, D.-Y.; Kim, T.; Lee, J.J.; Ha, J.H.; Kim, H.J. Validity analysis of respiratory events of polysomnography using a plethysmography chest and abdominal belt. Sleep Breath. 2020, 24, 127–134. [Google Scholar] [CrossRef]
  55. Immanuel, S.A.; Pamula, Y.; Kohler, M.; Saint, D.A.; Baumert, M. Characterizing respiratory waveform regularity and associated thoraco-abdominal asynchrony during sleep using respiratory inductive plethysmography. In Proceedings of the 2013 IEEE Eighth International Conference on Intelligent Sensors, Sensor Networks and Information Processing, Melbourne, Australia, 2–5 April 2013; pp. 329–332. [Google Scholar]
  56. Dietz-Terjung, S.; Martin, A.R.; Finnsson, E.; Ágústsson, J.S.; Helgason, S.; Helgadóttir, H.; Welsner, M.; Taube, C.; Weinreich, G.; Schöbel, C. Proof of principle study: Diagnostic accuracy of a novel algorithm for the estimation of sleep stages and disease severity in patients with sleep-disordered breathing based on actigraphy and respiratory inductance plethysmography. Sleep Breath. 2021, 25, 1945–1952. [Google Scholar] [CrossRef]
  57. Lopez-Meyer, P.; Tiffany, S.; Patil, Y.; Sazonov, E. Monitoring of Cigarette Smoking Using Wearable Sensors and Support Vector Machines. IEEE Trans. Biomed. Eng. 2013, 60, 1867–1872. [Google Scholar] [CrossRef]
  58. Senyurek, V.Y.; Imtiaz, M.H.; Belsare, P.; Tiffany, S.; Sazonov, E. A Comparison of SVM and CNN-LSTM Based Approach for Detecting Smoke Inhalations from Respiratory signal. In Proceedings of the 2019 41st Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), Berlin, Germany, 23–27 July 2019; pp. 3262–3265. [Google Scholar]
  59. Fontecave-Jallon, J.; Tanguy, S. Inductive Plethysmography for aortic Pulse Wave Velocity. In Proceedings of the 2020 42nd Annual International Conference of the IEEE Engineering in Medicine & Biology Society (EMBC), Montreal, QC, Canada, 20–24 July 2020; pp. 2711–2714. [Google Scholar] [CrossRef]
  60. Husom, E.J.B.L.G.; Bernabé, P.; Sen, S. Deep learning to predict power output from respiratory inductive plethysmography data. Appl. AI Lett. 2022, 3, e65. [Google Scholar] [CrossRef]
  61. ElMoaqet, H.; Eid, M.; Glos, M.; Ryalat, M.; Penzel, T. Deep Recurrent Neural Networks for Automatic Detection of Sleep Apnea from Single Channel Respiration Signals. Sensors 2020, 20, 5037. [Google Scholar] [CrossRef] [PubMed]
  62. Guede-Fernandez, F.; Fernandez-Chimeno, M.; Ramos-Castro, J.; Garcia-Gonzalez, M.A. Driver Drowsiness Detection Based on Respiratory Signal Analysis. IEEE Access 2019, 7, 81826–81838. [Google Scholar] [CrossRef]
  63. Huang, P.-H.; Luo, Y.-F.; Yeh, C.-F.; Hsiao, T.-C. Towards instantaneous phase difference on the COPD pre-discrimination. In Proceedings of the 2018 IEEE 23rd International Conference on Digital Signal Processing (DSP), Shanghai, China, 19–21 November 2018; pp. 1–5. [Google Scholar]
  64. Hong, Q.N.; Pluye, P.; Fàbregues, S.; Bartlett, G.; Boardman, F.; Cargo, M.; Dagenais, P.; Gagnon, M.; Griffiths, F.; Nicolau, B.; et al. Mixed Methods Appraisal Tool (MMAT) Version 2018 User Guide. National Collaborating Centre for Methods and Tools. August 2018. Available online: https://www.nccmt.ca/knowledge-repositories/search/232 (accessed on 30 December 2023).
  65. Chien, J.-Y.; Ruan, S.-Y.; Huang, Y.-C.T.; Yu, C.-J.; Yang, P.-C. Asynchronous Thoraco-abdominal Motion Contributes to Decreased 6-Minute Walk Test in Patients With COPD. Respir. Care 2013, 58, 320–326. [Google Scholar] [CrossRef]
  66. Mandel, J.E.; Atkins, J.H. Hilbert-Huang Transform Yields Improved Minute Volume Estimates from Respiratory Inductance Plethysmography During Transitions to Paradoxical Breathing. Anesth. Analg. 2016, 122, 126–131. [Google Scholar] [CrossRef]
  67. Silva, L.; Dias, M.; Folgado, D.; Nunes, M.; Namburi, P.; Anthony, B.; Carvalho, D.; Carvalho, M.; Edelman, E.; Gamboa, H. Respiratory Inductance Plethysmography to Assess Fatigability during Repetitive Work. Sensors 2022, 22, 4247. [Google Scholar] [CrossRef] [PubMed]
  68. Sinton, A.M.; Suntheralingam, R. Respiratory inductance plethysmography with an electrical impedance plethysmograph. Med. Biol. Eng. Comput. 1988, 26, 213–217. [Google Scholar] [CrossRef] [PubMed]
  69. Cohen, K.P.; Panescu, D.; Booske, J.H.; Webster, J.G.; Tompkins, W.J. Design of an inductive plethysmograph for ventilation measurement. Physiol. Meas. 1994, 15, 217–229. [Google Scholar] [CrossRef] [PubMed]
  70. Brown, K.A.; Aoude, A.A.; Galiana, H.L.; Kearney, R.E. Automated respiratory inductive plethysmography to evaluate breathing in infants at risk for postoperative apnea. Can. J. Anesth. Can. Anesth. 2008, 55, 739–747. [Google Scholar] [CrossRef] [PubMed]
  71. Bar-Yishay, E.; Putilov, A.; Einav, S. Automated, real-time calibration of the respiratory inductance plethysmograph and its application in newborn infants. Physiol. Meas. 2003, 24, 149–163. [Google Scholar] [CrossRef]
  72. Zheng, J.W.; Zhang, Z.B.; Wu, T.H.; Zhang, Y. A wearable mobihealth care system supporting real-time diagnosis and alarm. Med. Biol. Eng. Comput. 2007, 45, 877–885. [Google Scholar] [CrossRef] [PubMed]
  73. Landon, C. Novel Methods of Ambulatory Physiologic Monitoring in Patients with Neuromuscular Disease. Pediatrics 2009, 123, S250–S252. [Google Scholar] [CrossRef] [PubMed]
  74. Brüllmann, G.; Fritsch, K.; Thurnheer, R.; Bloch, K.E. Respiratory Monitoring by Inductive Plethysmography in Unrestrained Subjects Using Position Sensor-Adjusted Calibration. Respiration 2010, 79, 112–120. [Google Scholar] [CrossRef]
  75. Wolk, R.; Shamsuzzaman, A.S.M.; Somers, V.K. Obesity, Sleep Apnea, and Hypertension. Hypertension 2003, 42, 1067–1074. [Google Scholar] [CrossRef]
  76. Sale, S.M. Neonatal apnoea. Best Pract. Res. Clin. Anaesthesiol. 2010, 24, 323–336. [Google Scholar] [CrossRef]
  77. Strollo, P.J.; Rogers, R.M. Obstructive Sleep Apnea. N. Engl. J. Med. 1996, 334, 99–104. [Google Scholar] [CrossRef] [PubMed]
  78. Mansukhani, M.P.; Kolla, B.P.; Ramar, K. International Classification of Sleep Disorders 2 and American Academy of Sleep Medicine Practice Parameters for Central Sleep Apnea. Sleep Med. Clin. 2014, 9, 1–11. [Google Scholar] [CrossRef]
  79. Adha, M.S.; Igasaki, T. Three-Stage Breathing Effort Quantification for Obstructive Sleep Apnea Detection Based on Thoracic and Abdominal Movement Signals. IEEE Access 2021, 9, 72781–72792. [Google Scholar] [CrossRef]
  80. Kurth, C.D.; LeBard, S.E. Association of Postoperative Apnea, Airway Obstruction, and Hypoxemia in Former Premature Infants. Anesthesiology 1991, 75, 22–26. [Google Scholar] [CrossRef] [PubMed]
  81. Malviya, S.; Swartx, J.; Lerman, J. Are All Preterm Infants Younger than 60 Weeks Postconceptual Age at Risk for Postanesthetic Apnea? Anesthesiology 1993, 78, 1076–1081. [Google Scholar] [CrossRef] [PubMed]
  82. Davidson, A.J.; Morton, N.S.; Arnup, S.J.; de Graaff, J.C.; Disma, N.; Withington, D.E.; Frawley, G.; Hunt, R.W.; Hardy, P.; Khotcholava, M.; et al. Apnea after Awake Regional and General Anesthesia in Infants. Anesthesiology 2015, 123, 38–54. [Google Scholar] [CrossRef]
  83. Theytaz, F.; Vuistiner, A.; Schweizer, V.; Crépin, A.; Sandu, K.; Chaouch, A.; Piquilloud, L.; Lecciso, G.; Coombes, K.; Diserens, K. Feasibility study of the Nox-T3 device to detect swallowing and respiration pattern in neurologically impaired patients in the acute phase. Sci. Rep. 2023, 13, 7325. [Google Scholar] [CrossRef] [PubMed]
  84. Yoshida, R.; Tomita, K.; Kawamura, K.; Setaka, Y.; Ishii, N.; Monma, M.; Mutsuzaki, H.; Mizukami, M.; Ohse, H.; Imura, S. Investigation of inspiratory intercostal muscle activity in patients with spinal cord injury: A pilot study using electromyography, ultrasonography, and respiratory inductance plethysmography. J. Phys. Ther. Sci. 2021, 33, 153–157. [Google Scholar] [CrossRef]
  85. Laouani, A.; Rouatbi, S. Thorax and Abdomen Motion Analysis in Patients with Obstructive Diseases. J. Pulm. Respir. Med. 2016, 06. Available online: https://www.hilarispublisher.com/abstract/thorax-and-abdomen-motion-analysis-in-patients-with-obstructive-diseases-36088.html (accessed on 30 December 2023). [CrossRef]
  86. Raoufy, M.R.; Hajizadeh, S.; Gharibzadeh, S.; Mani, A.R.; Eftekhari, P.; Masjedi, M.R. Nonlinear model for estimating respiratory volume based on thoracoabdominal breathing movements. Respirology 2013, 18, 108–116. [Google Scholar] [CrossRef]
  87. Kaplan, A.; Cao, H.; FitzGerald, J.M.; Iannotti, N.; Yang, E.; Kocks, J.W.H.; Kostikas, K.; Price, D.; Reddel, H.K.; Tsiligianni, I.; et al. Artificial Intelligence/Machine Learning in Respiratory Medicine and Potential Role in Asthma and COPD Diagnosis. J. Allergy Clin. Immunol. Pract. 2021, 9, 2255–2261. [Google Scholar] [CrossRef] [PubMed]
  88. Setiawan, M.; Malik, M.A.; Fobelets, K. Correlation between Smart Mask and Knitted Coil Sensors Breathing Data. In Proceedings of the E-Textiles 2023; MDPI: Basel, Switzerland, 2024; p. 15. [Google Scholar] [CrossRef]
  89. Chen, H.-Y.; Wu, H.-T.; Chen, C.-Y. Quality Aware Sleep Stage Classification over RIP Signals with Persistence Diagrams. In Proceedings of the 2023 IEEE 19th International Conference on Body Sensor Networks (BSN), Boston, MA, USA, 9–11 October 2023; pp. 1–4. [Google Scholar] [CrossRef]
  90. Luo, Y.; Li, J.; He, K.; Cheuk, W. A Hierarchical Attention-Based Method for Sleep Staging Using Movement and Cardiopulmonary Signals. IEEE J. Biomed. Health Inform. 2023, 27, 1354–1363. [Google Scholar] [CrossRef] [PubMed]
  91. Robles-Rubio, C.A.; Brown, K.A.; Kearney, R.E. Automated unsupervised respiratory event analysis. In Proceedings of the 2011 Annual International Conference of the IEEE Engineering in Medicine and Biology Society, Boston, MA, USA, 30 August–3 September 2011; pp. 3201–3204. [Google Scholar] [CrossRef]
Algorithms 17 00261 g001
Figure 1. Functional block diagram of RIP.
Figure 1. Functional block diagram of RIP.
Algorithms 17 00261 g001
Algorithms 17 00261 g002
Figure 2. Flow chart of the screening and selection process for the studies.
Figure 2. Flow chart of the screening and selection process for the studies.
Algorithms 17 00261 g002
Algorithms 17 00261 g003
Figure 3. Distribution of number of subjects per study.
Figure 3. Distribution of number of subjects per study.
Algorithms 17 00261 g003
Algorithms 17 00261 g004
Figure 4. RIP belt coil design with three different sine wave patterns with step sizes of (a.) 1 cm, (b.) 1.5 cm, and (c.) 3 cm [44].
Figure 4. RIP belt coil design with three different sine wave patterns with step sizes of (a.) 1 cm, (b.) 1.5 cm, and (c.) 3 cm [44].
Algorithms 17 00261 g004
Algorithms 17 00261 g005
Figure 5. (a) Patient asleep while using a CPAP device. (b) OSA patient with tissue blocking the airway.
Figure 5. (a) Patient asleep while using a CPAP device. (b) OSA patient with tissue blocking the airway.
Algorithms 17 00261 g005
Table 1. Respiratory monitoring devices and methods [16,17,18].
Table 1. Respiratory monitoring devices and methods [16,17,18].
Measured TechniqueCategoryMeasured ParametersSensor Position
SpirometryLung function testingForced expiratory volume in 1 s (FEV1), forced vital capacity (FVC), peak expiratory flow rate (PEF), forced expiratory flow (FEF)Mouthpiece or nasal/oral area
Full-body plethysmographyLung volume measurementTotal lung capacity (TLC), residual volume (RV), functional residual capacity (FRC)Enclosed chamber
EIP (electromyography of the inspiratory muscles)Inspiratory muscle activity measurementElectrical activity of inspiratory muscles (e.g., diaphragm and intercostal muscles)Surface electrodes on the chest or nasal/oral area
RIP (Respiratory Inductive Plethysmography)Lung volume measurementThoracic and abdominal movement, respiratory rateStraps around the chest and abdomen
Table 2. Keywords used in the search.
Table 2. Keywords used in the search.
KeywordsDatabasesTotal Publications IdentifiedFull-Text Article,
Peer-Reviewed Journal Articles, Conference Proceedings		Primary Selection
Google ScholarIEEE XploreSpringerScience DirectPLoS One
RIP validation906655841311375206165
RIP use in sleep study23794823803978072
RIP machine learning applications4461512295297963
RIP post-operative apnea17802119612795042
Total1767161751482812387415342
Table 3. Criteria for inclusion and exclusion of studies.
Table 3. Criteria for inclusion and exclusion of studies.
Criteria for InclusionCriteria for Exclusion
i.
Studies published between 2012 and 2023
i.
No actual experimentation on the performance of RIP
ii.
Journal articles, conference proceedings, or peer-reviewed articles
ii.
Studies without datasets collected by the authors themselves or with unverified data sources
iii.
RIP belts are implemented in the experimental setup
iii.
Lacking clear description of experimentation setup, procedure, and number of subjects
iv.
Contain results in the form of statistical data, numbers, or definite remarks on the effectiveness of RIP
iv.
Results lacking any definitive conclusion
v.
Results compared to another acceptable standard or validated by a corresponding healthcare professional
v.
Results not compared to any standards, or not verified or validated by professionals in the corresponding field
Table 4. Quality analysis of included studies.
Table 4. Quality analysis of included studies.
StudyQ1Q2Q3Q4Q5
Villar (2015) [9]YesYesYesNot specifiedYes
Harbour (2021) [34]YesYesYesYesYes
Phillips (2017) [35]YesYesYesYesYes
Rahman (2017) [10]YesYesYesYesYes
Strang (2018) [23]YesYesYesYesYes
Sabil (2020) [12]YesYesYesYesYes
Ngo (2013) [44]YesYesYesNot specifiedYes
Robles-Rubio (2020) [11]YesYesYesYesYes
Zhang (2012) [45]YesYesYesYesYes
Ramos-Garcia (2017) [27]YesNot specifiedYesNot specifiedYes
Montazeri (2021) [46]YesYesYesYesYes
Chaari (2020) [36]YesYesYesNot specifiedYes
Lopez-Meyer (2013) [57]YesYesYesYesYes
Tataraidze (2015) [52]YesYesYesYesYes
Retory (2016) [47]YesYesYesNot specifiedYes
Rétory (2017) [43]YesYesYesYesYes
Leutheuser (2014) [37]YesYesYesYesYes
Lo (2017) [48]YesNot specifiedYesYesYes
Magalang (2016) [13]YesYesNot specifiedYesYes
Kogan (2016) [53]YesYesYesYesYes
Park (2020) [54]YesYesYesYesYes
Senyurek (2019) [58]YesYesYesYesYes
Fontecave-Jallon (2020) [59]YesYesYesYesYes
Kovatis (2021) [51]YesYesYesYesYes
Chien (2013) [65]YesYesYesYesYes
Immanuel (2013) [55]YesYesYesYesYes
Mandel (2015) [66]YesYesYesYesYes
Cabiddu (2016) [49]YesNot specifiedYesNot specifiedYes
Gouna (2013) [40]YesYesYesYesYes
Dietz-Terjung (2021) [56]YesYesYesYesYes
Sivieri (2018) [50]YesYesYesYesYes
Leutheuser (2017) [38]YesYesYesYesYes
Silva (2022) [67]YesYesYesYesYes
Hollier (2014) [39]YesYesNot specifiedYesYes
Husom (2022) [60]YesNot specifiedYesYesYes
ElMoaqet (2020) [61]YesYesYesYesYes
Guede-Fernandez (2019) [62]YesYesYesYesYes
Ratnagiri (2021) [42]YesYesYesYesYes
Precup (2012) [41]YesYesYesYesYes
Huang (2018) [63]YesYesYesYesYes
Table 5. Synthetic overview of general applications of RIP.
Table 5. Synthetic overview of general applications of RIP.
StudyApplicationSystem DesignNo. of SubjectsStudy DurationSubject Health StatusTest Setup
Villar (2015)
[9]		Validation of RIP measurements in ambulatory conditionsHexoskin smart shirt20N/A 1HealthyLab
Harbour (2021) [34]Flow reversal (FR), breathing rate (BR)Hexoskin smart shirt12 participants (6 M + 6 F) 2N/AHealthyLab
Phillips (2017) [35]Determining accuracy of Hexoskin smart shirtHexoskin smart shirt6 participants3 days for 3 weeksHealthyLab
Rahman (2017) [10]Breath per minute (BPM) and labored breathing index (LBI), phase, and ribcage %pneuRIP™10 participants (10–17 years old)N/AN/ALab
Strang (2018) [23]Breath per minute (BPM) and labored breathing index (LBI), phase, and ribcage %pneuRIP™43 (21 healthy; 13 M + 8 F) + 22 with NM disease (19 M + 3 F); 5–18 years old; 1 excludedN/AHealthy + children with neuromuscular diseaseHospital
Sabil (2020) [12]Diagnosis and characterization of sleep apneaTS-RIP combined70 patients, adultsLimited 6-month periodSuspected sleep apneaSleep lab
Ngo (2013) [44]Detection of respiratory rate (BPM), identification of low respiratory rate below thresholdCustom RIP belts (chest and abdomen)10 participants (7 M + 3 F; age: 18–21 years)N/AN/AN/A
Zhang (2012) [45]Detection of respiratory volumesChest and abdomen RIP belts15 subjectsN/AN/AN/A
Ramos-Garcia (2017) [27]Detection of respiratory volumes, calibrationSingle thoracic belt, 3 methods: belt, belt with adhesive, belt sewn to shirt3 adults (age: 30 ± 2)N/ARespiratory problems were absentN/A
Montazeri (2021) [46]Detection of respiratory volumes and air flow, reliability comparison between belt typesChest and abdomen RIP belts767 datasets before applying exclusion criteriaJanuary 2009–May 2017Suspected sleep apneaClinical home sleep apnea testing
Retory (2016) [47]Validation of RIP during exerciseChest and abdomen RIP belts7 M + 23 F, adultsN/ANo related respiratory or neurologic diseaseLab
Retory (2017) [43]Validation of RIP in high-BMI 3 subjectsChest RIP belt20 subjects (14 M + 6 F)N/A10 with BMI < 25 kgm−2 and 10 with BMI > 30 kgm−2Lab
Leutheuser (2014) [37]Comparison of RIP calibration techniquesVivoMetrics Lifeshirt system186 subjects (98 M + 88 F), 13 excluded from resultsN/AHealthyN/A
Lo (2017) [48]Calibration of RIPChest and abdomen RIP belts11, adults (age: 18–35)N/AHealthy, non-smokingLab
Magalang (2016) [13]Detection of sleep apneaHome sleep testing, including RIP and nasal pressure monitoring15 random sleep study reportsAll sleep studies were conducted within a quarterSuspected sleep apneaHome
Kogan
(2016)
[53]		Detection of sleep apneaChest and abdomen RIP belts12 (7 M + 5 F)N/ASuspected sleep apneaLab
Park (2020) [54]Verification of RIP detecting sleep apneaChest and abdomen RIP belts200 (100 M + 100 F) sleep study reports, based on severityFebruary 2015–March 2018Suspected sleep apneaLab
Fontecave-Jallon (2020) [59]Detection of pulse wave velocity (PWV)Chest and abdomen RIP belts11 subjects (7 M + 4 F)N/AHealthy, age: 22–53N/A
Kovatis (2021) [51]Detection of work of breathing (WOB), impact of HFNC 4 flow adjustmentsRespibands plus RIP belts, PneuRIP software21 infants in final datasetN/ASubjects requiring respiratory supportNeonatal ICU
Chien (2013) [65]Observation of asynchronous breathing in COPD patients during 6 min walk testChest and abdomen RIP belts88 subjects (age: 18–80 years)March to July 2009COPD patients, FEV1 < 80%, FEV1/FVC <70%N/A
Immanuel (2013) [55]Breathing asynchrony associated with sleep stagesChest and abdomen RIP belts40 children (age: 3.1–12.2 years)N/AHealthyN/A
Mandel (2015) [66]Breathing asynchrony associated with anesthesia and sedationChest and abdomen RIP belts51 adultsNovember 2011–March 2012Patients with anesthesia during surgical procedureHospital
Cabiddu (2016) [49]Validation of RIP during rest and moderate exerciseChest and abdomen RIP belts7 MN/AHealthyN/A
Gouna (2013) [40]Effect of positioning on respiratory patterns and functioningRespitrace Plus19 infants6 monthsPatients in neonatal ICUHospital
Sivieri (2018) [50]Validation of RIP in infantsChest and abdomen RIP belts25 infantsN/APatients in neonatal ICUHospital
Leutheuser (2017) [38]Self-calibration of RIP without external validationVivoMetrics LifeShirtTM193 subjectsN/AHealthyN/A
Silva (2022) [67]Assessment of fatigue produced by repetitive workChest and abdomen RIP belts22 subjects (11 M + 11 F)N/AHealthyN/A
Hollier (2014) [39]Validation of RIP in hypoventilation syndrome in high-BMI subjectsVivoMetrics LifeShirtTM26 subjects (13 control, 13 high-BMI (BMI ≥ 30))N/APatients with obesity hypoventilation syndromeN/A
1 N/A: not available. 2 M: male, F: female. 3 BMI: body mass index. 4 HFNC: high-flow nasal cannula.
Table 6. Results synthesis for general applications of RIP.
Table 6. Results synthesis for general applications of RIP.
StudyAnalysis MethodSamplesCompared AgainstPerformance/Accuracy
Villar (2015)
[9]		Statistical analysis of RIP data5 min of data collection for each measurementSpirometryIntra-class correlation for breathing rate: Ranges from 0.96 to 1.00 for various ambulatory conditions
Max. error (%): 0.48 ± 0.63
Harbour (2021) [34]Custom algorithm for data pre-processing, described in a statistical manner7816 flow reversal events, 3907 breath eventsSpirometryFR 1 detection accuracy: >99% BR 1 detection: r = 0.98
Phillips (2017) [35]Pearson’s correlation coefficient (r)15 min of data collection: days 1 and 2: moderate; day 3: vigorousParvoMedic TrueOne® 2400 Metabolic Cart (VO2 Max)Breathing rate: vigorous exercise: r = 0.996, p = 0.000; moderate exercise: r = 0.962, p = 0.002
Rahman (2017) [10]Statistical representation of RIP data3 min of data collection per subjectManual countingpneuRIP™ is more accurate than Respitrace for BPM count. Differences between two systems compared to spirometry: normal breathing: 13.2% vs. 36.4%; labored breathing: 16.9% vs. 60.7%. LBI is identical for both systems
Strang (2018) [23]Statistical representation of data; ANCOVA model4 min of data collection per subjectN/AMean phase angle: 25.14 (healthy) vs. 55.26 (patient); mean labored breathing index: 1.08 (healthy) vs. 1.25 (patient)
Sabil (2020) [12]Statistical representation of RIP dataFull night’s sleepSignal validated through CIDELEC software’s automatic signal-quality checkingSensitivity and specificity: 96.21% and 91.34% for detecting apneas; 89.94% and 93.25% for detecting hypopneas; and 92.24% and 92.13% for no AHI event Apnea characterization (sensitivity): obstructive, mixed, and central apneas: 98.67%, 92.66%, and 96.14%, respectively
Ngo (2013) [44]Plot respiratory movement graph from RIP signal, detection of BPM1 min of data collection per subject, 2 measurements per subjectManual counting of respiratory rateStatistical difference: 5% Absolute error: ±1 BPM
Zhang (2012) [45]Statistical representation of RIP data30 min per subject, 10,734 breaths in totalFlow Analyzer PF-300Tidal volume (Vt): within ±10%: 93.85%; within ±15%: 98.1%; within ±20%: 99.03%; average Vt best error figure: 0.00%
Ramos-Garcia (2017) [27]Statistical representation of RIP data24 h continuous monitoring per subjectCustom 800 mL respiratory bags, calibrated with spirometerLowest error in belt + shirt combination Root-mean-square error values: 0.039, 0.495, and 0.194
Montazeri (2021) [46]Custom algorithm, random data verification by sleep technologistsDatasets from RIP belt types:
(i) Disposable cut-to-fit: 206;
(ii) Semi-disposable snap-on: 149;
(iii) Disposable snap-on: 256		Cannula flow signalHighest reliability for disposable snap-on belts (mean thorax: 98.5 ± 9.3%, mean abdomen: 98.8 ± 8.9%, p < 0.001); calibrated RIP flow (r > 0.8): highest for disposable snap-on belts (80.6%); higher reliability than cannula
Retory (2016) [47]Custom algorithm7 epochs of 1 min per subjectPnT 2Tidal volume:
Correlation coefficient: 0.81, p < 0.0001
Inspiratory time:
Correlation coefficient: 0.92, p < 0.0001
Expiratory time:
Correlation coefficient: 0.94, p < 0.0001
Retory (2017) [43]Statistical representation of RIP data7 epochs of 1 min per subjectPnTTidal volume:
Correlation coefficient: 0.82, p < 0.0001
in both BMI groups
Respiratory rate:
Correlation coefficient: 0.95 and 0.94 for low- and high-BMI groups, respectively, p < 0.0001 for both
Leutheuser (2014) [37]VivoSense softwareVariable per subject, 5 min of resting-phase data for calibrationFlowmeterBest calibration method: least-squares approximation (LSQ); second-best calibration method: support vector regression (SVR) Verdict: RIP is reliable for ambulant applications, such as exercise
Lo (2017) [48]Statistical representation of RIP data30 min per subjectPneumotachometerTidal volume in both resting and exercise: Correlation coefficient: Intra-subject: 0.7, ranging from 0.5 to 0.9 Group mean: 0.8
Magalang (2016) [13]Statistical analysisOvernight sleep studyPSG validated by expert technologistsAHI scoring bias: Between nasal flow and RIP: 2.9 ± 3.6 events/hour Between square-root-transformed nasal flow and RIP: 1.1 ± 3.6 events/h
Kogan
(2016)
[53]		Both manual and automatic scoring compared with RIPOvernight sleep studyPSG validated by expert technologistsSensitivity: 100% with RIP, 78% with PSG
Specificity: 78% with RIP, 75% with PSG
Park (2020) [54]Statistical analysisOvernight sleep studyPSGCorrelation coefficient with PSG: AHI score: r = 1.0, p < 0.001 Respiratory effort-related
arousal (RERA) score: r = 0.643, p < 0.001
Fontecave-Jallon (2020) [59]Statistical analysis3–4 min of quite breathing + 40 s apnea + 20 s recovery, repeated 10 times per subjectBrachial pulse wave velocity (PWV) measurementLinear correlation with brachial PWV: r = 0.86, p < 0.001
Kovatis (2021) [51]Statistical analysis10 min of data collection followed by 2–5 min stabilization periodSystem validated by studies [10,23]Increased WoB in preterm infants (<28 weeks gestational age): phase angle 87 ± 34° vs. 58 ± 34° for ≥28 weeks gestational age HFNC flow was adjusted based on phase asynchrony data from RIP
Chien (2013) [65]Statistical analysis6 min walk testSpirometry, full-body plethysmographyPhase angles: normal: 10.5 ± 1.1°; moderate COPD: 14 ± 1.1°; severe COPD: 24.4 ± 1.7°
Immanuel (2013) [55]Statistical analysisOvernight PSGSleep stages identified with EEG, EOG, and EMGAsynchronous breathing correlated with low-frequency (0.02–0.05 Hz) energy of RIP signals during REM sleep (r = 0.625, p < 0.01)
Mandel (2015) [66]Statistical analysis, Hilbert–Huang transform (HHT)5 min epoch for each subjectSpirometryCorrelation coefficient between RIP and spirometry: without HHT: 0.62 ± 0.20; with HHT: 0.93 ± 0.07
Cabiddu (2016) [49]Statistical analysis5 min epoch for each subjectErgospirometry(ES)Low agreement between RIP and ES.
For tidal volume, r = 0.23 (rest) and r = 0.25 (exercise)
For respiratory rate, r = 0.52 (rest) and r = 0.46 (exercise)
Gouna (2013) [40]Statistical analysis60–180 min after positioning the infantRIP calibrated with PnT prior to the experimentationLeft lateral and prone positions found to be superior to supine position for respiratory functioning
Tidal volume, Vt = 5.2 mL/kg, phase angle = 50°, %ribcage = 36% for left lateral position; Vt = 5.5 mL/kg, phase angle = 62°, %ribcage = 47% for left lateral position
Sivieri (2018) [50]Statistical analysis0.5–1 hPnTCorrelation coefficient between PnT and RIP: r2 = 0.981 for respiratory airflow, r2 = 0.995 for respiratory volume
Leutheuser (2017) [38]Algorithm developed by author5 min of standing still, exercise until exhaustion, 10 min of recoveryFlowmeterLimits of equivalence, adjustment with standing still prior to exercise: 82.85 ± 19.21% for treadmill running, 76.20 ± 21.15% for recovery phase
Silva (2022) [67]Statistical analysis, Hilbert transform10 min of work for each trial, 3 trialsSurface electromyography (EMG)Both the RIP signal correlation between chest and abdomen and synchrony of phase reduces during fatigue (p < 0.001 for both)
Hollier (2014) [39]Vivologic 3.0 software5 min epoch for each subjectSpirometryRIP only validated for measurement of respiratory rate in control group and not in high-BMI group (limit of agreement (LOA): ±12%)
Agreement with spirometry for respiratory volume measurement not accepted by the authors due to ≥20% LOA
1 FR: flow reversal, BR: breathing rate. 2 PnT: pneumotachography.
Table 7. Synthetic overview of artificial intelligence-based applications of RIP.
Table 7. Synthetic overview of artificial intelligence-based applications of RIP.
StudyApplicationSystem DesignNo. of SubjectsStudy DurationSubject Health StatusTest Setup
Robles-Rubio (2020) [11]Detection of post-operative apnea in infants, classified into 5 breathing patternsAutomated Unsupervised Respiratory Event Analysis (AUREA)21 newborns5–12 h post-surgeryPost-surgical-operation patientsHospital
Chaari (2020) [36]Human activity recognition (HAR)Hexoskin smart shirt40 participantsN/AHealthyLab
Lopez-Meyer (2013) [57]Smoking activity recognitionChest and abdomen RIP belts, hand-to-mouth gesture monitoring with other sensors20 adults (10 M + 10 F)N/ARegular smokersN/A
Tataraidze (2015) [52]Detection of stage of sleepChest and abdomen RIP belts29 participantsN/ANo sleep-related breathing disordersSleep medicine laboratory
Senyurek (2019)
[58]		Detection of smokingSleepSense Inductive Plethysmography30 datasets (19 M + 11 F), 1 removed laterN/AMedium to heavy smokersN/A
Dietz-Terjung (2021) [56]Identification of sleep stagesChest and abdomen RIP belts111 subjectsOctober 2019–January 2020Patients with suspected sleep disordersSleep medicine laboratory
Husom (2022) [60]Estimation of power output from physical activityChest and abdomen RIP belts, heart rate1 subjectN/AHealthyN/A
ElMoaqet (2020) [61]Automated detection of sleep apnea from signal from a single RIP beltChest and abdomen RIP belts 17 subjectsN/APatients with suspected sleep disordersSleep medicine laboratory
Guede-Fernandez (2019) [62]Detection of drowsiness in a driving personSingle RIP belt20 (10 M + 10 F)N/AHealthyDriving simulator, external observers
Ratnagiri (2021) [42]Detection of asynchronous breathing in childrenpneuRIPTM51N/ABoth healthy and patients with neuromuscular diseaseN/A
Precup (2012) [41]Measurement of extubation readiness in preterm infantsRespitrace + ECG56 infantsN/AExtreme preterm infants requiring endotracheal intubation and mechanical ventilationHospital
Huang (2018) [63]Assessment of COPD stageChest and abdomen RIP belts26N/ACOPD patientsHospital
Table 8. Results synthesis for artificial intelligence-based applications of RIP.
Table 8. Results synthesis for artificial intelligence-based applications of RIP.
StudyAnalysis MethodSamplesCompared AgainstPerformance/Accuracy
Robles-Rubio (2020) [11]Binary K-means classifierN/AExpectation maximization (EM)Overall accuracy: 80%
Category-wise confusion:
pause: 22.2%, synchronous breathing: 16.3%, asynchronous breathing: 24.2%, unknown: 25.5%
Chaari (2020) [36]Machine learning5 repeats of 10 different activitiesActual human activity dataOverall accuracy: 95.4%
Lopez-Meyer (2013) [57]Support vector machine (SVM) classifiers5 min of 12 specific activities, 19.56 h of total recorded data, 21,411 respiratory cyclesN/ARecognition of smoke inhalation: highest precision and recall with IAR model (90.11% and 90.04%, respectively)
Tataraidze (2015) [52]Bagging classifier, 33 extracted featuresRecording during polysomnographyPSG 1 validated by expert physiciansAccuracy: 77.85 ± 6.63 (mean ± SD),
Cohen’s kappa: 0.59 ± 0.11 With heuristics: accuracy: 80.38 ± 8.32 Cohen’s kappa: 0.65 ± 0.13
Senyurek (2019)
[58]		Machine learning, SVM, and CNN LSTM classifier models120 smoking sessions, 1694 smoke inhalationsActual smoking eventsSVM: precision: 0.53, accuracy: 0.8, recall: 0.83, F1 score: 0.63; CNN-LSTM: precision: 0.68, accuracy: 0.55, recall: 0.74, F1 score: 0.72
Dietz-Terjung (2021) [56]Nox BodySleepTM algorithom implementing MLOvernight PSGPSGAverage sensitivity and specificity of sleep stages: 0.70 and 0.66, respectively AHI index: r = 0.91 between Nox BodySleepTM and manual scoring
Husom (2022) [60]Machine learning predictive models21 datasetsActual dataModerate accuracy of prediction, best R2 = 0.56 with CNN with 6 feature set
Mean absolute percentage error: 0.20–0.24
ElMoaqet (2020) [61]Recurrent neural network, LSTM, and BiLSTM 2Overnight PSGPSGAccuracy of apnea detection: 84.4%, true-positive rate: 78.5%, true-negative rate: 85.9%, F1 score: 67.4% with abdominal RIP data
Results fell behind nasal pressure data
Guede-Fernandez (2019) [62]Novel algorithm, machine learning, binary classifier36 separate testsExternal observersSpecificity: 96.6%, sensitivity: 90.3%, Cohen’s kappa: 0.75
Ratnagiri (2021) [42]Machine learning model20 training datasets, 31 test datasets (11 healthy + 20 NM 3 patients)Expert scorersAccuracy: 61.3%, sensitivity: 45.5%, specificity: 70% with phase-angle feature set Accuracy: 90.3%, sensitivity: 100%, specificity: 85% with ICP 4 feature set
Precup (2012) [41]Support vector machine (SVM) classifier, AUREA [91]3000 samples of AUREA features for each infantActual success/failure of extubationTraining accuracy for success class: 89.7%; failure class: 85.4% Testing accuracy for success class: 73.6%; failure class: 83.2%
Huang (2018) [63]Hilbert–Huang transform, K-means classifier49 datasets (2 from each subject, 3 excluded)Actual diagnosis of COPD stageClass-wise error: 20%, 34%, and 25%; Error with no-classifying model: 36%
1 PSG: polysomnography. 2 LSTM: long short-term memory, BiLSTM: bidirectional long short-term memory. 3 NM: neuromuscular. 4 ICP: inverse cumulative percentage.
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Rahman, M.S.; Chowdhury, S.; Rasheduzzaman, M.; Doulah, A.B.M.S.U. Artificial Intelligence-Based Algorithms and Healthcare Applications of Respiratory Inductance Plethysmography: A Systematic Review. Algorithms 2024, 17, 261. https://doi.org/10.3390/a17060261

AMA Style

Rahman MS, Chowdhury S, Rasheduzzaman M, Doulah ABMSU. Artificial Intelligence-Based Algorithms and Healthcare Applications of Respiratory Inductance Plethysmography: A Systematic Review. Algorithms. 2024; 17(6):261. https://doi.org/10.3390/a17060261

Chicago/Turabian Style

Rahman, Md. Shahidur, Sowrav Chowdhury, Mirza Rasheduzzaman, and A. B. M. S. U. Doulah. 2024. "Artificial Intelligence-Based Algorithms and Healthcare Applications of Respiratory Inductance Plethysmography: A Systematic Review" Algorithms 17, no. 6: 261. https://doi.org/10.3390/a17060261

APA Style

Rahman, M. S., Chowdhury, S., Rasheduzzaman, M., & Doulah, A. B. M. S. U. (2024). Artificial Intelligence-Based Algorithms and Healthcare Applications of Respiratory Inductance Plethysmography: A Systematic Review. Algorithms, 17(6), 261. https://doi.org/10.3390/a17060261

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop