Brain Science PhD: Virtual Brain Networks to Inform DBS for Childhood Epilepsy

Brain Science PhD: Virtual Brain Networks to Inform DBS for Childhood Epilepsy

Quick Decision

The project would build patient-specific brain-network models from childhood-epilepsy records and multimodal imaging, simulate deep brain stimulation (DBS), and compare model predictions with measured clinical effects. It combines nonlinear dynamics, network science, parameter inference and uncertainty quantification, and best fits applicants with quantitative training and programming experience. Prior neuroscience or clinical-research experience is not required. Cohort access, outcome data and validation scope are the decisive feasibility questions; prospective treatment selection is not promised.

Research Problem and Doctoral Work

What Is Unsolved and Why It Matters

The project description frames the problem as a mismatch between growing use of electrical stimulation for children whose epilepsy is not controlled by medication and the difficulty of choosing an effective stimulation location or timing for an individual child. It says those choices are currently refined by trial and error over months. The proposed research treats seizures as activity arising from interacting brain networks and asks whether patient-specific mathematical models can help explain or predict responses. [S01]

This is a computational research question with clinical relevance. Public evidence does not yet establish that a model can reliably select treatment settings for an individual child. Recent virtual-brain work shows a technical direction, but its two cases were adults and the task was stimulation mapping; pediatric DBS studies remain small and heterogeneous. Accordingly, a useful research contribution could be improved model construction or evaluation, but clinical benefit should not be assumed. [S20–S24]

Work, Methods, and Potential Outputs

The project PDF describes three core tasks: (1) develop methods to construct patient-specific brain-network models from patient recordings and multimodal imaging; (2) implement and simulate stimulation in those models; and (3) evaluate how far model predictions correspond to measured effects using clinical data. Named methodological areas include physiological models of neurostimulation, nonlinear dynamics, network science, model and parameter inference, uncertainty quantification and efficient simulation of large dynamical systems. The description says there is scope to shape the direction according to the student's interests. [S01]

That wording supports a modelling and validation doctorate, not a promise that a specific device, target, imaging sequence or prospective clinical trial will be used. Public materials do not identify the data modalities, cohort size, model class, stimulation parameters, clinical endpoints, or whether comparisons will use retrospective, prospective or external data. Those details will determine how much time is spent on data preparation, model development, inference and validation. [S01]

Potential technical questions include whether patient-specific parameters can be identified from available recordings, how uncertainty propagates from measurements into predicted stimulation effects, and whether a model remains calibrated on patients not used for fitting. Efficient emulation could help when exploring large parameter spaces. These are reasonable research angles, not guaranteed project deliverables. [S17–S20]

School, Programme and Field Context

The stated programme is a full-time PhD in Mathematics and Statistics at the University of Exeter, with the project based at the Living Systems Institute in the Sir Steve Smith Building. Exeter is the degree-awarding institution. The named SickKids and Bristol collaborators are listed as additional supervisors; the materials do not indicate a joint degree or separate registration. [S01, S09]

Exeter's Mathematics PhD research areas include systems and dynamics, control, modelling living systems and health applications. The Mathematics for Health and Life Sciences activity connects mathematical modelling with biological and clinical questions, while the Quantitative Health Network describes a broader interdisciplinary community. These pages support environmental fit, but do not guarantee project-specific data, compute or training access. The programme offers wider doctoral training; ask which modules or methods courses are available to this student. [S05, S07, S08]

For subject-level research context, REF 2021 reports that Exeter's Mathematics and Mathematical Sciences submission received 42% 4-star and 52% 3-star ratings. REF assesses research quality in a subject area; it is not a university league table or a rating of this project. QS places Exeter at #136 in its 2027 overall World University Rankings, which is broad institutional context rather than a mathematics-specific measure. [S10, S11]

Exeter lists a maximum full-time PhD registration period of four years, while this award funds up to 3.5 years. A maximum is not a prediction of how long this project takes, but applicants should confirm the expected funded completion and submission dates, and what happens if registration continues after stipend funding ends. [S02, S06]

Supervisor and Lab Assessment

Supervisor Research Directions

Professor Marc Goodfellow leads the project. His Exeter profile focuses on brain dynamics and epilepsy. Dr Piotr Slowinski works on data science and dynamical models for complex biological systems, including translation to healthcare. Dr George Ibrahim is a pediatric neurosurgeon and senior scientist at SickKids whose work includes pediatric epilepsy and neuromodulation. The project also names Mr Will Singleton; an NHS source documents his pediatric neurosurgery role and the Bristol hospital's epilepsy surgery activity. This mix is relevant to computational modelling grounded in clinical practice. It does not, by itself, establish data-sharing agreements, access to a particular clinic, or how frequently each collaborator will supervise. [S01, S13–S16]

Recent Important Publications

• Pediatric EEG and neural-mass modelling (2025). A study of 15 children with epilepsy and 16 controls used neural-mass models to relate fitted parameters to hyperexcitability and slow-wave sleep changes. It shows recent work connecting pediatric recordings with mechanistic models. Its sample is small and it studies sleep EEG, not DBS response. [S17]
• Parameter mapping (2023). A global nonlinear method combines multiobjective evolutionary search with weighted visibility-graph features to map parameters of neural-mass models. It is relevant to inference from observed brain signals, but algorithm-derived distributions should not be read as full Bayesian posteriors; results can depend on model assumptions and search objectives. [S18]
• Efficient simulator emulation (2024). Random Fourier features are used to emulate flow maps of complex dynamical systems and represent uncertainty, with examples including Lorenz and van der Pol systems. This is a possible methodological reference for fast simulation, not evidence of validation on clinical brain data. [S19]

Lab, Resources, and Collaboration Environment

Exeter's health-and-life-sciences mathematics group works across neural and biological data, with clinical applications and epilepsy-related activity. The Living Systems Institute and Quantitative Health Network give the project an interdisciplinary institutional setting. The public pages refer to computational resources and wider training, but they do not state a dedicated GPU/HPC allocation, protected-data environment, or project-specific training budget beyond the advertised £5,000 research-training support grant. Confirm resource access and the operational roles of all supervisors. [S02, S05, S07, S08, S13–S16]

Field Evidence and Clinical Translation

Patient-specific virtual-brain modelling is an active research direction. A 2025 study combined MRI/connectivity, EEG or SEEG, stimulation and Bayesian inference in two adults with epilepsy. It supports the technical possibility of individualized models, while the small adult sample and diagnostic stimulation-mapping task leave pediatric therapeutic DBS prediction unproven. [S20]

Pediatric clinical evidence is developing but still limited. In the ADVANCE study, a partially randomized, non-blinded preference trial involving 18 children reported seizure reduction of 51.9% versus 12.3% in its comparison, but found no quality-of-life change. A pediatric centromedian DBS registry reports quality-of-life and functional outcomes, but observational registry findings do not establish causal or individualized prediction. A 2026 systematic review covered 19 studies and 149 patients; a 2026 multicentre retrospective cohort covered 12 centres. Together, these sources underline small samples, varied targets and designs, and the need for prospective, multicentre work. Pooled responder percentages are population summaries, not forecasts for a particular child. [S21–S24]

NICE HTG550 addresses DBS for refractory epilepsy in adults. Its governance and audit/research recommendations are useful context for adult practice, but it is not pediatric guidance and should not be used to infer a regulatory pathway for this project. Ask the supervisors whether the doctorate aims at methodological validation, clinical decision support in future research, or another translational endpoint. [S25]

Funding, Application, and Living Cost

Award and Application Conditions

The project is listed under Exeter's EPSRC Doctoral Landscape Award for 2027/28 entry. The award page states up to 3.5 years full-time, Home fees and a £5,000 research-training support grant; it also says successful international candidates receive fee-gap coverage. The central award information describes a limited international share of the cohort. Visa fees, the immigration health surcharge and relocation are identified as costs the student must cover. Confirm the details and timing of any fee waiver directly with the university. [S02, S03]

The advertised application deadline is 12:00 noon, 8 December 2026. The project-specific award page says two references must be received by 8 January 2027, but the general EPSRC DLA page says references are due by 15 February 2027. The official pages do not reconcile this difference. Plan for the earlier 8 January date and ask the project team in writing which deadline applies. Interviews are likely 15–24 February 2027, with a possible technical interview. The award page requests a personal statement, CV, transcripts, two references and English-language evidence where applicable; qualification documents must be in English or accompanied by certified translations. Check the application portal for any later notice. [S02, S03]

There are three details to reconcile:

1. The award summary calls £21,805 the 2027/28 stipend, while its detailed text calls £21,805 the 2026/27 rate. UKRI currently confirms £21,805 as the minimum for 2026/27, not a confirmed 2027/28 amount. Ask for the amount applying at the intended start date and whether it is uprated. [S02, S04]
2. The project-specific page gives 8 January 2027 as the reference deadline, while the general EPSRC DLA page gives 15 February 2027. These official dates conflict. To reduce the risk of a late reference, plan for 8 January and request written confirmation of the applicable deadline from the project team. [S02, S03]
3. The project PDF says applicants should have a First/Upper Second or a relevant quantitative master's. The general award page says a Lower Second may be considered with a master's or significant relevant non-academic experience. Ask whether that award-wide exception applies to project 6019. [S01–S03]

Exeter Living-Cost Comparison

Exeter estimates basic costs for a single student at £1,300–£2,000 per month, based on local average spending in April 2026. Dividing the currently advertised £21,805 annual stipend by 12 gives about £1,817 per month: roughly £517 above the lower estimate and £183 below the upper estimate. This is a simple comparison, not a personal budget; it excludes one-off moving costs, travel, visa charges and the immigration health surcharge, and the stipend year remains unclear. [S02, S04, S12]

Applicant Fit and Preparation

Required Background and Skills

The project welcomes applicants from mathematics, physics, engineering, computer science, quantitative neuroscience or related quantitative disciplines. Experience programming in any language is expected; familiarity with Python, Julia, MATLAB or R is helpful. Prior neuroscience or clinical-research experience is not required, and training will be provided. [S01]

The project-specific academic wording is a First/Upper Second or relevant quantitative master's; the broader award-wide Lower Second exception needs confirmation. Applicants should distinguish their formal qualification evidence from any relevant experience they hope may qualify under that exception. [S01–S03]

Project-specific enquiries are directed to lead supervisor Marc Goodfellow (m.goodfellow@exeter.ac.uk). The PDF also lists Piotr Slowinski (P.M.Slowinski@exeter.ac.uk), George Ibrahim (george.ibrahim@sickkids.ca), and Will Singleton (William.Singleton@uhbw.nhs.uk) as additional supervisors. [S01]

Preferred Experience and Suggested Preparation

Beyond naming Python, Julia, MATLAB and R as helpful, the PDF does not list other preferred experience. As further preparation, applicants could review dynamical systems, networks, parameter estimation and uncertainty, and prepare an example of reproducible scientific code or modelling. These are suggestions based on the advertised methods, not additional entry requirements. A concise explanation of modelling assumptions, what a model can and cannot predict, and willingness to learn the clinical context would help show readiness for interdisciplinary work. [S01]

Career Paths and Labour-Market Context

Exeter's Mathematics programme lists example destinations including research and postdoctoral work, quantitative analysis, data science, statistics and technical analyst roles, while cautioning that these examples may not correspond to this specific programme. For this project, dynamical-systems modelling, statistical inference, uncertainty quantification, scientific programming, clinical-data collaboration and simulation/emulation may transfer to computational neuroscience, academic research, biomedical modelling, health data science, quantitative R&D and neurotechnology. These are plausible skill pathways, not promised placements. The 2025 UK AI Labour Market Survey reports that 97% of respondents identified at least one skills gap, 57% a technical skills gap and 35% of organisations struggled to fill AI roles. This broad employer survey is not healthcare-specific and cannot be extrapolated to clinical hiring or PhD graduate employment or salary outcomes. No project-specific placement rate is published. [S05, S26]

Next Actions and Questions to Verify

Before Applying

• Confirm the 2027/28 stipend, any uprating, and the fee-waiver conditions for international applicants. [S02–S04]
• Confirm whether the award-wide Lower Second exception applies to this project. [S01–S03]
• Request written confirmation of whether the reference deadline is 8 January (project award page) or 15 February 2027 (general DLA page); plan for 8 January meanwhile. Also verify required English-language and transcript documents. [S02, S03]

At Interview

• Which patient recordings and imaging modalities, cohorts, stimulation targets and clinical outcomes are expected to be available?
• Are the data already collected, and what consent, ethics and data-access approvals are in place?
• How will predictions be compared with measured clinical effects? Is validation retrospective, prospective or on external patients, and what would count as a useful result?
• Which model class and stimulation parameters are initial priorities? Which aspects can the student shape?
• What are the respective roles of the four named supervisors, and how often will the student meet the clinical collaborators? [S01, S13–S16]

Before Accepting an Offer

• Confirm how up-to-3.5-year funding aligns with the expected submission schedule and the four-year maximum registration period. [S02, S06]
• Ask what compute, storage/security environment, methods training and clinical-site access are available to this specific project. [S05, S07, S08]
• Confirm the expected out-of-pocket costs, particularly visa, healthcare surcharge and relocation. [S03, S12]

Evidence and Sources

Project, award, and university

• [S01] Official project description (PDF): source
• [S02] Exeter award 6019, Mathematics and Statistics: source
• [S03] Exeter EPSRC Doctoral Landscape Award information: source
• [S04] UKRI support for UKRI-funded postgraduate research students: source
• [S05] Exeter Mathematics PhD and research degrees: source
• [S06] Exeter research-degree durations: source
• [S07] Exeter Mathematics for Health and Life Sciences: source
• [S08] Exeter Quantitative Health Network: source
• [S09] University of Exeter contact and location: source
• [S10] Exeter REF 2021 results: source
• [S11] QS University of Exeter profile: source
• [S12] Exeter international student budgeting: source

Supervisors and research environment

• [S13] Marc Goodfellow, Exeter profile: source
• [S14] Piotr Slowinski, Exeter profile: source
• [S15] George Ibrahim, SickKids profile: source
• [S16] Bristol Royal Hospital for Children pediatric neurosurgery source: source

Publications and clinical field evidence

• [S17] Neural-mass modelling of sleep changes in children with epilepsy (2025): source
• [S18] Global nonlinear parameter mapping for neural-mass models (2023): source
• [S19] Emulating complex dynamical simulators with random Fourier features (2024): source
• [S20] Virtual brain twins for stimulation in epilepsy (2025): source
• [S21] ADVANCE pediatric DBS trial (2024): source
• [S22] CHILD-DBS pediatric registry report: source
• [S23] Pediatric drug-resistant epilepsy DBS systematic review and meta-analysis (2026): source
• [S24] Multicentre pediatric thalamic neuromodulation cohort (2026): source
• [S25] NICE HTG550, adult refractory epilepsy DBS: source

Career-market context

• [S26] UK AI Labour Market Survey 2025 executive summary: source