Results and submissions : REF 2021

Advancing the Development, Public Understanding and Benefits of Global Weather F ...

Submitting institution: The University of Surrey
Unit of assessment: 10 - Mathematical Sciences
Summary impact type: Environmental
Is this case study continued from a case study submitted in 2014?: No

Download case study PDF

1. Summary of the impact

To increase the benefits to society of weather forecasting, meteorological agencies need to develop more accurate prediction algorithms and invest heavily in computing infrastructure. University of Surrey research on data assimilation algorithms and the nonlinearity and geometry of models underpinning weather forecasting has guided the development of improved algorithms at the UK Met Office and Météo-France. The research has also informed and inspired a book and television documentary series promoting public understanding of weather forecasting.

2. Underpinning research

A series of meetings organised by Professor Roulstone at the Isaac Newton Institute in 1996, accompanied by two volumes of proceedings published in 2002, launched a world-leading programme on Data Assimilation (DA) and Numerical Weather Prediction (NWP) focusing on incorporating qualitative knowledge about atmospheric behaviour into quantitative models and algorithms, and on the roles of nonlinearity and geometry in this process. This section describes key contributions made by Roulstone and colleagues to this programme since his appointment at the University of Surrey in 2004.

Data Assimilation for Weather Forecasting

Four-Dimensional (space + time) Variational DA (4DVar) has been central to NWP for two decades. Forecasts are made using parameters obtained by minimising cost functions with two terms representing observation errors and model-dependent ‘background’ errors. For the latter traditional 4DVar uses a linear climatological model to provide flow-dependence over the assimilation window while ensemble DA methods estimate it by integrating a set of model states with slightly different initial conditions. The ensemble approach is potentially more accurate, and avoids costly maintenance of linear climatological models, but suffers from significant sampling errors if the ensemble size is too small, and so is itself computationally expensive. In 4D Hybrid Ensemble Variational (4DEnVar) DA this is mitigated by replacing the linear climatological model in traditional 4DVar by an appropriate linear combination of ensemble trajectories, a process called ‘localisation’.

Unfortunately, comparisons using near operational NWP systems demonstrated inferior performance of ensemble DA methods relative to 4DVar. To understand the reasons for this, research undertaken with Met Office Scientists [R1], compared traditional 4DVar and 4DEnVar for a toy model. It showed that if a perfect model is used the 4DEnVar outperforms 4DVar if the ensemble is large enough, but is worse in the presence of large sampling errors and model errors. This is because the localisation operator in 4DEnVar does not commute with the model, resulting in the background errors not following the flow. The difference in performance is considerably reduced if some stochastic noise is added to the 4DEnVar ensemble members at each step, a process known as ‘additive inflation’.

Nonlinearity and Geometry in Forecasting

A fundamental challenge facing NWP is the development of low-cost, accurate simulators of atmospheric models. These simulators are integrated into ensemble and DA schemes to mitigate the ‘butterfly effect’ caused by strong nonlinearities.

Following a strategy successful in other contexts, Roulstone and colleagues have explored approaches preserving the geometry and conserved quantitates of fluid and atmospheric models. The conserved quantities include Potential Vorticity (PV), a central concept in weather forecasting. The study [R3] revealed the relationship between several atmospheric model approximations that preserve conserved quantities while further work [R4] used the conservation of PV to propose a new, more efficient splitting of model variables for DA schemes. More broadly, how the Hamiltonian structure that underpins many fluid and atmospheric models might be integrated into DA schemes was demonstrated [R2]. Further, it was shown that quaternionic geometry can be used in an efficient reformulation of fluid flow equations [R5] while the geometry of differential forms can be used to characterise coherent structures in fluids [R6].

4. Details of the impact

Research by Professor Roulstone and colleagues has led to the development of more accurate prediction algorithms for meteorological agencies and has promoted public understanding of the mathematics and science of weather forecasting.

More accurate, longer-term and more local weather forecasts are of considerable benefit to society through impacts on agriculture, transport, flood-control, and water and energy supplies. The Surrey research both advances the ability of meteorological agencies to increase these benefits by directly increasing agencies’ forecasting skills and by increasing public understanding of weather forecasting.

Developing accurate prediction algorithms

Surrey research has contributed significantly to guiding research and development at the UK Met Office and Météo-France, specifically in developing accurate prediction algorithms.

In 2020, the Met Office – a worldwide provider of weather forecasts – was awarded £1.2 billion of public funds for its next generation forecasting environment centred on a new supercomputer [S1]. The business case was underpinned by a major programme of work on Data Assimilation and Ensembles (DAE) designed to match forecasting algorithms to the computing infrastructure. One of the Met Office’s key strategic decisions was around how to develop 4DVar and if an ensemble approach (4DEnVar) offered a promising alternative. In making such a significant change to forecasting, the Met Office drew on evidence from Surrey’s research [R1], stating that it “ contributed significantly to the knowledge that underpinned strategic decisions about data assimilation methods the Met Office should develop” [S2].

As a direct result of the findings from the Surrey work [R1], the Met Office undertook further investigations over the period 2014-2018, culminating in the internal report “ The development of a hybrid 4D-ensemble variational assimilation system”. The Met Office Expert Scientist confirmed that “ solving or mitigating the localisation problems highlighted in [R1] is needed to realise the full potential of 4DEnVar in the new system” [S2]. The Met Office therefore, took the strategic decision to continue developing 4DVar alongside 4DEnVar [S2].

The research insights gained by the Surrey research have been further exploited, via Fairbairn, at the Météo-France (2014-2017), the national forecasting agency for France and its overseas territories, where research is also “ *actively addressing the issues related to 4DEnVar demonstrated in [R1], including those associated with localization*” [S3].

Increasing Public Understanding of Weather Forecasting

The book, Invisible in the Storm: The Role of Mathematics in Understanding Weather by Roulstone and John Norbury (Princeton University Press, 2013) describes “ how the development of mathematical ideas, combined with modern computer technology, has completely transformed our ability to understand and predict the weather” (endorsement by Roger Penrose). Reflecting the research at Surrey [R1-6] it “ picks apart the challenge” of forecasting chaotic systems, describing how DA and ensemble forecasting mitigate the butterfly effect and outlining the role of geometry in applications of PV conservation [S4]. The book has been translated into German (2019) and has sold over 2,000 copies worldwide (up to August 2020) [S5].

A review by the European Mathematical Society stated, “ The authors have done brilliant work to collect a huge amount of historical information, as well as mathematical information, but keeping always a level in the explanations that makes the text accessible to undergraduate students in the first years, and even to people not so familiar with mathematics” [S6].

Further reviews illustrate the reach of Invisible in the Storm:

“As a TV weather forecaster for over forty years, I have always maintained that meteorology depends on mathematics for meaning. Making this conclusive point, Invisible in the Storm takes readers on an intriguing journey through the history of meteorology, revealing the critical role of mathematics from the earliest days of weather predicating to the current age of computer-generated forecasts. This book guides you inside the storm, where maths’ importance is clearly visible.” – chief weather forecaster at ABC-7 News/KGO-TV, San Francisco [S7].

“I recommended Invisible in the Storm both to mathematics undergraduates and educators who are interested in applied mathematics, weather forecasting, or both.” – Mathematics Teacher [S7].

In 2015, Roulstone and Norbury were awarded the American Meteorological Society’s Louis J. Battan Author’s Award for illuminating “ the mathematical foundation of weather prediction with lucid prose that provides a bridge between meteorologists and the public.” [S8]

In 2016, Roulstone was a consultant for a three-part BBC Four documentary, Storm Troupers: The Fight to Forecast the Weather (BBC4, May 2019), initially watched by over half a million viewers before repeats and further distribution [S9].

Producer James Sandy from Keo Films said that Invisible in the Storm was “ an invaluable resource”, “ both accessible and captivating”, that enables them to present weather forecasting as a “ triumph of modern mathematics and physics that stands alongside breakthroughs such as mapping the human genome.” [S10]

Influencing school start times, permanent daylight saving and shift workers: Mathematical ...

Submitting institution: The University of Surrey
Unit of assessment: 10 - Mathematical Sciences
Summary impact type: Societal
Is this case study continued from a case study submitted in 2014?: No

Download case study PDF

1. Summary of the impact

Uniquely, our work has focused on the mathematical modelling of light, sleep, circadian rhythms and societal constraints to produce quantitative predictions of sleep timing across the human lifespan. Our resulting mathematical framework has: (i) informed policy on delaying school start times for adolescents in the USA and UK; (ii) influenced decisions on whether to move to permanent daylight saving time in the USA and Europe; (iii) resulted in changes to shift patterns for 2,600 engineers working for a major UK utility company; (iv) been used in educational tools aimed at the general public to explain the fundamental mechanisms underlying sleep and circadian rhythm and the role of light.

2. Underpinning research

In the natural world, the ~24-hour light-dark cycle is the primary regulator of our circadian rhythms. However, we have radically changed our exposure to the natural light-dark cycle by such means as using artificial lights after dusk and exposing ourselves to light from screens. This disruption of the natural light-dark cycle is known to result in reduced or mistimed sleep, with consequences to health, including an increased risk of diabetes, obesity, cardiovascular disease, and impaired vigilance and cognitive decline.

The biological mechanisms underlying the health problems resulting from disruption to sleep are still poorly understood. Mathematical models are an important tool to further our understanding of these underlying phenomena and the regulation of the sleep-wake cycle.

Mathematical modelling and analysis

Work by Prof AC Skeldon, Prof G Derks (Mathematics, Surrey), and Prof D-J Dijk (Director, Surrey Sleep Research Centre) has critically demonstrated that the seminal two-process model that encoded the two principal biological mechanisms underlying sleep – homeostasis and circadian rhythmicity (Borbély, Hum Neurobiol. 1982;1:195-204; Daan, Beersma, Borbély Am J Physiol Integr Comp Physiol. 1984;246:R161-R183) and the well-known extended neuronal model (Phillips & Robinson, J Biol Rhythms 2007;22:167-79), can be explicitly related to each other and that the two-process model can be reduced to a one-dimensional map that may have discontinuities [R1]. Furthermore, the team have comprehensibly analysed the bifurcations and dynamics of such models [R2]. These analyses provide a critical bridge between sleep specialists and computational scientists, underpin our subsequent models and explain many published simulation results.

Working with Prof AJK Phillips (Harvard/Monash), the team developed a quantitative mathematical framework that incorporates the primary biological mechanisms, the primary external drivers (the light environment) and social constraints (e.g., getting up for work/school) and used this model to explain observed changes in sleep timing and duration across the lifespan [R3] and to systematically examine the role of the modern light environment [R4].

No other models currently exist that have as their focus the interaction of light and societal constraints on the timing of sleep and circadian rhythms.

Understanding the implication of social constraints on light consumption

The quantitative mathematical framework has relevance for understanding the effects of enforced social constraints, such as school start times, daylight saving, and shift-working on the sleep-wake cycle. This is of particular value because light exposure is a modifiable risk factor for ill-health.

1. Relevance to school start times: Sleep deprivation during the week and catching up at the weekend is particularly characteristic of adolescents and has led to calls in both the UK and USA for schools to start later. The team’s work [R3, R4] highlights that unless we adequately manage our light environment, moving school start times will not reduce sleep deprivation. Instead, delaying school start times will result in a parallel delay to bedtimes and wake times, resulting in no/minimal reduction in sleep deprivation. Depending on the start time and individual characteristics, changing the light environment may be a more effective intervention.

2. Relevance to permanent daylight saving: There is international interest in abandoning the biannual clock change between standard time and daylight-saving time, with campaigners favouring a move to permanent daylight-saving time. Skeldon and Dijk [R5] demonstrated that under permanent daylight-saving, our endogenous biological clock, entrained by the light-dark cycle, would be out of sync with the numbers on the clock (i.e., noon would be solar noon +1 hour). This would exacerbate problems with waking in the morning in time for work/school, and instead, it is a move to permanent standard time that would reduce sleep-deprivation.

3. Relevance to shift work: Fatigue resulting from shift work is a health and safety concern for employers. Existing fatigue risk assessment tools are limited. Research conducted by Skeldon [R6] with Transport for London evaluated existing Fatigue Risk Tools, established their limitations and identified an urgent need for improved biomathematical models.

4. Details of the impact

1. Informing national and international legislation on school start times

USA

School start times for middle- and high-school children can be as early as 7:00 am and have been linked to weaker academic performance, more car accidents, and increased absenteeism and exclusions. Our research [R4] informed the Californian State Bill (No. 328, Portantino) requiring California’s middle and high schools to begin school no earlier than 8:30 am. The Bill’s proposer, State Senator Anthony Portantino, told the New York Times that the Bill's benefits would include “better grades, reduced risk of depression and fewer vehicle accidents – that he said was borne out by research” [S1]. Supported by the American Academy of Pediatricians and the Centres for Disease Control, the State Bill was passed by Senate on August 31, 2018 and must be implemented in all California middle- and high-schools by July 1, 2021 [S2], directly impacting three-million school children. Our research also informed accompanying debates held in 2017 and was quoted in the Senate Bill Policy Committee’s analysis conducted in 2019 for the amended Bill [S3].

UK

An online petition titled ‘School should start at 10 am as teenagers are too tired ’ was started in November 2018 and rapidly exceeded the threshold to trigger a Parliamentary debate (179,000 signatures). This debate, ‘ Secondary School Opening Hours’, took place on 11 February 2019. Our research [R4] and accompanying Press Release titled ‘Mathematicians predict delaying school start times won’t help sleep deprived teenagers ’ were included in the Debate Pack [S4] provided to MPs. Our conclusion that “[our] model suggests that an alternative remedy to moving school start times in the UK is exposure to bright light during the day, turning the lights down in the evening and off at night” was explicitly highlighted for MPs [S4, pp.6]. During the debate [see transcript, S5], Martyn Day (MP Linlithgow and East Falkirk, SNP) cited our conclusion by way of opposition, adding our further conclusion that “ … body clocks would drift even later in response to later start times, and, in a matter of weeks, [adolescents] would find it just as hard to get out of bed.” Nick Gibb (The Minister for School Standards) also cited our research as he spoke in opposition to the proposed delay to school start times, stating, “The mathematical model showed that delaying school start times in the UK would not help reduce sleep deprivation” and “The mathematical model shows that the problem for adolescents is that their light consumption behaviour interferes with their natural interaction with the environmental clock – getting up late in the morning results in adolescents keeping the lights on until later at night. ” However, Mr Gibb did acknowledge that our results “lend some support to delaying school start in the US,” where start times are earlier than the typical UK start time. The final response from the Government was that “The Department [for School Standards] has no plans to require secondary schools to start later .” A decision that aligns with our research findings.

2. Informing public debate on permanent daylight-saving time

Our paper [R5] has been cited frequently in international debates on daylight-saving, News Media: 20 news stories, including Fox, CNN, and shared 129 times via Social Media: including 33 Facebook pages; 96 Twitter Tweets. It has also been used by the Save Standard Time Campaign in California to support their position [S6]. Skeldon was invited to co-author the Op-Ed article ‘How New York should lock the clock: don’t switch to daylight saving time, ditch it and stick with standard time’ published in the New York Daily News (daily circulation ~200,000), which warned that the proposed permanent change to daylight saving would have a negative effect on the health of 19.5 million New Yorkers [S7]. In a May 2019 CNN interview, Assemblyman Kansen Chu – who proposed the California permanent daylight saving bill (AB7) – was countered by the interviewer, who asked Chu to comment on the findings of Skeldon and Dijk's research [R5]. Later in September 2019, the San Francisco Chronicle reported that Chu “is now considering a proposal for year-round standard time instead.” [S8]

3. Impact on shift working patterns

On-call shifts are common in the utilities and transport industries. These shifts involve engineers working a standard (9am-5 pm) day, then subsequently working out-of-hours call-outs in response to urgent or unpredictable situations (i.e., gas leaks). An on-call worker may receive no call-outs or several, taking varying times to resolve. This unpredictability makes it challenging to model potential sleep deprivation and effectively manage employee working schedules, resulting in a health and safety risk. Existing Health and Safety Executive (HSE) Risk Tools are based on fixed shift working patterns.

Since 2019, Skeldon has worked with Scotia Gas Network (SGN) who provide gas to 5.9 million homes/businesses in the South of England and across Scotland. Skeldon used 3 years of timesheet data for their approximately 2,600 engineers to develop bespoke methods to analyse and model fatigue in call-out shift workers. SGN changed their shift patterns for engineers in November 2019 in response to her work and incorporated Skeldon’s methods into their Business Intelligence tool allowing line managers and supervisors to more accurately monitor their staff working patterns and check safety-critical issues, such as back-to-back shift working. As a result of these changes, SGN saw a 74% reduction in people working over 16 hours. Initially, for engineers and maintenance operatives only, changes in working patterns were expanded to include support and management staff. Skeldon’s work was crucial in SGN’s negotiations with the HSE regulator, and ultimately their continued licencing (a further 5-year licence was awarded to SGN by the HSE on 6/12/2020) [S9].

Skeldon’s collaboration with SGN has generated industry-wide interest. Consequently, an industry-wide working group including representatives of all UK gas networks and the HSE was established in 2019 to share data, knowledge and understanding across the sector, resulting in a sector-wide understanding of the extent of fatigue from on-call working as a health and safety issue [S9].

4. Raising public awareness

Skeldon regularly engages and involves the public in her research by working with e.g., healthcare professionals and medics who treat sleep and circadian rhythm disorders, politicians and political advisors, and teachers and educators. She regularly presents her findings at major meetings e.g., World Sleep 2017 (~2,600 participants); Society for Research in Biological Rhythms Meeting 2018 (~800 participants); and the UK Clock Club 2017 (~100 participants). A Parliamentary Office of Science and Technology’s briefing note (POSTnote) for MPs and advisors on Sleep and Health [S10] cited conclusions on school start times and adolescent sleep from [R4]. Skeldon proactively raises awareness of the impact of delayed school start times on adolescents and the move to daylight saving. For example, she has been interviewed for or had work featured in The Times (daily circulation ~368,929) and The Daily Mail (daily circulation ~1,169,241), the TES (aimed at teachers and educators), Teen Ink (a by Teens, for Teens blog with 460,000 registered users), and Cosmopolitan magazine [S10]. Furthermore, Skeldon and Dijk used their interactive sleep-wake model at the Natural History Museum’s Universities’ Week (2014), and at a Teachers and Advisers Conference (Surrey, 2014). The mathematical model that sits at the heart of the interactive model is an early version of a published one [R4]. It has been downloaded by 955 users (10/09/2020).

Symmetric Projection Attractor Reconstruction (SPAR) Improves the Quality and Relevance ...

Submitting institution: The University of Surrey
Unit of assessment: 10 - Mathematical Sciences
Summary impact type: Technological
Is this case study continued from a case study submitted in 2014?: No

Download case study PDF

1. Summary of the impact

The University of Surrey’s work on analysing the data collected by Iceni Labs non-contact respiration monitoring device SafeScan has had a direct and significant impact on Iceni Labs’ ability to bring their technology to market in a timely manner. It has allowed them to have significant first mover market advantage and they are now confident that their technology is world leading. Our analysis showed that their data had a poor signal-to-noise ratio due to the combination of chest and abdominal movement being detected. Iceni Labs redesigned their hardware so that their device could be repositioned which resulted in much cleaner signals, in which the chest and abdominal motion could be clearly distinguished, which are more suitable for analysis.

2. Underpinning research

The underpinning research consists of a new method for visualising and quantifying “approximately periodic” time series data and is known as Symmetric Projection Attractor Reconstruction (SPAR) . It was developed in the context of cardiovascular time series (e.g., continuous blood pressure, electrocardiogram (ECG)) but the general methodology is more widely applicable. The impact occurred in the context of respiration data.

The problem that the research addresses is one of recognising changes in the morphology and variability of a waveform which, in a clinical context, can aid with diagnosis of many diseases. This is a challenging problem as physiological time series data are irregular, non-stationary and noisy.

Large volumes of data can be collected over hours or days at high sampling frequency. To see the detail of individual cycles in such data, it is necessary to zoom in to a very small-time interval and even then, detecting differences with the data in another time interval is not easy. Our approach consists of plotting the data in a three-dimensional phase space and then projecting down to a particular plane which provides a simple but effective way of filtering out much of the baseline variation, which is one type of noise that is commonly found in the data (see Fig. 1). In this two-dimensional view of the data, which we refer to as an attractor, a single loop corresponds to one cycle in the data. If the data is exactly periodic, this phase space representation corresponds to a single closed loop but if the data varies significantly from cycle to cycle, then this would result in a more disperse attractor. This makes it easy to quantify variability in the data. The shape of the attractor is determined by the morphology of the signal and this representation makes changes in morphology easy to recognise as well. Since the attractor is in a bounded domain, changes in morphology and variability are easy to detect visually and algorithmically.

Figure 1: Three points track through a signal (top) are plotted in three-dimensional phase space (bottom left).

When viewed in the direction of the red line, a two-dimensional attractor is obtained (bottom right).

The attractor is generated in three-dimensional phase space by using Takens’ delay coordinates. We project the attractor onto the plane orthogonal to a particular vector that corresponds to vertical translation of the signal. By considering the frequency response functions of the variables that define the plane, we have showed that the low frequency baseline variation is filtered out in these variables. We prove that our two-dimensional attractor derived from a periodic function has threefold rotational symmetry if the time delay parameter  is chosen to be one third of the period. For approximately periodic data, we thus choose  to be one-third of the average cycle length in a window of data and this gives our attractor an approximate threefold rotational symmetry. This means that the attractor does not change shape as the cycle length in the data changes so that changes in the attractor are associated with morphology changes [R1, R2].

Heart Rate Variability (HRV) methods consider only beat-to-beat intervals and are a popular way of analysing cardiovascular data. Indeed, a search shows over 38,000 papers that refer to HRV, with over 2,500 papers published in each of the last 4 years. But by considering only the beat-to-beat intervals, all the valuable morphology information in the signal is discarded. That is why we state that “it is time to move beyond HRV and to develop a new generation of methods of analysis of physiological data that analyse all of the data contained within a particular waveform” [R1] with SPAR being one such method. We have also showed that our SPAR method can detect changes occurring in time series data that are completely missed by HRV [R3].

Features can be extracted from the attractor which quantify particular aspects. These can be combined with machine learning to classify the data. One example of this is classification of gender from human ECG signals for which the SPAR features obtained an accuracy of 96% whereas using popular ECG intervals only gave 78% [R4]. When classifying mice as either wild type or having a genetic mutation associated with sudden cardiac death from their ECG signals we obtained an accuracy of 87% using only SPAR features, which jumped to 96% when SPAR features were combined with ECG intervals and amplitudes [R5]. An alternative is to use the attractor images as input for a deep learning neural network for classification [R6].

The paper [R1] describing the foundations of the SPAR method was chosen by the journal Physiological Measurement as one of their best six papers for the Highlights of 2018 collection. It has been downloaded from the journal website over 3,500 times since publication in March 2018.

3. References to the research

[R1] PJ Aston, MI Christie, YH Huang and M Nandi. Beyond HRV: attractor reconstruction using the entire cardiovascular waveform data for novel feature extraction. Phys. Meas. 39, 024001, 2018. DOI: 10.1088/1361-6579/aaa93d

[R2] M Nandi, J Venton, PJ Aston. A novel method to quantify arterial pulse waveform morphology: attractor reconstruction for physiologists and clinicians, Phys. Meas. 39, 104008, 2018. DOI: 10.1088/1361-6579/aae46a

[R3] PJ Aston, M Nandi, MI Christie and YH Huang. Comparison of attractor reconstruction and HRV methods for analysing blood pressure data. Comp. Cardiol. 41, 437-440, 2014. ISSN 2325-8861

[R4] JV Lyle, PH Charlton, E Bonet-Luz, G Chaffey, M Christie, M Nandi and PJ Aston. Beyond HRV: Analysis of ECG signals using attractor reconstruction. Comp. Cardiol. 44, 091-096, 2017. DOI: 10.22489/CinC.2017.091-096

[R5] E Bonet-Luz, JV Lyle, CL-H Huang, Y Zhang, M Nandi, K Jeevaratnam and PJ Aston. Symmetric Projection Attractor Reconstruction analysis of murine electrocardiograms: Retrospective prediction of Scn5a+/- genetic mutation attributable to Brugada syndrome. Heart Rhythm O2 1, 368-375, 2020. DOI: 10.1016/j.hroo.2020.08.007

[R6] PJ Aston, JV Lyle, E Bonet-Luz, CL-H Huang, Y Zhang, K Jeevaratnam and M Nandi. Deep learning applied to attractor images derived from ECG signals for detection of genetic mutation. Comp. Cardiol. 46, 097, 2019. DOI: 10.22489/CinC.2019.097

Key Grants:

Continuous Information Extraction Using Reconstruction of Attractors from an ECG Signal, EPSRC Impact Acceleration Account, £20,000, August 2014 – February 2015.
Using Reconstruction of Attractors Method to Extract Diagnostic Information from Engine Sensor Data, EPSRC Impact Acceleration Account, £19,678, March 2015 – August 2015.
Using Attractor Reconstruction Methods to Classify Mouse ECG Data, EPSRC Impact Acceleration Account, £10,365, January 2017 – March 2017.
Attractor Reconstruction for Early Detection of Fever and AF, EPSRC Impact Acceleration Account, £30,386, April 2017 – November 2017.
Detection of the Presence and Progress of Covid-19 from Data, EPSRC Impact Acceleration Account, £7,000, August 2020 – September 2020.

4. Details of the impact

Professor Philip Aston led the evaluation of the respiration data obtained by SafeScan, a non-contact respiration monitoring device being developed by Iceni Labs that uses radar technology to assist medical staff in the early identification of worsening patient symptoms requiring medical intervention. By applying the Symmetric Projection Attraction Reconstruction (SPAR) method to the respiration data collected from SafeScan, Professor Aston informed the redesign of the SafeScan hardware and in doing so enabled Iceni Labs to both access new algorithmic tools and gain a significant market advantage.

Informing the development of a non-contract respiration monitoring medical device, SafeScan

Early identification of the need for medical intervention is essential as it results in improved patient outcomes and survival rates, for example, after surgery or in critical care patients. Previous research and development had focused primarily on cardiovascular monitoring. However, respiration rate is one of the most important vital signs for identifying deteriorating health but it is currently under-utilised. One study found that ‘by-eye’ measurements by medical staff were unreliable with 52% of doctors missing abnormal respiratory rates using conventional visual spot assessment methods [Philip et al., J. Clin. Monit. Comput. 29, 455-60, 2015]. Existing devices for measuring respiratory rate are cumbersome, reduce patient mobility and, due to their unreliability, are under-utilised. SafeScan is radar-based and so is non-contact, which makes it ideal for medical use. It will give medical and healthcare staff a reliable, real-time indication of respiratory rate which will improve observation accuracy and reduce nursing workload. SafeScan is able to identify the need for medical intervention before other existing technologies on the market can do so [S1].

The SafeScan device collects multi-channel data at equally spaced distances from the antenna, thus detecting any motion at each distance from the device. It is conventionally placed with the radar normal to the body either from the front or the back. In a hospital setting, it is normally placed underneath the patient’s bed. Professor Aston and his team applied their SPAR method to each relevant channel separately. The data was very noisy but by smoothing it, it was possible to generate an attractor from some channels. However, since each channel of the radar device effectively detects any motion on a spherical shell of a particular radius centred on the radar, it became clear that the signal being detected was being generated not only by chest movement (which is desirable), but also by abdominal movement (which is not desirable). This combination of effects was producing very noisy signals that were hard to analyse and were not clearly detecting the motion of interest in the chest. As a direct result of Aston’s findings, Iceni Labs redesigned the hardware of their system so that the radar can be positioned with a direct line in free space between the radar and the patient. This has meant that the radar can be positioned at the foot of the bed so that the chest motion and the abdominal motion are detected on different channels and so can be clearly differentiated. Dr Dylan Banks (Founder and Knowledge Director, Iceni Labs) reported that this redesign “allowed for us to significantly improve our signal-to-noise ratio and has opened up a range of new algorithmic processing tools that were otherwise unavailable to us” [S2]. In summary, Aston’s work has resulted in a cleaner and more relevant dataset for distinguishing characteristic change in breathing patterns associated with disease.

Significantly improved market advantage for Iceni Labs

Iceni Labs is a UK-based SME developing science and technologies for use in the medical and human security sectors. Dr Dylan Banks reports that work with Professor Aston on SafeScan “has had a direct and significant impact on our ability to bring our SafeScan radar technology to market in a timely manner. This has allowed for us to have significant first mover market advantage and we are now confident that our technology is a world leader ” [S2]. Dr Banks further reports that “Iceni Labs has recently set up a US subsidiary and in part due to their work that we have taken collaboratively with University of Surrey we have now made the decision to position SafeScan as front and centre of our US development pathway” [S2].

Thus, Professor Aston’s work has had impact both on the technological development of the SafeScan device and on the market opportunities available to Iceni Labs.

Impact case study database

Search and filter

Filter by

Advancing the Development, Public Understanding and Benefits of Global Weather F ...

1. Summary of the impact

2. Underpinning research

Data Assimilation for Weather Forecasting

Nonlinearity and Geometry in Forecasting

3. References to the research

4. Details of the impact

Developing accurate prediction algorithms

5. Sources to corroborate the impact

Influencing school start times, permanent daylight saving and shift workers: Mathematical ...

1. Summary of the impact

2. Underpinning research

3. References to the research

4. Details of the impact

1. Informing national and international legislation on school start times

2. Informing public debate on permanent daylight-saving time

3. Impact on shift working patterns

4. Raising public awareness

5. Sources to corroborate the impact

Symmetric Projection Attractor Reconstruction (SPAR) Improves the Quality and Relevance ...

1. Summary of the impact

2. Underpinning research

3. References to the research

4. Details of the impact

Informing the development of a non-contract respiration monitoring medical device, SafeScan

Significantly improved market advantage for Iceni Labs

5. Sources to corroborate the impact

Filter by higher education institution

Filter by unit of assessment

Filter by continued case study

Filter by summary impact type

Filter by impact UK location

Filter by impact global location

Filter by underpinning research subject