ORUK-VA-0005

Section 1 - Basic information about you and your application:

Title of research project
Biomechanical Insights through Markerless Motion Capture: A Fellowship in Early-Osteoarthritis Management and Machine Learning

Duration
24

Start date
01/01/2024

Profession
Academic scientist

Your current job title/position
PDRA

Have you previously received funding from Orthopaedic Research UK or Versus Arthritis
No

Please provide project ref number for previous funding from Orthopaedic Research UK


Please provide project ref number for previous funding from Versus Arthritis


Did you attend the AI in Orthopaedic Conference organised by Orthopaedic Research UK in 2022?
No

Did you attend the Microsoft AI training courses organised by Orthopaedic Research UK?
Yes

What other AI training courses have you previously attended?
Oxford Summer School: Reproducible machine learning of wearables in data science: March 2023 AI in Orthopaedics Conference 2023 - [Future event] C-STAR Course: Machine Learning and Sensors to Enhance Rehabilitation Research online course Explainable AI (XAI) for Medical Applications SINAPSE/MathWorks MATLAB Course Strathclyde Training School 2020

Are you an early-career researcher (ECR)? (definition of ECR)
yes

Section 2 - Lay summary

Lay summary:

Osteoarthritis (OA) is a widespread condition that causes joint pain and disability, affecting millions of people around the world. The disease progresses slowly, often starting with minor discomfort that many ignore until the pain becomes debilitating. Detecting OA in its earliest stages could make a significant difference in managing the disease and improving the quality of life for people. That is where this pioneering fellowship comes in, aiming to transform how we detect and treat early-stage OA.

The focus of this fellowship is the innovative use of markerless motion capture technology that can look at the way people move without requiring them to wear any special markers on their body. Traditional methods of analysing movement patterns, called gait analysis, usually involve a specialized laboratory setting where people wear markers over their body. New markerless technology allows researchers to monitor lots of people as they go about their regular activities in public spaces, without the use of markers. This approach will allow us to perform rapid assessments of the way that people move allowing us to identify subtle, often-overlooked variations in gait and movement patterns that could be early signs of osteoarthritis.

By capturing this data in real-world settings and in large numbers, we can build a more accurate and comprehensive understanding of how OA affects daily life from its earliest stages. This, in turn, can help medical professionals create more effective, personalized treatment plans. It is a potential game-changer in managing OA, a disease that until now has long been considered an inevitable part of aging.

Furthermore, the project aims to incorporate advanced machine learning algorithms. These algorithms will sift through the enormous amounts of data collected to identify the most crucial features and patterns that signify the onset of osteoarthritis. In essence, this makes the project not only a translatable health study but also a technological innovation, combining biomedical engineering, data science, and healthcare into a single, impactful initiative.

To achieve this the fellowship will not work in isolation but will be collaborative drawing on key expertise from a variety of disciplines and global collaborators. For example, additional techniques and insights will be acquired from other research groups that have successfully initiated similar programs for other health problems, expanding the impact of the project globally as well as nationally.

In summary, this fellowship aims to revolutionize early OA detection and management through the cutting-edge use of markerless motion capture and machine learning. It promises not only to advance our understanding of a complex, debilitating disease but also to offer real, tangible benefits to those who suffer from it, ultimately improving patient outcomes and reducing healthcare costs.

Section 3 - Purpose of research

Purpose of research:

The purpose of this fellowship is to revolutionize the early detection, interventions of Osteoarthritis (OA) and the decline in knee function. Utilizing machine learning algorithms gait phenotype detection and associating these with a risk of developing the disease, we aim to create a holistic understanding of an individual’s musculoskeletal (MSK) health. Filling the existing gap between early-stage OA and the current late-stage interventions, allows timely, targeted treatments that can significantly improve people’s quality of life.

Research Aims:

1. Early Diagnosis: Develop a machine learning model capable of identifying early-stage-OA through gait analysis.
2. Comprehensive Profiling: Integrate gait phenotype data with activity levels, health questionnaires, and healthcare records, thereby forming a comprehensive health profile.
3. Risk Modelling: Utilize advanced analytics for identifying individuals at a higher risk of developing OA and associated comorbidities.

Objectives:

1. Ethical Compliance: Obtain ethical approvals and informed consent from participants.
2. Data Collection: Deploy markerless motion capture in diverse public spaces to collect gait data from a target sample size of 10,000 participants.
3. Feature Extraction: Identify relevant features from the collected gait data for model training.
4. Data Integration: Combine gait phenotype data with other data sources such as activity monitors, health questionnaires, and healthcare records.
5. Pattern Recognition: Apply advanced machine learning techniques for pattern recognition and risk modelling.

Deliverables:

1. AI Model: Machine-learning model that accurately classifies gait phenotypes linked to early-stage OA.
2. Integrated Health Profiles: Comprehensive dataset combining gait data, activity levels, health questionnaire responses, and healthcare records.
3. Predictive Algorithms: Risk models forecasting the development of OA and associated comorbidities.
4. Validation Reports: Clinical evidence supporting the efficacy of the model and its integration into healthcare systems.

By achieving these aims and objectives, this research will set a new standard in early-OA detection and management, with significant implications for healthcare economics.

Section 4 - Background to investigation

Background to investigation:

Brief Literature Review:

Knee osteoarthritis (OA) is a leading cause of pain and disability worldwide¹. Its prevalence increases with age posing a substantial economic burden due to both direct medical costs (e.g., surgeries, hospitalizations, etc) and the personal burden on patients. The current state of OA management is limited by its focus on late-stage diagnosis and symptom management rather than early detection and disease modulation. Current treatment options for OA primarily focus on symptom management through pain relief, physical therapy, and intra-articular injections², culminating in joint replacement surgery as a last resort³; however, these approaches offer only temporary relief and fail to address the underlying disease progression. Research indicates that early interventions for OA⁴, particularly for early to moderate knee OA⁵, are cost-effective and could significantly reduce long-term healthcare expenses. This presents a significant gap in healthcare, underscoring the need for innovative approaches like the one proposed in this fellowship. By emphasizing early diagnosis and lifestyle modifications, we can not only enhance patient well-being but also yield economic advantages.

Before reaching the debilitating stages of osteoarthritis, there exists a transitional phase known as Pre-osteoarthritis (Pre-OA), which serves as the bridge from healthy cartilage to early osteoarthritis⁶. Common scenarios that may initiate this Pre-OA phase include cartilage damage due to meniscectomy or trauma. However, these are not the only routes to Pre-OA; the full spectrum within the population remains unidentified due to current limitations in diagnostic capabilities. Although diagnostic tools like biomarkers, MRI, and arthroscopy can identify pre-osteoarthritis (Pre-OA), their clinical use is limited due to accessibility and financial constraints. These limitations hinder the ability to detect early changes in function or biomechanical loading as early markers of human OA, even though similar signs have been observed in animal models⁶. Emerging evidence suggests that shifts in walking patterns could serve as observable markers for individuals in the Pre-OA stage⁷.

Gait and neuromuscular adaptations in people with severe knee OA have been well documented⁸. However, often due to the difficulty in participant recruitment in the pre/early stage of OA there have been far fewer studies and often in smaller numbers^7,9. During this stage of the disease there is difficulty in tracking the pathogenesis outside of animal models, with many stratifications being suggested in the literature. The ability to monitor those with early signs/changes would require large sampling across diverse populations which often requires a scalable data capture approach that can be delivered across multiple locations.

Within OA there are a number of different reported clinical phenotypes that are thought to predict progression of the disease and ultimately efficacy of interventions¹⁰. Interestingly, even healthy populations exhibit clustering into specific gait phenotypes¹¹, which may have ramifications for different tibiofemoral kinematics within varying knee compartments¹². This phenotype or personalised approach holds promise of identifying more effective ways of selecting the right treatment for individuals with OA¹³.

Biomechanics provides a framework for understanding the onset, progression, and management of knee OA¹⁴. Recent research in OA aetiology has shown promising advancements in biomechanics, with innovations in technology and machine learning (ML) enhancing its precision and applicability¹⁵. This progress highlights the potential for biomechanics to inform targeted treatments, emphasizing the need for its inclusion in large-scale studies and future clinical decision-making tools¹⁴.

Much of the biomechanical research require detailed gait and functional analysis with a wide variety of equipment¹⁶, but these studies often struggle as the variability in gait across a population require large study numbers¹⁷. Markerless motion capture techniques offer greater data generation and adaptability, with the potential for clinical use and capturing at large scales¹⁸.

Personal Track Record:

Dr. Banger’s expertise in arthroplasty, biomedical engineering, and machine learning positions him well to exploit this osteoarthritis research^19,20. His experience managing a gait lab and conducting postdoctoral research at Imperial College, University College London and with industrial partnerships offer collaborative opportunities, bridging the gap between academia and industry.

Joining Dr. Banger is a distinguished team. Prof. McGregor specializes in integrating machine learning with biomechanical models ^21,22, a perfect complement to Dr. Banger’s work. Prof. Cobb, an expert in arthroplasty, brings rich experience in surgical outcomes and research translation, strengthening the team’s clinical impact potential²³. Dr. Lindsay’s interdisciplinary background adds versatility, especially in understanding how activity patterns affect osteoarthritis progression. His experience with large population cohorts, such as the Biobank Nature Project involving over 90,000 participants, enriches our understanding of the intricate relationship between osteoarthritis and holistic health, as well as its long-term impact on morbidity and mortality^24,25. Dr. Salman, focused on primary care and public health outcomes related to healthy aging and osteoarthritis, offers valuable insights for early interventions^26,27.

Together, this team is poised to develop a platform to access to pre-osteoarthritic patients and significantly innovate in osteoarthritis diagnosis, treatment, and management.

Preliminary Data:

Dr. Banger’s network permits him access to a substantial database featuring data from ~1,000 knee patients^28,29 and healthy controls. This rich dataset will serve as an invaluable resource in the project’s initial stages for reviewing key information related to osteoarthritis (OA) and gait analysis.

The availability of this extensive dataset greatly enhances the feasibility of this proposed research. We plan to use machine learning algorithms to extract salient features from this dataset that can be monitored through markerless systems, thereby advancing our understanding of OA. Additionally, we are actively seeking partnerships with groups such as Queens University group in Canada¹⁸, Theia³⁰ and Has-Motion³¹ who are embarking on similar markerless research projects in public and clinical settings. While our work draws inspiration from their use of markerless systems for extensive data gathering, our project stands out for its unique research questions, follow-up protocols, and the additional data sets that will inform our AI model.

Section 5 - Plan of investigation

Plan of investigation:

The machine learning process for detecting gait phenotypes involves a series of steps aimed at identifying distinct patterns and characteristics within an individual’s gait. These steps are then followed by integrating this gait data with activity levels, health questionnaires, detailed testing of “at risk” profiles, and healthcare records to form a comprehensive understanding of an individual’s health profile.

1.Gait Phenotype Detection using Machine Learning:

The fellowship will collect data in a variety of public spaces, including parks, shopping malls, and transport stations, to capture a diverse range of human movements. We seek to recruit a balanced demographic population considering factors like age, gender, ethnicity, and social demographics to ensure the study’s findings are valid. This strategic choice of locations and target demographics aims to provide a comprehensive understanding of movement patterns associated with preclinical osteoarthritis.

Ethical approval will be obtained from relevant institutional review boards before data collection begins. Clearly marked markerless motion capture camera setups will be installed in selected public spaces. The purpose of the technology will be transparently communicated to ensure participants are aware they are entering an area where data collection for joint health research is taking place.

The study will be promoted via Patient and Public Involvement (PPI) engagement within the department, as well as through social media, to attract willing participants to the test sites. The goal is to enrol approximately 10,000 participants, aiming to record around 100 participants per day over 100 days. Data capture is expected to take 2 minutes.

Participants will be guided through a designated walkway area enabling capture of multiple repetitions of each participant’s gait to ensure reliability. Questionnaires and additional outcomes will be recorded at the end of the walking trials. Data will be aligned to the questionnaires through a QR code shown on the video footage. This will be generated by the research team upon completion of consent and subsequent questionnaires.

Feature extraction will focus on mining critical features from captured gait data, including but not limited to step length, cadence, and joint angles at various walking phases. These features serve as key indicators to understand unique gait characteristics related to osteoarthritis.

Model training will use machine learning algorithms like clustering or classification methods to develop a model based on the extracted gait features. The aim is for the model to recognize and sort different gait phenotypes based on the available data.

The trained model will be applied to new sets of gait data to classify individuals into distinct gait phenotypes. Statistical parametric modelling will be deployed to analyse the statistical differences between these kinematic profiles.

2.Integrating Data Sources:

The project will deploy activity monitors sent to consenting participants through the post to gather additional data points such as the intensity, frequency, duration and volume of physical activity. This data complements the gait phenotype information and provides a more holistic view of an individual’s activity patterns, contributing to a richer dataset. Due to the limited number of monitors available, data collection will initially target individuals with an “at-risk” gait phenotype which will be ascertained through the main data collection.

Participants will complete a comprehensive battery of health questionnaires encompassing metrics such as pain levels, joint mobility, previous medical history, and lifestyle factors eg diet and exercise. This data will provide contextual background to interpret gait and activity data more accurately.

Based on the data collected, individuals who are identified as having at-risk gait phenotypes will be invited for more rigorous testing at the Musculoskeletal (MSK) lab. The invitation will be extended after obtaining proper consent, aiming to carry out detailed functional assessments that could further characterize the identified gait phenotypes.

3.Data-Fusion and Analysis:

These multiple streams of data—gait phenotype data, activity monitor readings, health questionnaire responses, and MSK lab results—will be fused into a unified, comprehensive dataset. This will allow for a more robust and detailed analysis of individual’s health profiles.

Sophisticated machine learning algorithms will be employed for pattern recognition within this integrated dataset. The goal is to identify meaningful correlations and patterns between variables like gait phenotypes and activity levels, as well as potential health-related comorbidities.

With the integrated dataset, predictive models will be developed to identify individuals at higher risk of developing conditions related to or exacerbating osteoarthritis. These models will serve as decision-support tools for healthcare providers to recommend preventative measures.

4.Future-Work and Healthcare-Tracking:

In the long term, healthcare records of participants, including diagnoses, medications, and treatment plans, will be integrated into the dataset for validation purposes. This will allow for a reality check of the predictive models, enabling their fine-tuning for more accurate risk assessments. The inclusion of healthcare records will not only enrich the dataset but also add an additional layer of credibility to the research findings.

Based on the risk modelling and pattern recognition, the project aims to create personalized interventions for individuals. These could range from specific physical activity recommendations to more involved treatment plans or lifestyle changes, all aimed at mitigating identified risks.

Future iterations of the research project may include even more data types, such as biomarkers and medical imaging, gathered in collaboration with other research groups or institutions to further enrich the dataset.

The integration of research findings into clinical care will be a multistep, iterative process, across this collaborative network of surgeons,physiotherapists,GPs and engineers considering various factors like clinical workflows,training needs, and technological adaptations.

5.Dissemination:

Regular updates will be sent to VA and ORUK, to keep them informed about the research’s progress. Similarly, the group will disseminate results in reputable journals, conferences, workshops, and seminars. Individuals with knee conditions are the core focus of our study. We will organize PPIE sessions, fostering a mutual exchange of insights between patients and our research team and identifying routes for dissemination of results to communities of people interested in, and affected by, OA.

By combining gait phenotype detection with activity data, health questionnaires, and healthcare records, this comprehensive approach offers a powerful means to understand the complex relationships between gait patterns, activity, health, and potential co-morbidities.

Section 6 - Research environment and resources

Research environment and resources:

Imperial College London presents an ideal setting for research on markerless motion capture technology and its applications in healthcare, specifically in gait analysis for osteoarthritis. Hosted within the MSk Lab in the Department of Surgery and Cancer, the research will benefit from state-of-the-art motion capture suites, cutting-edge computing resources, and a dedicated team that includes two experienced technicians and a PPI engagement officer. Located at the Sir Michael Uren Hub on the rapidly expanding White City campus, the research environment encourages interdisciplinary collaboration among engineers, clinicians, and data scientists.

The site offers not just technological resources but also a rich intellectual ecosystem. Imperial College is globally renowned for its research and innovation in biomechanics, data science, and healthcare technology. It features a variety of Centres and Networks of Excellence that span fields critical to this research, including injury studies, neurotechnology, and AI. The institution’s robust data storage and management systems ensure secure handling of sensitive medical data, a crucial factor in studies involving patient information.

Direct access to the Imperial Hackspace and the Innovation Rooms on campus facilitates the innovation cycle, allowing for rapid prototyping and community engagement. Furthermore, strong links with local healthcare providers, such as the Imperial College NHS Trust and several GP practices, offer avenues for real-world data collection and validation of the study’s findings. The research group also enjoys active collaboration with a Public and Patient Involvement and Engagement (PPIE) group, enhancing the public relevance and impact of the research.

Within the campus, the research team will have opportunities for meaningful interactions with world-class researchers specializing in imaging and biomarkers. Additional support in the form of funding opportunities, grants, and fellowships available through Imperial College provides a strong financial foundation for the research.

In summary, the confluence of advanced facilities, a multidisciplinary team, high-calibre institutional partnerships, and an ethos of innovation makes Imperial College London a compelling choice for conducting this cutting-edge research. With a reputation for impactful studies that address real-world healthcare challenges, the institution offers a potent blend of resources and expertise well-suited for a project of this complexity and importance.

Section 7: Research impact

Who will benefit from this research?

This fellowship’s research on markerless motion capture and gait profile integration has broad implications that could benefit multiple stakeholders. Patients stand to gain from early detection of osteoarthritis and co-morbidities, leading to timely interventions and personalized treatment plans. Healthcare professionals will find value in improved diagnostic accuracy and the potential for tailored treatments based on individual gait patterns. The research community will benefit from new datasets and predictive models that could serve as a foundation for future work. For healthcare institutions, early detection could lead to cost savings and more efficient resource allocation. Public health organizations may utilize the research for data-driven policies and public awareness campaigns. Finally, technology developers in healthcare could leverage findings to innovate and open new market opportunities. Overall, the fellowship aims to create a multidimensional impact, advancing individual patient care, clinical practices, public health, and technological innovation in healthcare.

How can your research be translated in real-life?

This research on markerless motion capture and gait profiling aims to revolutionize the management of joint disease, akin to how blood pressure metrics transformed cardiovascular care. It has the potential to translate into clinical screening tools for early diagnosis, personalized treatment plans, and decision-support tools for clinicians. Importantly, the data collected can shape healthcare policies, specifically those focusing on arthritis and preventive care. The project aspires to create an ecosystem of care where data-driven, timely interventions are made possible. By collaborating with healthcare professionals, technologists, this research seeks to bring about holistic changes in patient care, healthcare systems, and public health strategies. The multi-disciplinary approach ensures the research has broad and actionable real-world implications.

How will your research be beneficial for Orthopaedic Research UK, Versus Arthritis and their purpose?

Revolutionizes management of common musculoskeletal diseases through:

Advance knowledge and understanding by

• Providing unique insights into the biomechanics of gait and health
• Evolve early markers of joint ill-health
• Create new care paradigms to manage joint health.

Innovative and timely

• Facilitate early identification risk and subsequent tailored management
• Develop technologies scalable to translate to primary and secondary care

Clinically impactful and patient centred

• Facilitate personalisation of care and management
• Potential to improve outcomes
• Raise public awareness and engagement with joint health
• Empower patients to take control of their joint health

Collaborative

• Our research collaborators and patient community are broad and collaborative and have the skills and knowledge to support this ambitious work
• Interdisciplinary clinical colleagues in primary and secondary care to translate the findings and make them impactful
• Local partners including our councils to facilitate data collection

Section 8: Outreach and engagement

Outreach and engagement

The fellowship aims to pioneer a paradigm shift in how both clinicians and the general public perceive and manage osteoarthritis (OA). While traditionally viewed as an unavoidable outcome of ‘wear and tear’ or aging, this research endeavours to position OA as a condition that can be effectively managed through early intervention strategies but requires a means of profiling large numbers of people into at risk groups. Our outreach and engagement initiatives are critical levers to instigate this change.
In collaboration with our Patient and Public Involvement and Engagement (PPIE) group, we will roll out a comprehensive dissemination strategy designed to reach a broad and diverse demographic, covering various social and ethnic backgrounds. We will harness a mix of traditional and digital communication platforms to get our message across. Currently the group is 50 patients with a history of osteoarthritis, but we will look to expand this to capture the views of those with early joint health problems. An interactive website will act as the central hub for all project-related information, updates, and resource materials. We will employ social media to maintain a constant dialogue with our community, sharing real-time advancements, taking feedback, and answering queries.
In focusing on the medical community, we understand the importance of buy-in from healthcare providers for sustainable change. To this end, our co-investigators will actively engage with the primary healthcare, physiotherapy, and surgical sectors. Their roles will be pivotal in exploring how to integrate this research findings and joint health metrics into existing medical practices. Collaborative endeavours will extend to working with key organizations such as Versus Arthritis and ORUK to solidify our standing and reach within the healthcare community. These partnerships will serve to enrich our research while simultaneously facilitating its translation into everyday clinical practice.
Public engagement events form another cornerstone of our outreach efforts. These events alongside data collection will not be mere information dissemination sessions but interactive platforms that allow for two-way communication. Attendees will not only learn about the importance of joint health and the role of lifestyle choices in preventing OA but will also have the opportunity to share their own experiences and challenges, offering real-world insights that can help refine our research and intervention models.
This outreach and engagement plan is a multi-pronged strategy aimed at shifting societal and medical perspectives on OA. By employing a blend of digital and physical platforms to engage with a diverse audience, fostering academic projects to instil new practices among healthcare providers, and creating collaborative partnerships with key organizations, we aspire to make a lasting impact on the management of osteoarthritis.

Section 9: Research budget

Requested funding from Orthopaedic Research UK and Versus Arthritis

University fees (if any)
£0

Salary
£69896.34

Consumables
£30000

Publications
£0

Conference attendance
£0

Other items
£

Total 'requested fund'
£99896.34



Other items




Other secured funds

Internal funding
£0

Partner (University)
£0

Partner (Commercial)
£0

Partner (Charity)
£0

Other sources
£0

Total 'other funds'
£0



List all the 'other sources' and explain how their funds are used to cover the costs of your research.

Section 10: Intellectual property and testing on animal

Is there an IP linked to this research?
No

Who owns and maintains this patent?


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?