FP-00031

Section 1 - Basic information about you and your application:

Title of research project
Development of Artificial Intelligence to diagnose ligament health and injury potential.

Grant Type
The ORUK Inspiration Fund

Research area
Diagnostic

Duration
12

Start date
March 4, 2024

Have you previously received funding from ORUK?
No.

Profession
Academic scientist

Your current job title/position
Senior Lecturer

Are you an early-career researcher (ECR)? (definition of ECR)
no

Section 2 - Lay summary

Lay summary:

Ankle sprain is one of the most common injuries across the general population and particularly in sport. Globally, an ankle sprain accounts for 5-10% of all accident and emergency visits per year. This equates to approximately 2 million reported incidences in the UK alone and has been estimated to cost the economy more than £1bn. Over 50% of first-time sprains involve chronic symptoms, including pain and instability, leading to a high risk of re-injury and development of longer-term ankle instability. In most sprain cases, the outer side of the ankle is affected usually by a tear in one particular ligament – the anterior-talofibular ligament (ATFL).

Ankle ligaments are crucial for daily activities such as standing, walking and running because their main function is to stabilise the ankle joint. With increasing load, the ligament structure gets longer and stiffer, but will continue to function and absorb energy until loading reaches the tissue’s elastic limit. After this limit, strain in the ligament increases at the expense of stiffness and the likelihood of tissue damage is heightened. Expensive diagnostic technologies such as magnetic resonance imaging and computed tomography can diagnose ligament injury and give some indication of the tissue’s health during rehabilitation. However, both techniques fall short as they cannot quantify a ligament’s capacity to withstand physical stress (e.g., mechanical loading), which is an important consideration for injury prevention and effective rehabilitation.

A tissue’s tolerance to stress can be assessed by measuring its stiffness. An increasingly popular technique being used in musculoskeletal research for estimating tissue stiffness is shear-wave elastography via ultrasound imaging. Ultrasound beams generate tiny forces inside the body, which spread sideways to the beam and displace tissue. The shear-wave elastography technique detects the velocities (speed and direction) of the resulting displacements and tissue elasticity can be measured. For example, if a ligament has a low shear-wave velocity, then the tissue is lax. This is a threat for injury.

The ambition of this research project is to provide the Orthopaedic community with a cost-effective and practical solution to predict and manage ankle sprain injury. The longer-term aim is to reduce the incidence rate. The proposed project will recruit a broad cross-section of the population differing in age, gender, ethnicity, health and physical activity. It will specifically recruit members of the public and athletes suffering from current or previous ankle sprain injury to participate in the study and assist in disseminating the study’s aims and its findings; as well as advising future research. Participants will have their ATFL ligaments imaged, and the measures of stiffness will be entered into a database. We will then use artificial intelligence to produce an innovative diagnostic tool for the prevention and management of ankle sprain injury.

Section 3 - Purpose of research

Purpose of research:

This project will be part of on-going PhD research. The overall aim is to deliver a ligament screening test that objectively measures the capacity of ATFL to tolerate mechanical loading, which can be used in practice for acute and chronic assessment of the ligament. To achieve this, a database of ATFL stiffness values will be created and subsequently managed using artificial intelligence methods to provide a score on ligament health and injury potential.

The project will have three deliverables as the result of following work packages:

• W-P 1: Data collection – the objective is to collect data on ATFL health representing a broad cross-section of the population. This will be an ongoing work package throughout the length of the project, where participants will visit the Musculoskeletal Biomechanics laboratory at LSBU and undergo an experimental protocol (section 5) that images the ATFL of both extremities with shear-wave ultrasound elastography. The deliverable will be a minimum (minimum n=300) number of ATFL stiffness records indexed according to demographics, health and physical activity status and entered to a normative database.
• W-P 2: Building of database – the objective is to create a population database of ATFL health. Each measure of ligament stiffness will be entered into the database, and cumulatively, used for the identification of ‘norms’ and ‘outliers’ for a given ATFL based on demographics, health and physical activity status. The deliverable will be a relational database system created in Postgre SQL.
• W-P 3: Programming of artificial intelligence – the objective is to develop a learning algorithm that returns a visual score on ligament health. The algorithm will be written in Python and the deliverable will be a software that acts as a front end in accepting location specific ATFL stiffnesses, cross-references a normative database, and returns an individual’s potential for injury.

Section 4 - Background to investigation

Background to investigation:

The prevalence rate of ankle sprain injury continues to rise [1] and still, a comprehensive understanding of the risk factors underlying the injury is not clear [2], particularly factors associated with joint anatomy [2]. In most ankle sprain cases, the ATFL is the affected ligament, and this is because it has the poorest strain behaviour of all the lateral ankle ligaments [3].

There is sufficient evidence that an injured Achilles tendon can be reflected by its mechanical properties (e.g., its stiffness) [4,5,6,7,8]; therefore, the same should be true for the ATFL, given that tendon and ligament share a similar structure [9]. However, the typical approach for measuring Achilles tendon stiffness (via synchronized ultrasound and dynamometry) is problematic for ligament because load and displacement are difficult to measure. An alternative approach is shear-wave elastography; this is an emerging technology in musculoskeletal research, but not without its limitations. Primarily, these are artefactual challenges that influence the speed of the shear-wave such as tissue heterogeneity and depth, and underlying osseous structures [10]; although, these are more of a concern for skeletal muscle and deeper tissue elastography than shear-wave investigations of ATFL. Indeed, the ATFL is a passive and superficial tissue and correct positioning of the transducer can overcome the reflection of bone within an elastogram [10].

Notwithstanding these limitations, shear-wave elastography has demonstrated that the stiffness of the ulna collateral ligament in the throwing arm of collegiate baseball pitchers significantly reduces during the season compared to pre-season (p=0.001) [11,12] and non-throwing arm (p<0.001) measures [11]. This is indicative of ligament laxity and what should be a flag for therapeutic intervention. Similar observations have been shown in collegiate football players [12]. The technique has also been shown to be sensitive to changes in ankle position with shear-wave velocities being significantly greater (p<0.001) in ATFL when plantarflexed and under greater strain than a relaxed position [13]. In addition, inter- [14] and intra-observer [15] reliability of ATFL measurements has been demonstrated, but construct validity is still missing from the literature. In the tendon, there is equivocal opinion on this issue [16,17].

In summary, shear-wave elastography is sensitive enough to detect changes in the stiffness of ligament with training and competition, or when placed under different loads; and its output – either shear-wave velocity or shear modulus, is reliable and repeatable.

The investigatory team are internationally recognized for research on muscle-tendon biomechanics spanning over 20 years and including over 30 publications in high-ranking journals, invited keynote lectures, and knowledge transfer / income generating activities with private healthcare companies, Olympic organizations and professional football clubs. We are also internationally recognized for work in super-resolution ultrasound imaging techniques, sports injuries and foot and ankle biomechanics. Combined with a senior UK foot and ankle consultant orthopaedic surgeon, the investigatory team has the expertise and track record to deliver the proposed work packages and add significant knowledge to the area of ankle sprain injury.

Dr Darren James (PI): is a Senior Lecturer in Biomechanics. He has led external research and development consultancy projects for the footwear industry and has supervised PhD projects in the areas of clinical gait analysis and foot and ankle biomechanics. He has led on two projects funded by the British Orthopaedic Foot and Ankle Society and in total, has secured over £500k (as PI or Co-I) in research income, capital investment funding and enterprise activity. Darren has co-authored 22 scientific papers within the Biomechanics area and one book chapter on ankle sprain injury.

Miss. Fenja Deister (Co-I, doctoral researcher): completed her BSc in Biomedical Engineering at Hochschule Mannheim, MSc in Biomedical Engineering at University of Stuttgart, and in January 2023 accepted a fully funded PhD scholarship at LSBU to investigate the mechanical properties of ligaments.

Dr. Sevan Harput (Co-I, project adviser): is an Associate Professor and head of South Bank Applied BioEngineering Research (SABER). He has secured over £1.2m in research income as a PI or Co-I and has given 10 invited talks on ultrasound imaging, including the first clinical super-resolution ultrasound imaging study. He has authored 27 peer-reviewed journal articles; is an Associate Editor for IEEE T-UFFC and sits on the Royal Society panel for the International Exchanges scheme.

Prof. Kiros Karamanidis (Co-I, project supervisor): is a Professor of Ageing and Exercise Science and is head of the Musculoskeletal Biomechanics Research group at LSBU. He is an author and co-author of more than 100 peer reviewed journal publications and book chapters and has secured over €4.3m, as a PI or Co-I, in research grant income and enterprise projects.

Mr. Matthew Solan (Co-I, project recruitment and translation outside academia): is a Consultant Orthopaedic Surgeon at Royal Surrey County Hospital. He is past chairman of the BOFAS Scientific Committee and has co-authored over 100 peer-reviewed journal articles predominantly on foot and ankle surgery.

Prof. Wolfgang Potthast (Co-I, project collaboration in sport injury research): is a Professor of Clinical Biomechanics within the Institute of Orthopaedics and Biomechanics at the German Sports Institute, Cologne. He is currently president of the German Society of Biomechanics and has worked on several funded football-related projects with different research institutions (e.g., FIFA) and industry partners.

 

Preliminary work so far has revealed that B-mode ultrasound imaging of ATFL is feasible and reliable. Intra-observer reliability was assessed from multiple measurements within a day (n=3) and between days (n=5) in one healthy female (white American) participant aged 21 years (Fig. 1). ATFL length (mean ± SD: 25.4 ± 2.4mm), thickness (1.8 ± 0.2mm) and 2D area (29.6 ± 4.7 mm²) was shown to be reliable (ICC_(3,1): 0.997, 0.983 and 0.967, respectively) and in line with reported values in the literature.

Our preliminary data shows that ligament shear-wave velocity measurements are feasible with measurement errors of 0.05, 0.12 and 0.07 m/s for mid-section, distal and proximal regions of the ATFL (Fig. 2), respectively. The preliminary investigations have emphasised that a standardized experimental procedure is essential for repeatability of measurement, and we will proceed in future with an ankle fixation device as described in Figure 3.

Section 5 - Plan of investigation

Plan of investigation:

Work package 1 (W-P 1): Data collection.

A cross-section of the population (minimum n=300; [18]) will be recruited for this study to reflect varying functional capacities of the ATFL, including injured and previously injured ATFLs, and injuries to other areas of the lower extremity. Recruitment methods will include internal (LSBU) and external social media advertising, student engagement and word of mouth advertising around the University campus, our physiotherapy and industry partner networks, and from the private clinics of Mr Solan. Participant incentive has been included within the costing for the project. Exclusion criteria will be severe injuries, pathologies and any other condition that would prevent participation in the experimental procedures. Each participant will complete a demographic and health questionnaire and provide written informed consent to participate in the study.

The experimental procedures will require the participant to lie on their side, with the spine, hip and knee in extension, on a physio plinth. At the end of the plinth, a custom-made ankle arthrometer (Fig. 3) will be fixated to induce and control passive ankle inversion movement at pre-defined and standardized loads. Each ankle will be positioned and secured to the device in neutral position and its joint centre aligned, in the coronal plane, to the axis of rotation of the arthrometer. An electro-goniometer, embedded within this axis, will be used to identify four ankle inversion positions of 0°, 5°, 10° and 20° with which the ankle will be passively displaced into by the addition of increasing load onto the tilting platform. In doing so, the ATFL will be subjected to increased tensile strain and its stiffness will be measured via B-mode ultrasound using a 50 mm linear array transducer imaging at a depth of 15 mm. The ligament will be measured in its longitudinal axis with the probe positioned at the antero-distal aspect of the lateral malleolus and parallel to the plantar surface of the foot.

Once ATFL has been located, shear-wave elastography will be activated within the ultrasound system (an elastogram superimposed over the B-mode image) and three regions of interest (ROI) will be manually positioned at the proximal, middle and distal aspects of the ligament (Fig. 2). The shear-wave stiffness will be recorded at each ROI, saved and the ankle moved into the next position of inversion.

To ensure that the angular displacement of the ankle is performed passively and without antagonist muscle activity, surface electromyography (EMG) will be collected from the antagonist muscle to ankle inversion – m. peroneus longus. A bi-polar surface EMG sensor will be attached the skin surface overlying the belly of the muscle once shaved and cleaned to mitigate skin resistance. The signal will be amplified x1000 at the source, sampled at 1000 Hz and band-pass filtered between 10-400 Hz, then digitally converted and continuously recorded throughout the experimental procedures.

This application requests funding to support a Research Assistant (~0.5 FTE) to work alongside the doctoral researcher (Fenja Deister) to ensure this work package is delivered.

Work Package 2 (W-P 2): Building of database.

A relational database system will be created in Postgre SQL to index all measures of ATFL. The database will store normative ATFL stiffness measures as a function of its location and tensile strain; therefore 14 x elastograms will be recorded from all participants and entered into the database: a proximal, middle and distal ATFL stiffness at four ankle joint angles (0°, 5°, 10°, 20° inversion) from each limb. The elastograms will also be made available to the broader research community for qualitative analysis and further development via the LSBU research repository. Each stiffness measure will be indexed with respect to participant demographic, health and physical activity status.

Work Package 3 (W-P 3): Programming of artificial intelligence.

Once work package 2 has been completed, a user interface (in Python) will be developed as a front end to read the database with a search function capability. At the back end, regression-based machine learning algorithms will be used to process the labelled ATFL elastograms to generate a score. Here, we will combine two methods: (1) Linear Regression – a supervised machine-learning algorithm will be used and trained by the labelled elastograms, combining the location (ROIs) and tensile strain (inversion angle) to map to the most optimized linear function; (2) Decision Trees – a non-parametric supervised learning method used for classification and regression, will be used to create a model that predicts the score of the ligament health based on the inputs from our database of stiffness with respect to age, gender, health, medication, current/previous injury, physical activity status and nutrition and supplementation. Although, these are the most effective machine learning algorithms for the proposed project, in the case of poor accuracy, sensitivity, or specificity we will employ different methods such as logistic or polynomial regression for (1) and naïve bayes classifier for (2).

It is envisaged that a number of tests of functional capacity will be needed, and this has been made clear in the project workflow diagram (Fig. 4). This application requests funding to employ a senior software engineer (~0.2 FTE) to work alongside the doctoral researcher (Fenja Deister) to ensure this work package is delivered.

Whilst there are barriers to consider that will initially prevent the widespread clinical adoption of the ligament screening test (e.g., practitioner access to ultrasound systems, variability between different systems and implementation of the experimental procedures underpinning the database), there are a couple of immediate opportunities following delivery of the work packages. Firstly, we will include the ligament screening test within our existing portfolio of scientific methods that are utilized in practice by stakeholders of the Sport and Exercise Science Research Centre (SESRC); and secondly, look to translate the test outside of LSBU and into the private practices of Mr Solan. Longer-term, the barriers to widespread adoption will need to be addressed with support from orthopaedic funders.

Section 6 - Research environment and resources

Research environment and resources:

The project will be managed by the Musculoskeletal Biomechanics Research Group in collaboration with South Bank Applied BioEngineering Research (SABER). The former is part of the SESRC, which is internationally renowned for scientific excellence and for leading expertise in the following areas of emerging significance: musculoskeletal biomechanics, environmental psychophysiology and sports coaching. SESRC submitted to the recent 2021 Research Excellence Framework (REF21) and returned an overall grade point average of 2.93, ranking 31^st out of 61 Higher Education providers who submitted to unit of assessment 24 (note: 12 staff, ~100% / 10.6 FTE, were submitted).

Aside from the REF 21 sub-panel committee noting 75% of all research outputs as being ‘world-leading’, and ‘very considerable impact in terms of reach and significance in relation to health behaviour change and injury prevention’, they commented on a research environment conducive to ‘vitality and sustainability’, specifically noting that the ‘continued investment in infrastructure and facilities supported a clear research strategy’.

This project will draw upon the SESRC environment. It will constitute two of four objectives of on-going, fully funded (by SESRC), PhD research being conducted in one of our three Musculoskeletal Biomechanics laboratories (J-207). The laboratory houses the essential equipment and software required for successful delivery of the project, including brand new ultrasound imaging hardware (LOGIQ^TM P9 XDclear) with enabled shear-wave elastography. This has recently been invested in by LSBU’s Research Capital Investment Fund as part of a collaborative bid put forward by SESRC and SABER.

The doctoral researcher will be well-supported throughout the project. Technical support for shear-wave elastography will be available from the project advisor (Dr. Harput). Technical support for B-mode ultrasound imaging will be available from all members of the team who have varying experience of the technique in research and practice. The project requests funding to support the large collection of data by way of a part-time research assistant (~£22.7k), and specialist technical support (~12.3k) for the doctoral researcher in software engineering. In addition, there will be directly incurred operating expenditure to support and disseminate the project and its output. Specifically, £5k is requested for a reserve ultrasound probe and £3k on manufacturing costs to replicate the experimental apparatus in Fig. 3; £2k on patient / participant incentives and finally £5k on article processing charges, conference fees and attendance, and subsistence.

Aside from the practical skills developed during the project, the doctoral researcher will develop the essential professional skills expected of an early career researcher, including (but not limited to): project management and communication skills, grant writing and working with funding bodies, and reporting and dissemination skills. The project will also provide the opportunity for travel to a collaborating laboratory in Germany to scaffold existing, and develop, new practical skills.

Section 7: Research impact

Who will benefit from this research?

Ankle sprain injury occurs through many factors and can be sustained either through acute or chronic loading of the ankle ligaments. It is impossible to prevent the injury, but to have a diagnostic tool at hand that can give some insight into the ATFL’s capacity to tolerate mechanical loading should be effective to at least lower the incidence rate. Beneficiaries of this research will be wide-ranging, extending up to the National Health Service and the Economy in general, given that ankle sprain injury represents an economic burden with respect to medical costs and time away from work.

How can your research be translated in real-life?

Post-project, our pathway to real-world impact will start with research outputs. Then, we will work with SESRC stakeholders in elite sport and private practice and undertake longitudinal monitoring of athletes, which we will disseminate at conferences and specialist meetings. We then hope to collaborate with industry and further develop the screening test for robustness and make it, and the experimental apparatus, more widely available.

The research team has a track record of translating laboratory methods and techniques into practice. TemuLab® is a mobile diagnostic and training device that identifies an athlete’s susceptibility to over-use tendon injury. This was developed by Prof. Karamanidis and colleagues at the German Sports University and multiple devices are operational across Europe including German Olympic training centres and professional football clubs in the UK. TemuLab® was rated as a 3* impact case study by REF21 for ‘reach and significance’ in relation to injury prevention.

How will your research be beneficial for ORUK and its purpose?

The project aligns with strategic objectives 1, 3 and 7 of the ORUK 2020-23 strategy. It will deliver an innovative diagnostic tool for the prevention and management of ankle sprain injury, which is embedded in the surrounding ecosystem of orthopaedic well-being.

Section 8: Outreach and engagement

The Musculoskeletal Biomechanics Research group led by Prof. Karamanidis has stakeholders in elite sport, sports medicine and industry. Prof. Karamanidis regularly communicates academic findings and new knowledge to these stakeholders and keeps them abreast of new developments within our labs. This project will feature in discussions between Prof. Karamanidis and his network.

We will communicate the project’s findings to the academic sport medical community principally by research outputs and conference attendance. The team have professional memberships with the British Association of Sport and Exercise Sciences and/or the European College of Sport Sciences and German Society of Biomechanics, and we have previously presented at the International Ankle Symposium and Isokinetic Sport Rehabilitation and Traumatology Conference; so, we would expect to target the main beneficiaries of our research at these events. We will communicate to clinical practitioners, who may not necessarily be academics, by way of our strong connection to the British Orthopaedic Foot and Ankle Society and its annual conference as well as other relevant practitioner conferences in the areas of Sport Therapy and Podiatry.

We will embrace social network platforms to additionally communicate the ligament screening test. Specifically, a YouTube channel will be created to introduce the concept and serve as a repository for educational videos on data collection and experimental considerations. In addition, an Instagram channel will provide regular updates on the project throughout the term.

This year Dr Harput and Miss Deister participated in the Royal Society’s 2023 Summer Science Exhibition where they showcased ultrafast ultrasound imaging and its application to musculoskeletal research – identifying members of the public who possessed a Fabella – a rare sesamoid bone in humans. Dr. Harput has previously been invited to talk at STEM for Britain, an event designed to give members of Parliament an insight into UK University research. We will look to engage in these events again as part of our dissemination of the project.

Section 9: Research budget

Requested funding from ORUK

University fees (if any)
£0

Salary
£34935

Consumables
£10000

Publications
£2000

Conference attendance
£3000

Other items
£

Total 'requested fund'
£49935



Other items




Other secured funds

Internal funding
£0

Partner (University)
£58822

Partner (Commercial)
£0

Partner (Charity)
£0

Other sources
£0

Total 'other funds)
£58822

Section 10: Intellectual property and testing on animal

Is there an IP linked to this research?
Yes

Who owns and maintains this patent?

The research will develop a novel and transformative solution to assess ligament health and injury prediction. Once validated, LSBU’s technology transfer office will work with ORUK to protect the technical idea underpinning how the algorithm works.

Does your research include procedures to be carried out on animals in the UK under the Animals (Scientific Procedures) Act?
No

If yes, have the following necessary approvals been given by:

The Home office(in relation to personal, project and establishment licences)?


Animal Welfare and Ethical Review Body?


Does your research involve the use of animals or animal tissue outside the UK?
No

Does the proposed research involve a protected species? (If yes, state which)



Does the proposed research involve genetically modified animals?


Include details of sample size calculations and statistical advice sought. Please use the ARRIVE guidelines when designing and describing your experiments.


There should be sufficient information to allow for a robust review of any applications involving animals. Further guidance is available from the National Centre for the Replacement, Refinement and Reduction of Animals in Research (NC3Rs), including an online experimental design assistant to guide researchers through the design of animal experiments.

Please provide details of any moderate or severe procedures


Why is animal use necessary, are there any other possible approaches?


Why is the species/model to be used the most appropriate?