Abstract
Damage to the smallest blood vessels in the lungs (the microvasculature) is associated with several lung diseases, but is difficult to measure or image.

Hyperpolarised 129Xe MRI is a powerful tool for functional lung imaging and can be used to quantitatively measure gas exchange. 129Xe is soluble in the lung alveolar membrane, capillaries and red blood cells (RBCs). The 129Xe RBC signal oscillates with the same frequency as the heart beat, as a result of changes in the capillary blood volume over the cardiac cycle. A reconstruction method has been developed by the POLARIS group which allows regional mapping of the 129Xe RBC oscillation amplitude and phase, from which spatial variability in microvascular function can be inferred [1]. Further work is needed to validate the RBC oscillation mapping technique against other modalities such as dynamic contrast enhanced (DCE) MRI.

A numerical model of the pulmonary circulation has been developed by researchers at the ABI [2]. This project will involve comparing patient-specific perfusion simulations from this model to 129Xe MRI and DCE-MRI, in order to help validate these novel techniques and extend our understanding of blood flow limitation in diseases such as pulmonary hypertension.

[1] Pilgrim-Morris JH, Collier GJ, Takigawa M, Strickland S, Thompson R, Norquay G, Stewart NJ, Wild JM. Mapping the amplitude and phase of dissolved 129Xe red blood cell signal oscillations with keyhole spectroscopic lung imaging. Magn Reson Med. 2025 Feb;93(2):584-596. doi: 10.1002/mrm.30296.

[2] Ebrahimi, B.S., Tawhai, M.H., Kumar, H., Burrowes, K.S., Hoffman, E.A., Wilsher, M.L., Milne, D. and Clark, A.R. (2021), A computational model of contributors to pulmonary hypertensive disease: impacts of whole lung and focal disease distributions. Pulmonary Circulation, 11: 1-15 20458940211056527. https://doi.org/10.1177/20458940211056527

Aims and Objectives
1) Perform analysis of spectroscopic 129Xe lung MRI imaging data using existing analysis tools developed by the supervisor in MATLAB, in order to map the  oscillations of 129Xe red blood cell (RBC) signal

2) Compare RBC oscillation amplitude and phase maps to metrics from established lung MRI methods

3) Compare simulations from a model of pulmonary blood flow to lung MRI

Workplan
Week 1: The start of the project will involve getting familiar with lung MRI theory, through reading the literature and tutorials given by the supervisor and other members of the POLARIS group. The student will meet virtually with our collaborator at ABI to learn about the numerical pulmonary vasculature model.

Week 2:  The student will be given a demonstration of the RBC mapping code and methodology and their first task will be to perform analysis of a previously analysed dataset, in order to gain understanding of the technique. From this point, the student can begin to analyse patient data. 

Weeks 3-5: The student will work on validation of the RBC mapping technique, by comparing with metrics from other established lung MRI techniques, such as DCE perfusion MRI, and a numerical model developed by ABI.

Week 6: The final week will be dedicated to the student finishing outstanding tasks and making a poster of their work.

Throughout the placement, the student will have the opportunity to take part in other group activities, for example, the group scientific seminar.

Skills needed
Essential skills: programming in Python and Matlab

Abstract
This project offers an innovative approach to understanding pulmonary arterial hypertension (PAH), a progressive and life-threatening condition. While conventional cardiovascular modelling focuses on beat-to-beat mechanisms, key clinical biomarkers of disease progression manifest over months to years, yet this temporal disconnect remains largely unaddressed in current literature. Working at the intersection of bioengineering and clinical medicine, the successful candidate will develop and analyse novel non-pulsatile cardiovascular models to investigate long-term disease progression in PAH.

The research will explore how non-pulsatile models can capture fundamental PAH disease mechanisms and long-term regulatory processes crucial for diagnosis and therapy planning. Key areas of investigation include right ventricular adaptation and pulmonary vascular remodelling. Through comprehensive sensitivity analyses, the project will identify critical parameters influencing disease progression. This clinically-driven approach focuses on developing models that directly correspond to clinical disease markers, diverging from traditional temporally-resolved modelling approaches however we will still make comparisons to these models.

The successful candidate will develop expertise in:

1) Mathematical modelling of physiological systems

2) High-performance computing and sensitivity analysis

3) Cardiovascular physiology and pathophysiology

4) Clinical research collaboration

This project is ideally suited to candidates with strong mathematical and computational skills, particularly those with backgrounds in bioengineering, applied mathematics, or related fields. Whilst experience in physiological modelling or programming is beneficial, comprehensive training will be provided. This project will aim to utilise the Julia language however experience in Python and Matlab will be sufficient.

This research forms part of a broader programme developing digital twins for PAH, offering opportunities to contribute to cutting-edge research with direct clinical applications. The project benefits from established collaborations with the world-renowned Auckland Bioengineering Institute, known for their pioneering work in physiological modelling, and Sheffield Medical School. These partnerships ensure both technical excellence and strong clinical translation, whilst providing unique opportunities for international collaboration and knowledge exchange.

Aims and Objectives
1) Conduct a comprehensive literature review examining non-pulsatile cardiovascular models, with particular emphasis on gold standard applications in spaceflight and orthostatic intolerance. 

2) Develop a non-pulsatile mathematical model of the cardiovascular system. 

3) Integrate pulmonary arterial hypertension disease mechanisms into the model, conducting sensitivity analyses to compare influential mechanisms with pulsatile equivalents.

4) To implement and analyse a long-term regulatory mechanism associated with pulmonary arterial hypertension, focusing on either right ventricular hypertrophy, ventricular-arterial coupling or pulmonary vascular remodelling.

5) Maybe: If to a sufficient level prepare a peer-reviewed publication.  

Workplan
Week 1: Literature Review and Theoretical Foundations

1) Systematic collection and review of relevant academic papers

2) Undertake structured tutorials on non-pulsatile model theory

3) Comprehensive review of PAH pathophysiology and progression

Week 2: Model Development - Initial Phase

1) Critical analysis of existing pulsatile cardiovascular models

2) Development and implementation of standard ODE pulsatile cardiovascular model

3) Theoretical framework development for non-pulsatile equivalent model

Week 3: Model Development - Core Implementation

1) Implementation and validation of non-pulsatile cardiovascular model

2) Comparative analysis of outputs between non-pulsatile and pulsatile models

3) Documentation of model development and initial results

Week 4: Disease Mechanism Integration

1) Structured review of PAH disease mechanisms

2) Implementation of identified mechanisms in both model types

3) Validation of implemented disease mechanisms

Week 5: Sensitivity Analysis and Regulatory Mechanisms

1) Theoretical foundation work on sensitivity analysis methodologies

2) High-Performance Computing (HPC) training for sensitivity analyses

3) Execute and analyse local sensitivity analysis results

4) Literature review of PAH-associated long-term regulatory mechanisms

Week 6: Advanced Analysis and Mechanism Implementation

1) Analysis and visualisation of global sensitivity analysis results

2) Implementation and testing of selected long-term regulatory mechanism

3) Documentation of sensitivity analysis findings

Week 7: Comprehensive Analysis and Initial Documentation

1) Conduct sensitivity analyses incorporating regulatory mechanisms

2) Validate model outputs against established PAH progression patterns

3) Begin manuscript preparation, focusing on: Introduction, Literature review & Scientific novelty

4) Develop presentation materials for dissemination

Week 8: Final Analysis and Research Dissemination

1) Complete comparative analysis of sensitivity studies

2) Continue manuscript development

3) Finalise presentation materials

4) Present findings at CVD-Net weekly meeting (PAH Digital Twin programme)

Skills needed
This project is ideally suited to candidates with strong mathematical and computational skills, particularly those with backgrounds in bioengineering, applied mathematics, or related fields. Whilst experience in physiological modelling or programming is beneficial, comprehensive training will be provided. This project will aim to utilise the Julia language however experience in Python and Matlab will be sufficient.

Abstract
Individual exercise responses provide useful -possibly the most useful- diagnostic information available to the clinician across a range of cardiovascular diseases. These include those with the highest morbidity-  ischemic heart disease, valve disease and hypertension. Medics currently use a range of proxies to assess cardiac energetics in the exercise state, tolerating a certain amount of ambiguity. What is required are more objective, direct energetic measures. To provide these, personalisable models are an obvious place to start. 

We hypothesize that computational 0D cardiovascular models have the scope to deliver objective mechanical metrics of exercise. Currently, such tools very seldom represent the heart outside of the rest state. Further, what exercise models exist are difficult to validate, due to the challenge of acquiring meaningful cardiac data in the exercise state. 

Exercise MRI data is currently being acquired within the Mathematical Modelling in Medicine Group (MMMG) in Sheffield University, which will eventually allow the validation of exercise models. In this project, the student will gain an understanding of this cutting edge data collection method and its encapsulating protocol and participate in developing the 0D models of the cardiovascular system under exercise. These models will eventually provide useful clinical metrics of cardiac energetics which, if accurate and verifiable, could be used in the near future to aid diagnosis, across a range of cardiovascular conditions.

We understand that this project is ambitious, but given the paucity of research in this area, the student will benefit from this project- even a thoughtful review and critique of exercise effects to include in cardiovascular models and the attending protocols would comprise a valuable outcome. 

Aims and Objectives
1. To gain familiarity with exercise data collection, and 0D modelling principles

2. To explore and test LV pressure estimation methods

3. To analyse the differences in model output when data from different exercise patients is inputted into the same model  

Workplan
Week 1: Introductory research

Learn the basic concepts of 0D modelling and perform a literature search for existing models of exercise within the literature. To understand the data collection method used during exercise MRI.

Week 2: 0D model building

Code 0D models of increasing complexity, and briefly test the solver method. 

Week 3: Model calibration

Choose parameter values for the model to maintain reasonable physiological results. 

Week 4: Cardiac energetics

Learn how to compute the cardiac energetics from the model output, and make predictions about the change under exercise based on the literature

Week 5: Exercise effects

Choose which exercise physiology effects to represent and test in the model, based on the literature, and begin to code. 

Week 6: Exercise models

Complete inclusion of exercise effects into the model. Run the model using the parameter values chosen in week 3 to check implementation.

Week 7: Test exercise model

Use the model parameter values found in week 3 to personalise the exercise model. Run the model to predict the cardiac energetics and compare with hypotheses from week 4. 

Week 8: Data curation and dissemination

Prepare data and work for a poster presentation

Skills needed
Candidates from any background are invited to apply. Previous experience with at least one of MATLAB, Julia or Python is required, but most important is a willingness to learn.

Abstract
This project aims to use an already developed QCT-FE pipeline from medical images of human vertebrae to the estimation of ultimate failure load. This pipeline will deviate from the existing pipeline in separate segmentation of bone and the lytic lesion followed by development of homogenized finite element models of the bone and assigning different material models to the lytic lesion. Non-homogenous isotropic elastoplastic material model will be assigned to the bone and different material models (elastic, hyper-elastic, poro-elastic, poro-hyper elastic) from the existing literature will be assigned to the lesion. Ultimate failure load will be predicted through an engineering based suitable failure criterion. Results from this study will provide an insight into the fracture risk assessment of vertebral bodies with lytic metastatic lesions.

Aims and Objectives
-Segmentation and reconstruction of vertebra and the lesion separately from QCT based medical images

-Development of homogenized finite element models of reconstructed vertebrae

- Application of different material models (elastic, hyper-elastic, poro-elastic) to the reconstructed lesions

-To analyse the minimum principal stress, minimum principal strain and ultimate failure load on a loaded vertebral body

Workplan
WP 1: Identification and segmentation of CT datasets with lytic lesions (2 weeks).

This work package involves identifying the CT datasets with metastatic lytic lesions from an already available database procured from 15 cadaveric donors. Ethical approval for this database is already in place. All the identified CT datasets will be segmented using a suitable image processing tool, 3D Slicer and further reconstructed into faceted bodies in stereolithographic formats.

WP 2: Development of end plates and FE models (2 weeks)

This involves developing the superior and inferior bony endplates and stitching them with the faceted body in ANSYS Space Claim to convert those faceted bodies into solid bodies for further simulation.

WP 3: Assigning material properties to vertebral body and lesion (2 weeks)

Densitometric CT Calibration will be performed from the CT datasets with in-line phantoms through an opensource image processing tool, ImageJ and heterogeneous isotropic elastic material property will be assigned to the vertebral body using an opensource software Bonemat through a power-law relationship between elastic modulus and apparent density of bone. Three separate FE models will be developed for each selected vertebra with lesion to evaluate the influence of three different material models (elastic, hyper-elastic, poro-elastic, poro-hyper elastic) of the lesion. 

WP 4: Evaluation of local stress/strain and preparation of report (2 weeks)

Finally, applying a suitable uniaxial compression on the cranial endplates and restraining the caudal endplate, stress/strain distribution across the vertebral body will be analysed and ultimate failure load will be predicted in each case which will provide an insight into fracture risk assessment which will be evaluated as a future scope of this present study.

Skills needed
Essential Skills: ANSYS Mechanical APDL, ANSYS SpaceClaim, 

Desirable Skills: Basic knowledge of APDL and Matlab scripting is desirable.

Preferable Skills: ImageJ, Bonemat, any medical image processing software such as Mimics/3D Slicer/Amira

Abstract
Background:

The primary function of the lung is gas exchange, in which ventilation plays a major part; ventilation is the process by which air flows in and out of the respiratory system to supply oxygen and expel carbon dioxide. 

The lung is a highly heterogeneous organ both functionally and structurally. Capturing regional information enables greater sensitivity to local tissue-level changes and detection of dysfunction in both localised and systemic disease (Hill and van Beek 2004). 

Assessing regional lung ventilation traditionally involves the use of contrast agents, which come with various limitations, the primary of which is their limited accessibility. Ventilation images acquired without the use of contrast agents, or “Non-contrast” Ventilation CT and MRI images present an attractive avenue for the extraction of regional lung information, whereby changes in lung volume between lung inflation states such as full inspiration and full expiration, extracted by image segmentation, can be inferred via deformable image registration. Image registration also aligns image voxels so changes in image intensity between inflation level can also be derived. Together, using these outputs, we can model regional ventilation as a 3D map of local function. 

The sensitivity of these non-contrast regional ventilation maps will be investigated in healthy subjects and across various disease cohorts. This validation is a critical step in establishing the utility of non-contrast imaging techniques and in advancing their adoption as accessible and effective tools for regional lung function assessment.

Methods:

The student will perform each of the steps of data curation, image segmentation, image registration and synthesis of the surrogate ventilation images, on a novel dataset containing a variety of diseases, and healthy subjects.

Evaluation:

The student will extract a variety of quantitative parameters such as: lung volumes from the segmentations, ventilation distribution and heterogeneity. 

Using these parameters the student will also perform comparative analysis on the disease cohorts. 

Aims and Objectives
•        Perform a brief review of ventilation imaging literature 

•        Perform image segmentation on CT images to delineate the lungs. 

•        Perform deformable image registration on segmented lung images, and generate Ventilation images 

•        Analyse differences in ventilation patterns between health and disease

Workplan
Students will primarily be assisting with image segmentation on a novel CT dataset for the bulk of the project. This will involve the delineation of lung regions within various disease cohorts on CT and MRI images. 

Students will also be taught how to extract parameters such as lung volume from their segmentations, and how to perform deformable image registration on their segmented images on the Sheffield High Performance Computer system.

Additionally, students will also be shown how to generate regional ventilation maps from their registered images and analyse their data.

The students will also be given the opportunity to present work to supervisor, and a wider research group. 

Week 1 – Review literature, and receive training on image segmentation

Week 2 – Begin Image segmentation

Week 3 – Image segmentation

Week 4 – Image segmentation / Training on image registrations

Week 5 – Image segmentation / Training on generation of ventilation images

Week 6 – Finalise image segmentations, Generation of Image Registrations & Ventilation Images

Week 7 – Analysis of Data (Segmentations & Ventilation images)

Week 8 – Preparation of figures and content for poster

Skills needed
Training will be provided; however, students should be familiar with a computer and installing software. Coding experience (Python, MATLAB, R, BASH) is beneficial but not required.

Abstract
Primary Care Support England (PCSE) provides support services to key front-line health services such as General Practitioners (GP). As part of this support, PCSE issues guidance to GPs in the sensitive situation of adoption and gender reassignment. During a patient’s transition in these settings it is essential that patient medical records are updated appropriately. The current established guidance requires a transfer of all patient information to a new medical record, however all prior identifying information with regards to the change must be redacted.

For example, adoption legislation mandates issuing adopted patients a new NHS number and the creation of a new medical record. Importantly, the new record excludes any information about the identity or location of birth parents. Similarly in gender reassignment, a new medical record is issued, however all gender identifying information must be removed.

Existing guidance in this setting relies on manual redaction, which is time-expensive, and potentially prone to error and inconsistencies. This project looks to address key questions: can artificial intelligence (AI) be used to automate the redaction processes? Can AI ensure sensitive information is handled accurately and efficiently?

The project aims to develop a language model or multimodal AI system to identify and redact sensitive information related to gender and pre-adoptive identities in medical records, demonstrating improvements in accuracy, speed, and reliability over manual practices. 

Through data collection and analysis, the study will evaluate the effectiveness and real-world applicability of AI-based redaction in Primary Care settings. The findings aim to streamline processes, reduce administrative burdens, and maintain patient care while safeguarding sensitive information, laying the groundwork for future research and scalable AI solutions in medical record management.

Aims and Objectives
- Review the current NHS England and Primary Care Support England (PCSE) guideline on manual redaction of sensitive personal information (SPI) in patient medical records during adoption and gender reassignment. This will help gather knowledge of where and how redaction can be automated.

- Design, implement and test a system that utilises existing AI models for the redaction of personal information necessary to preserve changes in adoption and gender reassignment. This will be used to demonstrate the models ability to identify and redact SPI during gender reassignment and pre-adoptive identities.

- Collect and prepare public medical record data for understanding the effectiveness and limitations of using AI for redaction in adoption and gender reassignment. This will involve sampling from public records to identify medical records that display gender and familial information. These samples will be necessary for assessment of the system.

- Assess the performance of the AI system against current manual redaction practices in terms of accuracy, efficiency, and reliability. Potentially, collaborating with local general practitioners (family doctors) in Sheffield and Wolverhampton to compare with real clinical samples.

Workplan
Week 1-2: Orientation, and Patient and Public Involvement and Engagement.

- Review current NHS England and PCSE guidelines on redaction in medical records during adoption and gender transition.

- Review potential software and tools for AI-based text redaction.

- Establish a way of working between supervisors and students.

- Establish access to compute resources, and public datasets.

- Begin discussions with relevant clinicians to identify real-world challenges in medical record redaction.

Week 2-3: Prepare and understand medical record samples.

- Familiarise with publically available and anonymized medical records.

- Gather example medical records that contain specific information about a patient's  family or gender. 

- Define sensitive data patterns (e.g., names, pronouns, familial information, gender specific interventions and treatments, pre-adoptive details).

Week 3-6: Build a prototype model for identifying and redacting sensitive information.

- Familiarise with publicly available AI-models.

- Define AI model prompts and queries to help elicit the redactive capabilities of an AI model.

- Conduct initial training and testing using prepared datasets.

Week 5-6: Evaluate model effectiveness with PPIE to benchmark against manual redaction.

- Measure the accuracy of the AI system by comparing results with data from manual redaction practices.

- Measure the time taken to manually redact a medical record, while also measuring the time taken to manually verify the redactions produced by a model.

- Use these measurements to identify strengths, weaknesses, and potential areas for improvement.

- Conduct follow-up PPIE discussions to validate AI usability and effectiveness in stakeholder-defined scenarios.

Week 6-8: Consolidate findings.

- Simulate real-world use cases for redacting adoption and gender reassignment records.

- Compile results into a structured report, highlighting key outcomes.

- Draft a summary of pilot data for potential use in future research proposals.

Week 8: Dissemination of findings.

- Create a poster and/or presentation of the research outcomes, including challenges, achievements, and future directions.

Skills needed
This project will be well suited to most students with an interest in health and/or artifical intelligence. It’s not essential, however, a student having programming skills with Python or another language, such as Javascript/R/Rust, will make analysis and interaction with an AI model easier and may allow for a deeper dive into experimentation with a variety of AI models.

Abstract
Electrical impedance spectroscopy is an emerging technology permitting a non-invasive identification of cancerous changes in various organs, such as cervical epithelium, skin or breast tissue. Electrical impedance instrumentation measures the opposition the tissue exhibits when exposed to a small alternating current. The measured passive electrical properties depend on the structure and composition of the biological material on the cellular and tissue scale levels. Knowing the structural and chemical properties changes the tissue undergoes with the progression of cancer we should be able to distinguish cancerous lesions based on their electrical impedance. Computational modelling methods can be used to broaden our understanding of the connection between tissue structure and their passive electrical properties improving the cancer diagnosis.

During the project, the student will investigate various mathematical modelling methods to compute electrical properties of a healthy and cancerous oral cell. This will be preceded by a literature review on current state of the art research concerning electrical impedance spectroscopy in cells and tissue applications. The outcomes of the project will consist of a qualitative and quantitative comparison of the computed electrical properties of healthy and cancerous cells obtained with different modelling methods along with a critical analysis of the feasibility of these methodologies in cell modelling.

The work carried out in the summer project is directly related to the objectives of our ongoing multidisciplinary study concerning cancer diagnosis in oral tissues using electrical impedance spectroscopy. In our current work, we develop tissue engineering models which inform our virtual tissue constructs using finite element methods to simulate the electrical impedance spectra for healthy and cancerous oral mucosa.

Aims and Objectives
•        Literature review on electrical properties of cells, and healthy and cancerous oral epithelium;

•        Literature review on mathematical modelling methods for electrical properties of cells (such as equivalent circuit modelling, maxwell’s mixture theory); 

•        Computational modelling of electrical properties of healthy and cancerous oral cells using chosen modelling methods (with a simple cell equivalent circuit model as a starting point, an investigation of additional methods will be encouraged);

•        Comparison of the obtained results with the numerical data provided by the research group;

•        Discussion of the advantages and limitations of selected cell modelling methods in the assessment of electrical properties of healthy and cancerous oral cells.

Workplan
The project can be divided into four phases, after each phase the student will be asked to present and discuss their outcomes with the project supervisor. 

Phase One:  Duration 2 weeks, the student will carry out a literature search to investigate current studies on electrical impedance measurements in healthy and cancerous oral epithelium. Additionally, the student will gain an in-depth understanding on the physical principles underlying the electrical behaviour of biological cells and the recent computational methods aiming to predict and interpret them. Outcomes: Written summary of the literature review (around 3-5 pages). Student will propose initial ideas for the modelling work. 

Phase Two: Duration 3 weeks, the student will explore different mathematical modelling approaches to compute the electrical impedance of healthy and cancerous oral cells. The tissue engineering and computational labs will provide the student with the necessary data required to complete this part of the project (e.g. initial model parameters). Outcomes: Student will present a qualitative and quantitative comparison of computed impedance of a healthy and cancerous cell, and a comparison of the results obtained with different methods (the latter being optional). The student should be able to interpret the results based on the previously gained knowledge on the structural changes in cancerous tissue.

Phase Three: Duration 2 weeks, the student will compare their results against the numerical data provided by the project supervisor. Outcomes: Student will present a critical analysis of the benefits and limitations of the chosen mathematical models when compared to the numerical solution.

Phase Four: Duration 1 weeks, Student prepares a poster summarising their summer project as the preparation for Insigneo or NIHR BRC event.

Skills needed
•        Bioengineering/Physics/Maths/Other Engineering student interested in computational modelling and medicine;

•        Good Physics understanding, especially in Electromagnetism;

•        Proficient coding skills (Matlab or Python).

Abstract
Healthcare data has shown that around 60% of deaths resulting from colon cancer are due to patients avoiding early screening, due to the stigma around the screening process. The gold standard diagnostic tool for colon cancer is flexible colonoscopy, a procedure which is often viewed as embarrassing, uncomfortable and/or painful. In an effort to increase the number of patients attending screening earlier, robotic solutions have been in development over the past two decades, with robotic flexible colonoscopes, inchworm-locomotion devices and wireless capsule robots all having been in development.

When testing new devices for use in the human colon, testing environments have previously either been inaccurate test rigs that do not accurately simulate the passive biomechanical properties of the human colon, or animal models (anesthetised pigs) that are unethical, and require animal deaths. These environments are used due to a lack of an accurate, sustainable and ethical alternative, all of which are criteria that can be met by using virtual simulation to model the colon and its interaction with new devices.

In order to validate the performance of such a simulation, a force-sensing model is required with which to compare force and deformation data. To do so, a program will be needed that processes the data from a force-sensing phantom colon model interacting with a shape-sensing colonoscope, and presents it visually in the same (or a similar) format to the visual component of the virtual simulation.

Aims and Objectives
Project aim: to program a visualisation interface to help validate a force-sensing phantom colon model, using a shape-sensing colonoscope.

Project objectives:

-        Create a program which renders a section of the phantom colon model, as well as the colonoscope;

-        Program a function which manipulates the shape and position of the virtual colonoscope based on the data input from the shape-sensing colonoscope;

-        Program a function which maps the magnitudes and locations of force data input from the force-sensing phantom colon onto the virtual section of the phantom colon.

Workplan
The student will mostly be working towards creating the code required to visualise the data from the force- and shape-sensing setup, with the possibility of some work to help design and assemble the physical setup itself, and some academic writing about the project.

In weeks 1-2, the student will have the opportunity to familiarise themself with the physical setup and virtual simulation, and will look at the existing code used to gather data from a flat, interaction-sensing e-skin to start adapting it to the force-sensing colon, which uses the same technological principles.

In weeks 3-6, the student will work on creating the code to process the data from both the force-sensing colon section and the shape-sensing colonoscope and visualise it in a format similar to that of the virtual simulation set up, allowing comparisons between the physical and virtual models to be easily visualised, as well as allowing direct comparisons of magnitude and location data of the forces generated/calculated respectively.

In weeks 7-8, the student will participate in the data collection using the physical and virtual models, and they will have the opportunity to co-author a research paper documenting the results of the project alongside creating a poster to highlight the results.

Skills needed
Programming using C

Programming using Processing

Abstract
In a collaboration with the biomechanics group of Thor Besier at the Auckland Bioengineering Institute, we will develop a multiscale model of the knee joint to study patients with osteoarthritis. Using a current mechano-inflammatory model of the knee joint developed by Juntong Lai at the University of Sheffield, we will combine it with the MSK tools developed at ABI by the group of Thor Besier. This will enable us to determine the importance of body level data and patient-specificity. 

Aims and Objectives
- Develop patient-specific finite element model of the knee joint

- Developing multiscale modelling to link rigid body motion of the leg with the FE model of the knee joint

- Initiate collaboration with the MSK group of Thor Besier who already has patient database and MSK modelling tools

Workplan
The following tasks will be performed:

- Explore the MSK tools available at ABI for the study of musculoskeletal modelling in a regular stance

- Integrate the Sheffield mechano-inflammatory model of the knee joint with the MSK model tool of ABI at the body level

- Select 5 patients from ABI or the public database OA Initiative to study the importance of the body-tissue coupling on the cartilage degradation over time

Skills needed
Any experience previously with the finite element method or other modelling skills would be useful.

Abstract
Blood pressure (BP) is a key indicator of cardiovascular health, yet traditional BP monitoring methods are invasive, intermittent, and unsuitable for continuous daily use. Wearable technologies, such as photoplethysmography (PPG) devices, offer a promising non-invasive and continuous alternative for BP estimation by capturing blood volume variations indicative of cardiovascular dynamics. This study focuses on utilising multiple wearable devices to monitor BP during physical activity by analysing PPG morphology features and pulse arrival time (PAT) in conjunction with ECG signals. The primary objective is to collect and analyse a pilot dataset for developing enhanced BP estimation models.

PPG and ECG signals will be collected from two wearable devices (e.g., Polar H10), and BP readings using Omron cuff-based monitors as the ground truth. Key PPG features, including waveform morphology, first and second derivatives, and spectral components, will be extracted and evaluated. Signal synchronisation of different wearable devices will done using accelerometer signals. PAT will be calculated as the time difference between R peaks in ECG and fiducial points in PPG signals. Machine learning models such as random forests will be trained to predict BP, with performance validated using leave-one-subject-out cross-validation and metrics such as RMSE, correlation coefficients, and Bland-Altman analysis. This pilot project will advance wearable-based BP estimation research while enabling continuous cardiovascular monitoring for daily use.

Aims and Objectives
Collect PPG and ECG signals from wearable devices.

Extract features from PPG and ECG signals

Apply machine learning techniques to estimate blood pressure.

Workplan
Data collection will involve 10 healthy young participants over a 20-minute protocol: baseline (3 minutes), walking (3-5 minutes), resting (3 minutes), treadmill running (3-5 minutes), and final resting (3 minutes). BP will be measured during rest using a cuff-based Omron device. PPG and ECG will be collected using two wearable devices (e.g., Polar H10). PPG morphology and spectral features will be extracted, and PAT will be calculated as the time difference between R peaks in ECG and fiducial points in PPG. Device synchronisation will use accelerometer data from the two wearable devices.

Machine learning models, including random forests, will be developed using PPG features and PAT as input. Leave-one-subject-out cross-validation will ensure robust performance evaluation. Statistical analyses such as Bland-Altman plots, RMSE, and correlation coefficients will assess model accuracy and reliability. The results will inform future research on wearable BP estimation using wearable devices.

Skills needed
The student should have a basic understanding of signals (time-series data) and programming experience in Python, including libraries like NumPy, pandas, and scikit-learn. Students with backgrounds in computer science, biomedical engineering, electrical engineering, mechanical engineering, or data science, or those familiar with signal processing or machine learning, are preferred.

Abstract
This project aims to investigate the potential utility of different machine learning algorithms and multimodal AI to predict the presence of heart murmurs from recorded heart sounds (phonocardiograph) Successful candidates will spend 10 weeks working with the CIRCOR heart sounds dataset to learn how to apply preprocessing and machine learning techniques to audio data. Algoirthms developed will then be applied prospectively to clinical research projects investigating phonocardiography in neonates in the future. The CIRCOR dataset contains over 6000 recordings of different patients with expert labels to define the presence of murmurs. The multiple auscultatory points and coexistence of tabular data makes this a complicated multivariable problem using different modalities, and thus well suited to multimodal AI. The application of deep learning methods including artificial neural networks will be explored. Data analysis can be conducted in python or R. Basic skills in computer coding is beneficial, but this project is well suited for someone keen to gain extra experience in these languages or for those who are keen to understand more about appropriate real world applicatoins. A keen desire to learn will be complemented by guidance through tutorials and weekly support sessions. 

Aims and Objectives
Carry out preprocessing of a rich audio data set of heart sounds

Assess different methods of feature extraction on the data set of heart sounds

Perform segmentation of heart sounds to assess the variability of individual heart sounds within patients

Conduct categorisation of heart sounds within patients using a variety of machine learning categorisation algorithms

Apply multimodal artificial intelligence to categorise the audio data of heart sounds alongside tabular data 

Workplan
Basic data preprocessing has already been conducted, allowing the student to make progress early in the project. Different techniques of feature extraction will be examined in the first two weeks. Categorisation tasks will then proceed for three weeks, with one week to address data visualisation and broad explainability. The final two weeks of the project will be spent collating an abstract and poster for presentation or publication.

Skills needed
Knowledge of machine learning in python or R would be beneficial but can be at a low level. A keen desire to learn and apply these techniques will suffice, as the student will be directed towards appropriate resources and given regular update meetings to help direct learning and application.