Availability and performance of image/video-based vital signs monitoring methods: a systematic review protocol
Harford et al. Systematic Reviews
Availability and performance of image/ video-based vital signs monitoring methods: a systematic review protocol
Mirae Harford 0 1
Jacqueline Catherall 0
Stephen Gerry 2
Duncan Young 0 1
Peter Watkinson 0 1
0 Nuffield Department of Clinical Neurosciences, University of Oxford , Oxford , UK
1 Kadoorie Centre for Critical Care and Trauma Research and Education, John Radcliffe Hospital , Headley Way, Oxford OX3 9DU , UK
2 Centre for Statistics in Medicine, University of Oxford , Oxford , UK
Background: For many vital signs, monitoring methods require contact with the patient and/or are invasive in nature. There is increasing interest in developing still and video image-guided monitoring methods that are non-contact and non-invasive. We will undertake a systematic review of still and video image-based monitoring methods. Methods: We will perform searches in multiple databases which include MEDLINE, Embase, CINAHL, Cochrane library, IEEE Xplore and ACM Digital Library. We will use OpenGrey and Google searches to access unpublished or commercial data. We will not use language or publication date restrictions. The primary goal is to summarise current image-based vital signs monitoring methods, limited to heart rate, respiratory rate, oxygen saturations and blood pressure. Of particular interest will be the effectiveness of image-based methods compared to reference devices. Other outcomes of interest include the quality of the method comparison studies with respect to published reporting guidelines, any limitations of non-contact non-invasive technology and application in different populations. Discussion: To the best of our knowledge, this is the first systematic review of image-based non-contact methods of vital signs monitoring. Synthesis of currently available technology will facilitate future research in this highly topical area. Systematic review registration: PROSPERO CRD42016029167
Monitor; Non-invasive; Non-contact; Image; Camera; Vital signs
Monitoring of the basic functions of the human body is
routine practice in healthcare settings from primary care
to critical care. The term ‘vital signs’ typically refers to
measurements including heart rate (HR), respiratory rate
(RR), blood pressure (BP), oxygen saturations (SpO2)
and temperature. Vital signs monitoring methods such
as the electrocardiograph (ECG) and pulse oximetry
have the advantages of being easily available, quick and
non-invasive. In many clinical settings, these widely
available monitors may be sufficient. However, they do
have disadvantages which may limit their use in high
Examples of their limitations include the following:
ECG monitoring relies on application of three or
more ECG electrodes on the patient. These consist
of conducting gel surrounded by an adhesive area.
Signal acquisition relies on good contact between
the gel and the patient’s skin and this process is
prone to errors from electrostatic and
electromagnetic interference and skin impedance.
The small amplitude bioelectrical signals measured
are prone to interference from other biological
signals or machinery.
A single point of measurement may limit the
amount of information being collected [
broader receiving system may offer more detailed
spatial information simultaneously from multiple
sites, allowing derivation/mapping of physiological
parameters. Ultimately, this may facilitate insights
that would be difficult to appreciate from a
There are infection risks from being physically
attached to monitors and the use of the same
machines on multiple patients.
Monitors that require contact with the skin may
lead to skin barrier breakdown. This is a concern in
critical care patients who are more prone to
hospital-acquired pressure ulcers and impaired skin
], especially in neonates [
Multiple lines for various forms of monitoring create a
hazard for both patients and staff, particularly during
transfer. They may also interfere with clinical care.
Multiple contact monitors can interfere with sleep
which is known to adversely affect recovery [
Non-contact monitoring is likely to be less
emotionally traumatic for visitors in critical care
Some of these limitations and complications may be
avoided by using non-contact, non-invasive monitoring
methods in the form of image/video-based monitoring.
There is a huge potential for this type of monitoring in
the current era of telemedicine, and the required
technology has developed rapidly in the last decade. Several
image-based technologies have been explored for vital
signs monitoring, and it is our aim to provide a
summary of their current availability and performance.
We aim to review non-contact, non-invasive image/
video-based vital signs monitoring devices in health care
and health-related fields. To our knowledge, this will be
the first systematic review of the evidence for this type
of monitoring. The review will provide a basis for further
research using similar technology in a rapidly evolving
Protocol and registration
Methods of the systematic review have been developed
in accordance with the Preferred Reporting Items for
Systematic Reviews and Meta-Analyses (PRISMA) [
guideline. The completed PRISMA-P [
] checklist is
available as a supplementary file to this protocol
(Additional file 1).
This proposed systematic review has been registered
with PROSPERO (the International prospective register
of systematic reviews): CRD42016029167.
validated reference device will be included without
limitation on the setting. We will exclude temperature
monitoring which lends itself to infrared imaging and has
been extensively studied [
]. Technology which
require placement of trackers or monitors on the body will
be excluded. Studies using projected markers but require
image/video analysis for monitoring purposes will be
included. No limitations will be set on the primary objective
of the studies; for example, studies aiming to refine image
analysis will be included as long as they measure one of
the specified vital signs and compare it to a clinically
validated reference device as part of the methodology.
Initial screening process will select any literature
reviews to ensure all original studies are included. Only
human subject studies will be included, and no age
limits will be set. Date and language restrictions will not
be applied. Authors of studies published in languages
other than English will be contacted by email for
assistance with data extraction.
All unpublished studies found and articles in the
media will be included wherever possible to minimise
We will include randomised controlled trials,
interventional studies, observational studies (including
casecontrol or cohort studies) and pilot studies that compare
an image-based monitoring system to any established
forms of monitoring already used in clinical practice.
Only studies that compare results from an image
methodology with a clinically validated reference non-image
methodology will be included. Any case series or reports
will be included in the first instance with no limitations
on the total number of subjects.
The primary intended outcome is a summary of the
current availability of image/video-based vital signs
monitoring methods and a comparison of the
performance of the new technology with other clinically available
Secondary outcomes of interest are the quality of the
studies’ methodology with respect to the Guidelines for
Reporting Reliability and Agreement Studies (GRRAS)
], current limitations in image/video-based
monitoring technology and the feasibility of its use in different
It is probable that there will be considerable variability
in the outcome measurements and the reporting of the
technology’s efficacy. We will aim to produce a
qualitative synthesis if the heterogeneity between the studies
does not allow a quantitative summary.
All papers looking at monitoring of HR, BP, RR or SpO2
using image analysis with comparison to a clinically
Searches will be performed in multiple databases
including MEDLINE, Embase, CINAHL and Cochrane library.
Given the subject topic, IEEE (Institute of Electrical and
Electronics Engineers) Xplore Digital Library and ACM
(Association for Computing Machinery) Digital Library
will also be searched. OpenGrey will be searched for any
unpublished grey literature. Given the popular culture/
commercial applications of the technology,
Google/Google Scholar will also be searched.
If any websites are available, primary original research
data from which the technology is based will be sought.
Identified references will be downloaded to reference
library software for the initial title and abstract
screening. All references proceeding to the next stage will be
handled using Distiller SR software (Evidence Partners,
Ottawa, Canada). All searches will be saved for
referencing, including a full list of papers and timeline of
searches. The process of each paper will be clearly
documented from the screening and review process.
The search strategy will be guided by a medical librarian
who will advise on the conduct of the searches and
limitations of the search terms used. Broad search terms will
be used in order to capture all publications on this topic.
Initial search terms of interest include
Measure/Respiratory rate/Breathing rate/Pulse/Heart
Where appropriate, the terms will be exploded and
MeSH (Medical Subject Headings) terms will be used.
As demonstrated in the example, word stems will be
used to generate results where appropriate.
Once searches are complete, the findings will be
pooled and any duplicates removed.
Draft search strategy is outlined in Table 1.
software. The extraction output will be
compared and a final data extraction form will be
created. All full texts will then be assessed by
the two reviewers. Authors will be contacted
by email for clarification or for further data
Stage 4: Once articles are selected, hand-searching
strategy will be used to review their references
to look for relevant articles not already
identified. This is of particular importance as
pseudohealthcare use of non-contact monitoring or
non-published articles may be identified using
Study selection and data extraction
A four-stage review process will be used to select final
articles for inclusion.
The returned papers will be categorised under the
predominant technology used. Anticipated headings include
Stage 1: Once searches are generated and checked for
duplicates, the titles will be screened by two
reviewers (MH, JC) to reject any which are
clearly unrelated. Any articles with uncertainty
or where there is a disagreement between the
reviewers will be included for abstract review.
Stage 2: Abstract review will be undertaken by two
reviewers (MH, JC). Where there is a
discrepancy, the article will proceed to a
fulltext review along with others selected by both
Stage 3: A selection of full-text articles will be assessed
by two reviewers (MH, JC) using a pilot data
extraction form on the reference management
Non-contact photoplethysmography (PPG)/imaging PPG
Multi-slice image and tracking
Tracking of projected light
The above categories will be included in the data
extraction form and effort will be made to group similar
technologies together for analysis. We expect to stratify
studies by predominant imaging modality, vital signs of
interest and the target population.
Outcomes and prioritisation
The following data will be extracted from each study:
the type and date of publication, funding source, conflict
of interest, study type, vital sign of interest, setting, the
number of participants, population and cohort data,
eligibility criteria, image technology used, reference
method, distance from camera to subject, body part
imaged, frames per second, simulated physiological changes,
performance of new technology compared to reference
method, limitations reported, statistical analysis method,
main conclusions and data to allow quality of reporting
assessment according to GRRAS guidelines.
The primary intended outcome is a qualitative review
of the availability and performance of different
technology used for each vital sign. The studies will be
categorised according to the video/image-based monitoring
used (e.g. PPG, thermal) and the vital sign being
measured (i.e. HR, BP, RR, SpO2) during the data extraction
A secondary intended outcome is an appraisal of
the quality of reporting among the method
comparison studies included in the review with respect to the
modified GRASS scoring system (Table 2).
Furthermore, a summary of the current limitations of
noncontact, non-invasive technology, potential methods
of improving image acquisition and the application of
the technology in different clinical populations will be
Assessment of study quality
The appraisal will be performed by two reviewers (MH,
JC) who will agree on the final assessment. In the event
of a discrepancy, this will be resolved by discussion with
a third reviewer (PW or DY).
Many studies of interest for the purpose of this review
are likely to be experimental/pilot/proof-of-concept
studies using a small number of subjects from a single
centre. It is anticipated that it may not be possible to
weight studies numerically in the final analysis due to
the likely variability in the study designs and variable
GRRAS recommendation for method comparison
studies reporting is designed to cover a broad range of
clinical test scores/classification/diagnosis and not all
applicable for our use. Therefore, they have been adapted
to suit our objectives (Table 2).
A key aspect of quality assessment is the
appropriateness of the statistical analysis. Pearson’s correlation
coefficient is not an appropriate statistical method to assess
reliability and agreement, as originally discussed by
Bland and Altman [
]. Despite this, a review of
statistical methods used to test for agreement of medical
instruments showed that the correlation coefficient
remains the second most frequently used statistic [
order to assess this aspect of each study, the statistical
analysis will be checked against the published method
agreement analysis guidelines [
]. The classification
strategy based on the guidelines is summarised in
Table 3. Each study will be assessed for statistical
appropriateness. This has been incorporated to the modified
GRRAS scoring system (Table 2).
Where one method is used, or all methods used are in
the same category, the overall rating will be given as that
category. Where multiple methods are used and they are
in adjacent categories, the study will be categorised as
the better. If gold standard and inappropriate methods
are used together, the study will be classified as
It is expected that Bland and Altman’s method will
be used most frequently for measuring agreement.
Specifically, it is important that where multiple
observations are made per individual (e.g. multiple HR
measurements from one subject using two methods),
this is accounted for in the statistical assessment [
This has been incorporated into the statistical rating
scale (Table 3).
Each included study will be given a rating based on
the modified GRASS scoring system as below:
All items are “+”: high quality
All items are “+” except the statistical analysis rating
of “?” OR one of the other sections is given “-”:
Statistical analysis is given “-” OR two or more
sections are given “-”: low quality
Finally, the GRADE methodology will be used to
systematically rate the strength of the overall review as
high/moderate/low/very low [
There is likely to be a high degree of heterogeneity
across the studies given the likely variation in reported
outcomes. It is hoped that it would be possible to
compare methods post stratification by the vital signs of
interest (e.g. HR) and by the predominant technology
We will seek missing data by contacting original authors.
Effort will be made to include all available studies.
Subgroup analysis will be undertaken for the group of
studies with acceptable statistical methods and outcomes
summarised in numerical format if possible. Should the
extracted data be deemed sufficiently homogenous from
a clinical standpoint for quantitative synthesis to be
undertaken, we anticipate performing numerical analysis.
If we identify studies that are sufficiently homogenous,
we will attempt a meta-analysis of the results, using
Title and abstract Identify in title or abstract that interrater/intrarater reliability
or agreement was investigated
Explain how the sample size was chosen. State the determined
number of raters/subjects/objects/replicate observations
Explains how the sample size was chosen and/or state
the determined number of subjects/replicate observationsb
The image-based vital signs monitor is described
Specifies the subject population of interest
No population specified
Describes what is already known about the reliability
of image-based monitoring method and provide a
rationale for the study
No statement on the current knowledge of the method
and no rationale stated
No explanation of sample size. Number of subjects/replicate
observations not stated
Describes sampling method
No description of sampling method
Describes the measurement process
No description of measurement process
Two (or more) methods of measurements conducted
Measurements not conducted independently
Describes the statistical analysis planned
No description of statistical analysis planned
States the actual number of subjects who were included
and number of observations (e.g. duration of recording)
No statement of the number of subjects/observations
Reports estimates of reliability and agreement including
measure of statistical uncertainty using gold standard
Reports estimates of reliability and agreement including
measure of statistical uncertainty using acceptable standard
Reports estimates of reliability and agreement including
measure of statistical uncertainty but using inappropriate
measures/no estimates of reliability/agreement given
Discusses the practical relevance of results
No discussion of practical relevance of results
Describe sampling method
Describe the measurement/rating process (e.g. time interval
between repeated measurements, availability of clinical
State whether measurements/ratings were conducted
Describe the statistical analysis
State the actual number of raters and subjects/objects
which were included and the number of replicate
observations which were conducted
Describe the sample characteristics of raters/subjects
Report estimates of reliability and agreement including
measure of statistical uncertainty
Discuss the practical relevance of results
Provide detailed results if possible (e.g. online)
N/A not applicable
aStudies will not be penalised for not stating the gold standard/reference method of monitoring in the title/abstract as we aim to include studies where this
comparison was performed with an alternative primary aim (e.g. improve image analysis protocol)
bPilot studies will not be penalised for not stating how the sample size was chosen
random-effects methods to combine measures of bias
and agreement from multiple Bland-Altman analyses
]. Similar techniques may also be used to combine
measures of accuracy and error. If significant
heterogeneity exists between the types of study making the
numerical subgroup analysis impossible, the analysis will
be qualitative. If there is sufficient stratified data for the
application or performance of any particular imaging
modality or detected vital sign, a statistical subgroup
analysis will be performed.
Amendments to protocol
Any deviation from this protocol will be dated and
documented to ensure a transparent review process. As per
PRISMA guidelines, no changes will be made to the
main body of the protocol and any unanticipated
additional findings will be discussed in the final review.
This systematic review will synthesise a summary of the current
technology available for image-based monitoring of vital signs.
The main strength of the proposed review is that a
summary of the availability and effectiveness of non-invasive,
non-contact monitoring methods and their utility in the
clinical environment will provide a basis for the direction
of future work in the field. Furthermore, an appraisal of
the quality of reporting among these studies will guide
authors for reporting of future studies. Similarly, a review of
the statistical methods employed among these studies will
improve the analysis of future studies.
Despite the outlined wide search strategy, we anticipate
a number of difficulties in carrying out the proposed
systematic review. There is high likelihood of publication
bias as the published studies/imaging methods are more
likely to have better performance compared to those that
are not published. We also anticipate that studies
conducted as part of doctoral theses may not be found
despite the wide search strategy and/or may be difficult to
access. It is likely that a proportion of these studies will
have been published in peer-reviewed journals with
further likelihood of publication bias. Every effort will be
made to include grey literature available with review of
citations/references and advice from a librarian.
We also anticipate that a large number of studies
included in the review will be proof-of-concept studies
with very few subjects. Therefore, we anticipate a
heterogeneous group of studies to be included in the review
which may make quantitative summary difficult.
We believe that this review will provide a useful
summary of the current technologies available for
imagebased monitoring of vital signs and guide the direction
of future research for more complex monitoring,
especially in critical care settings.
Additional file 1: PRISMA-P. (DOCX 21 kb)
BP: Blood pressure; ECG: Electrocardiograph; GRRAS: Guidelines for Reporting
Reliability and Agreement Studies; HR: Heart rate; ICU: Intensive care unit;
LOA: Limits of agreement; PPG: Photoplethysmography; PRISMA: Preferred
Reporting Items for Systematic Reviews and Meta-Analyses; RR: Respiratory
rate; SpO2: Oxygen saturations
The authors would like to thank Tatjana Petrinic for her guidance during
development of search strategy.
This work is supported by the National Institute for Health Research. The
funder supports salaries for this work but has had no role in the
development of this protocol.
Availability of data and materials
MH and PW conceived the original idea. MH and JC developed the search
strategy. MH and SG developed the statistical assessment and analysis
strategy. MH has written this manuscript with revisions by JC, SG, DY and
PW. All authors read and approved the final manuscript.
Ethics approval and consent to participate
Consent for publication
The authors declare that they have no competing interests.
Springer Nature remains neutral with regard to jurisdictional claims in
published maps and institutional affiliations.
1. Sun Y , Papin C , Azorin-Peris V , Kalawsky R , Greenwald S , Hu S . Use of ambient light in remote photoplethysmographic systems: comparison between a high-performance camera and a low-cost webcam . J Biomed Opt . 2012 ; 17 ( 3 ): 037005 .
2. Mestha LK , Kyal S , Xu B , Lewis LE , Kumar V . Towards continuous monitoring of pulse rate in neonatal intensive care unit with a webcam . Conf Proc IEEE Eng Med Biol Soc . 2014 : 3817 - 20 .
3. Johnson DE. Extremely preterm infant skin care: a transformation of practice aimed to prevent harm . Adv Neonatal Care . 2016 ; 16 ( Suppl 5S ): S26 - 32 .
4. Kamdar BB , Needham DM , Collop NA . Sleep deprivation in critical illness: its role in physical and psychological recovery . J Intensive Care Med . 2012 ; 27 ( 2 ): 97 - 111 .
5. Hardicre J. Meeting the needs of families of patients in intensive care units . Nurs Times 2003 ; 99 ( 27 ): 26 - 27 .
6. Moher D , Liberati A , Tetzlaff J , Altman DG , The PRISMA Group. Preferred Reporting Items for Systematic Reviews and Meta-Analyses: the PRISMA Statement . PLoS Med . 2009 ; 6 ( 7 ):e1000097. doi: 10 .1371/journal. pmed1000097.
7. Moher D , Shamseer L , Clarke M , Ghersi D , Liberati A , Petticrew M , Shekelle P , Stewart LA . Preferred Reporting Items for Systematic Review and MetaAnalysis Protocols (PRISMA-P) 2015 Statement. Syst Rev . 2015 ; 4(1):1 . doi: 10 .1186/2046-4053-4-1.
8. Aubakir B , Nurimbetov B , Tursynbek I , Varol HA . Vital sign monitoring using Eulerian video magnification and thermography . Conf Proc IEEE Eng Med Biol Soc . 2016 : 3527 - 30 .
9. Wang K , Gill P , Wolstenholme J , Price CP , Heneghan C , Thompson M , Pluddemann A . Non-contact infrared thermometers for measuring temperature in children: primary care diagnostic technology update . Br J Gen Pract . 2014 ; 64 ( 627 ): 681 - 3 .
10. Lam H , Kostov Y , Tolosa L , Falk S , Rao G . High resolution non-contact fluorescence based temperature sensor for neonatal care . Meas Sci Technol . 2012 ; 23 ( 3 ): 035104 .
11. Kottner J , Audige L , Brorson S , Donner A , Gajewski BJ , Hrobjartsson A , Roberts C , Shoukri M , Streiner DL . Guidelines for Reporting Reliability and Agreement Studies (GRRAS) were proposed . J Clin Epid . 2011 ; 64 : 96 - 106 .
12. Bland JM , Altman DG . Statistical methods for assessing agreement between two methods of clinical measurement . Lancet . 1986 ; 1 ( 8476 ): 307 - 10 .
13. Zaki R , Bulgiba A , Ismail R , Ismail NA . Statistical methods used to test for agreement of medical instruments measuring continuous variables in method comparison studies: a systematic review . PLoS One . 2012 ; 7 ( 5 ): e37908 .
14. Watson PF , Petrie A . Method agreement analysis: a review of correct methodology . Theriogenology . 2010 ; 7 : 1167 - 79 .
15. Bland JM , Altman DG . Agreement between methods of measurement with multiple observations per individual . J Biopharm Stat . 2007 ; 17 ( 4 ): 571 - 82 .
16. Belsham H , Helfand M , Schunemann HJ , Oxman AD , Kunz R , Brozek J. GRADE guidelines: 3. Rating the quality of evidence . J Clin Epidemiol . 2011 ; 64 : 401 - 6 .
17. Tipton E , Shuster J. A framework for the meta-analysis of Bland-Altman studies based on a limits of agreement approach . Stat Med . 2017 ; doi: 10.1002/sim.7352. [Epub ahead of print]