AI and Healthy Ageing PhD: What Can Everyday Movement Reveal About Mood and Fatigue?

AI and Healthy Ageing PhD: What Can Everyday Movement Reveal About Mood and Fatigue?

Quick Decision

This Exeter DLA project asks whether movement features from existing two-dimensional functional videos can provide useful information about depression, anxiety and fatigue in adults aged 60 and over. Its dataset contains more than 1,000 assessments across healthy participants and people with several long-term conditions; that is a count of assessments, not necessarily unique people. The planned sequence is pose estimation, interpretable statistical baselines, nonlinear movement features and explainable machine learning. Official project PDF

The central analytical risk is confounding: slower or asymmetric movement may reflect pain, medication, diagnosis, cognition or physical function rather than mood or fatigue. The project does not identify the mood/fatigue instruments, sample counts by group, repeated-measure structure, model hierarchy, participant-level validation split or external test. These details determine whether a model finds a robust signal beyond clinical and demographic information, and whether it generalises across impairments. The project PDF also leaves the doctoral registration title incomplete, so confirm the programme route and location.

The award offers up to 3.5 years of funding, tuition and £5,000 total RTSG. Official pages disagree on the application deadline (8 or 12 December 2026), the referee date (8 January or 15 February 2027), and whether £21,805 is the current 2026/27 rate or a 2027/28 amount. Plan around 8 December at noon and 8 January for references while requesting written confirmation.

Research Problem and Doctoral Work

What is unresolved and why it matters

Depression, anxiety and fatigue can accompany multiple long-term conditions and affect mobility, rehabilitation and independence. But self-report scores, clinical interviews and movement limitations do not measure the same thing. For an older adult, slower or asymmetric movement might reflect mood, pain, medication, neurological disease, frailty, cognition, or several factors together. A movement classifier must therefore show that it captures information about mood or fatigue beyond what can already be inferred from diagnosis, age or physical function. The project argues that infrequent clinical encounters may miss changes between visits; that is a rationale for research, not evidence that a camera-based tool can already monitor patients at home.

The project describes an existing dataset with more than 1,000 functional movement assessments from adults aged 60 and older, including healthy adults and people with stroke, Parkinson’s disease, frailty, arthritis or mild cognitive impairment. Associated information includes demographics, clinical characteristics, functional outcomes, mood assessments and fatigue scales. The description does not name the instruments, counts by group, whether participants contributed repeated assessments, or how symptoms were confirmed clinically.

Advertised work and plausible research questions

The project proposes to apply OpenPose, MediaPipe and MoveNet to 2D videos of posture, balance, transfers, walking and activities of daily living. Named features include spatiotemporal characteristics, joint kinematics, movement smoothness, symmetry, postural alignment and inter-limb coordination. Traditional statistical methods will establish interpretable baseline relationships with depression, anxiety and fatigue measures. Entropy, fractal dynamics and recurrence-based metrics will then examine movement structure that conventional biomechanical summaries might miss. The next stage is a hierarchical AI framework using explainable Random Forest, XGBoost and LightGBM models.

This is a coherent methodological progression, but the PDF ends after listing those model families. It does not specify the framework’s hierarchy, target definitions, feature-selection procedure, sample balance, training/test design, calibration, external validation or performance criteria. Those are open design questions, not details to assume.

The analytical challenge is separating psychological signal from physical impairment. A 2023 cohort of 72,359 people associated wrist-sensor gait and activity features with incident depression; a 2024 feasibility study examined in-home movement patterns in older adults against Geriatric Depression Scale outcomes. Both make the question plausible, but they use different sensors and settings from this project’s functional 2D videos and do not validate its proposed models. A 2024 review of depression biomarkers found substantial variation in outcome reporting and feature engineering, limiting conclusions about individual features. A 2025 fatigue review found gait measures sparse and low-certainty, with limited longitudinal, cross-condition and diagnostic-performance evidence. These findings support a validation-first approach, not a claim that movement is already an accepted mental-health biomarker.

Possible evidence-led angles—rather than promised project tasks—include testing whether movement features add information beyond age, diagnosis, demographics and standard functional measures; checking performance across conditions and mobility levels; and testing robustness to camera viewpoint, occlusion and incomplete movement. Even an explainable model would initially produce a candidate association or marker. Explainability alone does not establish diagnosis or clinical usefulness.

School, Programme and Field Context

The degree-awarding institution is the University of Exeter. The project PDF places the work in Public Health and Sport Sciences at the Medical School. Exeter’s public health research-degree page offers an MPhil/PhD Public Health route and describes an interdisciplinary department spanning epidemiology, psychology, sociology, medicine and nursing. However, the project PDF’s “Potential PhD programme of study” field stops at “PhD in computer Science or”. The award page describes Public Health and Sport Sciences studentships, while the team also includes Computer Science academics. Confirm the registration programme, campus and any cross-departmental training arrangement before applying.

Exeter’s REF 2021 results show relevant breadth: 35% of its Public Health, Health Services and Primary Care submission was rated 4-star and 54% 3-star; corresponding profiles were 42%/50% for Allied Health Professions, Dentistry, Nursing and Pharmacy, 53%/39% for Psychology, Psychiatry and Neuroscience, and 33%/50% for Computer Science and Informatics. These are subject-level institutional profiles, not league-table rankings or a project-level quality score. Exeter sets a four-year maximum for a full-time PhD, while this award funds up to 3.5 years; ask how the funded submission date fits the registration timetable. Relevant broader context includes Exeter’s AI for Humans and Health theme and Exeter Digital Health network, but membership does not guarantee access to specific facilities or placements.

Supervisor and Lab Assessment

Professor Helen Dawes is lead supervisor. Her public profile describes a clinical rehabilitation and movement-science background and work on technology, explainable AI and translation with clinicians, industry and the public. She leads rehabilitation themes at the NIHR Exeter BRC. Dr Pavlos Evangelidis brings exercise physiology, biomechanics and digital-health work with older adults and people with long-term conditions. Professor Chris Fox’s profile focuses on dementia, healthy ageing and technology-supported trials. Professor Lu Liu works on trustworthy AI for healthcare, including cognitive decline and dementia risk. Professor Hang Dong is also named in the project PDF, but his specific role in this project is not clear from the reviewed material. This breadth could support the project if supervision responsibilities are coordinated; the PDF does not specify the day-to-day division of work.

Recent Exeter-associated work helps show the surrounding research activity:

• A 2026 conference abstract by Mansoubi and Dawes describes markerless computer-vision assessment of walking and joint range of motion in Parkinson’s disease, using clinical measures for comparison. It is a close local methodological example, but an abstract rather than a complete validation paper.
• A 2024 intensive longitudinal cohort study linked exertion across daily activities with immediate and some delayed symptom fluctuations in long COVID. It reinforces that fatigue and mood-related symptoms are time- and context-sensitive; it does not validate video biomarkers in older adults.
• A 2024 qualitative study explored how people with acquired brain injury and occupational therapists view digital fatigue monitoring. Its practical lesson is that outputs must be interpretable to patients and clinicians; it is not an automated classification trial.
• The 2024 Knee4Life protocol proposes smartphone-camera computer vision for movement assessment after knee replacement and plans comparison with established measures. It is a protocol, not an efficacy result, and does not establish methods or resources for this PhD.

The NIHR Exeter rehabilitation theme describes the wider VSimulators environment, including force plates, motion capture, virtual reality and biomedical sensing. The project itself is based on existing videos and does not say these facilities will be used. Ask which data, compute and clinical resources are actually available and how the named supervisors will divide clinical, statistical and computer-vision guidance.

Funding, Application, and Living Cost

This project is listed within Exeter’s 2027/28 EPSRC Doctoral Landscape Award. Official central guidance describes up to 3.5 years of full-time funding: university tuition, an index-linked stipend currently set at £21,805 per year for 2026/27, and a £5,000 total Research Training Support Grant (RTSG). Successful international award holders are stated to receive full tuition coverage through a university fee waiver, but international places are limited to up to 30% of the cohort. Visa, Immigration Health Surcharge and moving costs are excluded. The 6008 award summary labels £21,805 as the 2027/28 value, while its body and central DLA page call it the current 2026/27 rate. Ask which rate applies in the offer.

Plan to submit by noon on 8 December 2026. The central DLA page and award-page summary give that date, while the body of the 6008 page says 12 December. References are also inconsistent: the award page says 8 January 2027; central information says 15 February 2027. Use 8 January as the safe date pending confirmation. The award page schedules interviews for 15–24 February 2027, but the central page includes stale wording for February/March 2026. The central page says offers are expected in early March 2027.

Exeter estimates basic living costs of £1,300–£2,000 per month, based on local spending in April 2026. Dividing £21,805 by 12 gives about £1,817 per month. That sits within the published range but below its upper estimate; visa, health-surcharge, travel and initial moving costs are separate for international applicants. Individual housing choices matter. The university’s general PhD maximum is four years and may exceed the funded period, so ask how the 3.5-year award fits registration and thesis-submission requirements.

The general DLA application asks for a project-specific personal statement, CV, transcripts, two references and, where relevant, English-proficiency evidence. Documents must be in English, with certified translations of academic qualifications. Applicants may apply to at most three projects, with a separate application for each, and should identify their preferred project. The project PDF includes an “Apply Here” instruction, but the route is not readable in the extracted PDF text.

Applicant Fit and Preparation

The project-specific PDF says applicants should have Computer Science, mathematics, statistics or associated training “at some level.” It does not specify a degree title, grade threshold, coding language or clinical qualification. General EPSRC DLA rules call for a First or Upper Second UK honours degree, or overseas equivalent, in an appropriate science, engineering or technology area. Applicants with a Lower Second may be considered if they also have a Master’s degree or significant relevant non-academic experience. Confirm how the project-specific line and general funding eligibility interact; do not assume that a Computer Science degree is mandatory.

The work should suit applicants who combine quantitative analysis with careful interpretation of health outcomes. Evidence of Python, R or MATLAB, statistical modelling, computer vision or time-series analysis, reproducible machine learning and clinical/functional outcome interpretation would be useful preparation, but only the broad training line is specified as a project-level expectation. A biomedical, psychology, rehabilitation or sports-science applicant should show quantitative readiness; a computing applicant should demonstrate awareness of confounding, measurement labels and the limits of diagnosis. No prior clinical credential is listed. In an application statement, a strong example would describe a dataset, the baseline model, how leakage or imbalance was handled, and how uncertainty was explained to non-technical collaborators.

Career Paths and Market Context

Adjacent routes include academic work in digital health or movement research, biomedical data analysis, clinical-technology evaluation and applied machine learning in health research or health technology. Skills England describes broad life-sciences growth and demand for digital, data and AI capabilities, but its sector-wide outlook does not estimate vacancies for video-based mood and fatigue biomarkers or placements from this PhD. Translating a research signal into clinical technology would also require validation, data governance and clinical collaboration. The project does not promise a clinical product or employment outcome. Skills England life-sciences assessment

Next Actions and Questions to Verify

Before applying

1. Ask PGR Admissions and the lead supervisor to confirm the noon 8 December deadline, the 8 January reference date, the 2027 interview schedule and the 2027/28 stipend amount.
2. Ask whether the award registers the student in Public Health, Computer Science or another route; request the full programme title and location. Clarify how “associated training at some level” is assessed alongside general DLA degree criteria.
3. Confirm fee-status eligibility, international award availability, fee-waiver terms and the application link. The project PDF names Helen Dawes (h.dawes@exeter.ac.uk) as lead and lists Hang Dong, Pavlos Evangelidis, Chris Fox and Lu Liu as additional supervisors; the general application contact is pgrapplicants@exeter.ac.uk.

At interview

• What are the counts by condition, age and mood/fatigue level? Which questionnaires or clinical assessments define the outcomes, when were they collected relative to the videos, and are there repeated observations per person?
• How will participant-level leakage be prevented? What baselines, missing-data strategy, class-balance handling, calibration and internal/external validation are planned? Will errors be reported separately across conditions and mobility levels?
• Is the intended use association, monitoring, screening or clinical decision support? What evidence would be required before making a diagnostic claim, and how will patients and clinicians contribute to interpreting features?
• Who leads the computer-vision, statistics and clinical interpretation components, and what dataset, computing and video-processing resources are available?

Before accepting

Obtain written confirmation of the registering programme, stipend amount/year, funded period, fee coverage, start/end dates, dataset access and governance, RTSG use, and thesis deadline.

Evidence and Sources

Official project, programme and funding

• Official project description PDF
• EPSRC DLA award 6008
• EPSRC DLA applicant information and eligibility
• EPSRC DLA 2027 entry page
• Public Health MPhil/PhD route
• Research degree durations
• REF 2021 subject profiles
• Exeter student budget estimates

Supervisors and research environment

• Helen Dawes, University of Exeter
• Helen Dawes, NIHR Exeter BRC
• Pavlos Evangelidis, NIHR Exeter BRC
• Chris Fox, NIHR Exeter BRC
• Lu Liu, University of Exeter
• AI for Humans and Health
• Exeter Digital Health network
• NIHR Exeter BRC rehabilitation theme

Recent papers and field evidence

• Parkinson’s walking and computer vision, 2026 conference abstract
• Long COVID symptom fluctuations, 2024 cohort study
• Digital fatigue monitoring perspectives, 2024
• Knee4Life computer-vision protocol, 2024
• Digital home symptom monitoring protocol, 2023
• Large-scale digital gait and incident depression study, 2023
• In-home movement and depression feasibility study, 2024
• Systematic review of digital biomarkers in depression, 2024
• Systematic review of digital fatigue biomarkers, 2025

Population, career and governance context

• Major conditions strategy evidence
• Skills England 2026 life-sciences skills assessment
• Health Research Authority guidance on data-driven health technology