MPBE Summer Studentship Opportunities

Each year the Department of Medical Physics and Biomedical Engineering offers a number of summer placement opportunities for undergraduate students. These placements are an excellent way for students to gain project and research experience during the summer months.

This year, we have two types of studentships available: Research Support and Teaching and Learning Support.

Applicants are welcome to apply for 2-3 projects from either category. 

These posts are open to all undergraduate students, including those in their final year and iBSc students. Projects will start in early July 2025.

Research Support Summer Studentship

These placements are intended to allow undergraduate students to undertake paid work during the summer with academics. The key aims of the studentships are to give students the opportunity to work on a collaborative research project, gain experience developing and writing a research proposal and presenting their research to their peers.

Available projects:

3D printed patient-specific dental stents for auto-transplantation: AI-assisted automation and retrospective review of efficacy 

Supervisor: Dr Kirill Aristovich 

Project Aim: 

1. To evaluate the procedure efficiency: analyzing the existing anonymized patient data, specifically the surgery time and procedure outcomes, and compare this to pre-implementation data where the ‘fit trialing’ procedure without stents was used.    
2. To improve the segmentation algorithm which is used in clinical 3D printed stent construction process. 

Project Summary: 

Auto-transplantation (AT) is the surgical repositioning of a donor tooth within the same individual. It requires precise surgical, orthodontic, and restorative techniques. If successful, the treatment can be an excellent alternative to traditional methods with the maintenance and continuation of alveolar bone growth. Preservation of soft tissues around the donor tooth’s root is critical, influenced by the length of extra-alveolar time and handling during surgery. Using an exact 3D printed replica of the donor tooth that incorporates the Hertwig epithelial root sheath (HERS) and periodontal ligament (PDL) eliminates the need for ‘fit trialling’, increases surgical precision and efficiency, and minimises tissue damage. The existing process of precise 3D-printed stent creation involves the manual segmentation step, which is inefficient. The aim of the project is to investigate available automated segmentation tools, implement the most robust for this clinical setting and test it using a variety of existing clinical CT images.    

Tooth segmentation and analysis play a crucial role in dental research and clinical applications, particularly in diagnosis, treatment planning, and orthodontic procedures. Traditional methods for tooth segmentation rely heavily on manual annotation by clinicians, which is time-consuming, labor-intensive, and prone to variability. Recent advancements in artificial intelligence (AI) and deep learning, particularly in medical imaging, have provided promising solutions for automating tooth segmentation with high accuracy and efficiency. 

This project aims to develop a robust and automated AI-based segmentation model using U-Net and other deep learning architectures tailored for dental CBCT (Cone Beam Computed Tomography) or intraoral images. By leveraging a well-curated dataset of dental images, the model will be trained and validated to achieve precise segmentation of individual teeth, including complex cases such as overlapping or misaligned teeth. This approach not only enhances segmentation accuracy but also improves reproducibility, reducing reliance on expert annotation. 

Work Plan:

1. Data Collection and Retrospective Analysis (Weeks 1-2). To understand the impact of the new service and how it compares to the outdated procedures 
2. Improvement of the segmentation tool, 3D Modeling and Stent Design (Week 3-4).  
3. 3D Printing of stents and physical evaluation (Week 5-6). 
4. Clinical Feasibility and Data Comparison(Week 6-7) 
5. Writing of a report, and potentially a scientific publication (Week 8), followed by the adoption of the new routine by the hospital staff

Applications of imaging in dentistry 

Supervisor: Prof Adam Gibson

Project Aim: There appears to be a gap where the latest applications of medical imaging have not been applied to dentistry. The aim of this project is to conduct a literature review and explore this under-researched area, and identify future research potential. 

Project Summary: Dental science appears to lag behind other areas of medicine in the uptake of the latest medical imaging techniques. In this project, the student will review the literature and meet with research scientists working in dentistry to establish the current state-of-the-art for both research imaging and clinical imaging. They will identify current unmet needs in both research and clinical dentistry and explore how imaging is currently being used to meet these needs, and where the future research gaps lie. 

The output of the project will be a literature review and annotated database of research papers which will form the basis of future undergraduate, MSc and PhD projects, and potentially grant applications. 

This is not suitable to offer as an undergraduate or MSc project in itself as there is no intention to collect or analyse data; however, it is expected to underpin future projects. 

Work Plan: 

1. Week 1: Preparation. Student to learn what is a literature review, test software that will assist with literature review, prepare database and begin to scope out the project 
2. Week 2-4: Search literature, keeping a record of all papers found with notes on their content and quality 
3. Week 5: continue literature search, and visit Institute of Dentistry, Queen Mary University of London 
4. Week 6: continue literature search, and visit Eastman Dental Institute 
5. Week 7-8: Write report, summarising literature and research gaps 

Can entrainment of oscillatory molecular signalling pathways explain the efficacy of “oscillatory radiotherapy”? 

Supervisor: Dr Jamie Dean

Project Aim: Reveal the most likely molecular signalling pathways underlying the effectiveness of oscillatory radiotherapy, for future experimental validation.  

Project Summary: Glioblastoma is an incurable brain tumour with a median survival of 15 months. Radiotherapy, a standard-of-care treatment for glioblastoma, is typically administered once daily over several weeks. Recent research demonstrates that delivering multiple doses per day, separated by 3-hour intervals (“oscillatory radiotherapy”), significantly improves survival in a glioblastoma mouse model compared to standard schedules. However, the mechanisms underlying this effect remain unclear, making it difficult to identify patients likely to benefit or to design drug-radiotherapy combinations to further improve survival. 

Notably, oscillatory radiotherapy is effective only with a 3-hour interval, not 1-hour or 6-hour gaps, suggesting a frequency-dependent mechanism. This observation points to the involvement of an oscillatory molecular signalling pathway with a resonant frequency of 3 hours. Oscillatory stimuli can trigger resonance in such pathways, influencing cell behaviour. Several molecules critical to the radiotherapy response of glioblastoma exhibit oscillations with a period close to 3 hours, making them potential candidates for mediating this effect. 

In this project you will develop mathematical models of these oscillatory signalling pathways to simulate their dynamics in response to different radiotherapy schedules. In doing so you will reveal the most likely molecular signalling pathways underlying the effectiveness of oscillatory radiotherapy, which will be experimentally validated in the future. These findings could guide the development and optimisation of novel glioblastoma treatment strategies. 

Work Plan: 

1. Week 1: Background reading of papers to familiarise the student with the context of the work; discussions with supervisor; meeting members of the Computational Radiation Biology Lab. 
2. Weeks 2 – 3: Develop model of p53 dynamics in response to radiotherapy and simulate dynamics in response to different radiotherapy administration schedules. 
3. Weeks 4 – 5: Develop model of NFkB dynamics in response to radiotherapy and simulate dynamics in response to different radiotherapy administration schedules. 
4. Weeks 6 – 7: Analyse the properties of a molecular network necessary to generate “band pass filter”-like behaviour. 
5. Week 8: Report write-up 

Dynamic Gamma: A novel approach to simulating radiotherapy emission patterns 

Supervisor: Dr Rob Moss and Sergio Lopez Martinez 

Project Aim: To develop a cost-effective system able to replicate complex gamma emission distributions by moving a gamma source, enabling accurate detector testing and calibration in the early stages of development without the need for expensive clinical setups. 

Project Summary: The goal of this project is to develop a cost-effective and versatile system capable of replicating gamma emission distributions such as the ones observed in Boron Neutron Capture Therapy (BNCT) or Proton Beam Therapy (PBT) treatments. These radiotherapy techniques produce complex gamma radiation patterns, and their identification is key for treatment verification and detector calibration. However, replicating these distributions in a controlled setting typically requires expensive clinical setups, including actual treatment beams, which are expensive and inaccessible to many researchers and institutions. 

This project proposes a novel and practical solution: using a moving gamma source to simulate these complex profiles in the detectors. By understanding the spatial distribution of gamma emissions from these treatments, we will design a system that translates these patterns into speed profiles for a gamma source mounted on a precision rail system. By varying the speed and position of the gamma source at different points along the rail, the system will create dynamic exposure patterns that accurately replicate the desired gamma distributions in a controlled environment. 

This approach aims to enable researchers to test and calibrate detectors, optimize imaging systems, and conduct experiments without the need for costly infrastructure. 

Work Plan:

Along the 8-week project, the student will work in the four key stages of developing this technique: 

1. Analyzing a model of the gamma emission profiles of BNCT and PBT treatments. 
2. Developing an algorithm to convert these profiles into speed and exposure parameters for a moving gamma source. 
3. Constructing and validating a prototype system capable of accurately reproducing the target distributions. 
4. Testing the effect with a detector, comparing the result to a model. 

Eliciting homosynaptic depression in upper limb posterior root reflexes using transcutaneous spinal cord stimulation

Supervisor: Dr Lynsey Duffel and Dr Sarah Massey

Project Aim: To investigate whether posterior-root reflexes elicited using transcutaneous spinal cord stimulation can be used as an equivalent diagnostic test to the H-reflex. 

Project Summary: The Hoffman (H)-reflex is a standard diagnostic test of neurological function, which involves electrical stimulation of a peripheral nerve while measuring the elicited responses in a relevant muscle using electromyography. In the upper limbs, the H-Reflex can only be reliably elicited during a sustained muscle contraction; therefore, in people with upper limb paralysis due to a spinal cord injury or stroke, upper limb H-reflexes cannot always be elicited, which limits the diagnostic tests available to this population.  

In this project, we are investigating whether posterior-root reflexes (elicited using electrical stimulation applied close to the spinal cord) can be used as an equivalent to the H-reflex without requiring a sustained muscle contraction. We will apply pairs of stimulation pulses to the afferent nerve fibres at the peripheral nerve in the upper limb (H-Reflex) or close to the spinal cord (posterior-root reflex) and the amplitude of the responses will be compared. The extent to which first pulse causes a reduction in the amplitude of the response to the second pulse, depends on the interval between the pulses (homosynaptic depression). We will measure if the level of homosynaptic depression is comparable between H-reflexes and posterior-root reflexes. This experiment has already been performed on 15 participants, so the student would be adding to this data and performing further statistical analysis.  
 
This project would also work towards publication, which the student may lead if they would like to. Otherwise, the student would be named as an author upon publication after the project. 

Work Plan: Ethical approval is already in place for this research study. In the first week of the studentship, the student will be appropriately trained on the placement of electromyography and stimulation electrodes, the equipment used and good clinical and ethical practices when running a research study with human participants.  

In the following 2-6 weeks, the student will recruit participants and gain informed consent, before running the study with up to 8 more participants under supervision. Throughout the studentship, they will analyse the collected data and run appropriate statistical analyses on the final dataset. 

If there is time left, the student will prepare the data and manuscript for the study to be published. They will be invited to continue this work after the studentship has ended and will be a named author in the eventual publication. 

Miniature optical ultrasound for external imaging 

Supervisor: Dr Erwin Alles and the MISI team (Nyma Nassren, Shaoyan Zhang, Semyon Bodian)

Project Aim: Adapt an existing fibre-optic imaging system to achieve the first-on-human external ultrasound imaging with a miniature (<2 mm) probe. 

Project Summary: Ultrasound imaging is a highly versatile modality, offering real-time, depth-resolved imaging in a compact and affordable package. However, current imaging probes are fist-sized and need to be manually kept in contact with the skin, which means the operator does not have both hands available to deliver treatment. In addition, the probe size restricts the scenarios where ultrasound can be applied and prevents use in confined spaces such as the bore of MRI scanners.

Recently, we have developed optical ultrasound imaging probes that use light delivered through fibre-optics to both generate and detect acoustic waves. The resulting imaging probes are highly miniature (measuring less than 2 mm in diameter), and yet capable of video-rate 2D imaging. Whilst originally developed for use in interventional (minimally invasive) surgery, in this summer studentship project you will explore adapting this fibre-optic ultrasound technology for external imaging applications.

In close collaboration with our research team, you will explore how such probes can either be affixed to the skin directly to enable hands-free imaging, or alternatively be embedded within surgical gloves to substantially reduce the impact ultrasound imaging probes have on clinical workflows. If successful, this project could culminate in a first-on-human demonstration of miniature fibre-optic external ultrasound imaging. You will be joining an active research team possessing world-leading expertise in fibre-optic ultrasound imaging, will be supported by experienced supervisors, and have access to (and will be trained in the use of) state-of-the-art experimental and rapid prototyping facilities.

Some experience in measurement and instrumentation and wet lab work would be beneficial, as would some familiarity with ultrasound imaging. However, training will be provided. 

Work Plan: 

This project will involve several aspects that require development or investigation; the plan below is just one of many possible ways to tackle this: 

• Week 1: Shadow the researchers and complete safety and equipment inductions (workshop and labs) 
• Week 2: Bench-top imaging with fibre-optic ultrasound probes to familiarise yourself with the technology (ultrasound lab) 
• Week 3: Investigate ways of achieving reliable acoustic coupling for consistent and effective imaging without requiring submersion in water (wet lab) 
• Week 4+5: Investigate methods for affixing the fibre-optic probes to human skin and/or to surgical gloves (wet lab) 
• Week 6: Identify appropriate anatomical sites for fibre-optic imaging (literature + ultrasound lab) and recruit volunteers for an imaging study (research ethics are already approved) 
• Week 7: Perform volunteer imaging studies with both the fibre-optic and conventional clinical ultrasound systems and compare performance 
• Week 8: Document findings and developed approaches 

Should imaging performance prove inadequate for on-human demonstration, weeks 6+7 will instead be used to perform detailed imaging studies on tissue-mimicking phantoms and other imaging targets. 

Optimising phase-based magnetic resonance electrical properties tomography (EPT) for multi-parametric mapping (MPM) MRI acquisitions in the brain

Supervisor: Dr Jierong Luo

Project Aim: This project aims to optimise phase-based EPT to reconstruct brain tissue conductivity from clinical MPM MRI acquisitions in the brain. The goal is to establish methods to broaden the clinical applicability of EPT.  

Project Summary: Phase-based magnetic resonance electrical properties tomography (MREPT) is a non-invasive technique to measure tissue electrical conductivity at relatively high spatial resolution without injecting any current. At clinical magnetic field strengths, tissue conductivity is mostly determined by tissue ion concentration, in particular, sodium ions. By measuring pathological tissue conductivity changes, EPT has shown promise in detecting brain and breast lesions, tumours, stroke, and aiding diagnosis of Alzheimer’s disease. 

In EPT, a rapidly developing MRI field, tissue conductivity maps are reconstructed from the measured MRI transceive phase using a technique based on the truncated Helmholtz equation, commonly by fitting the spatial phase to a 3D spatial parabola in a kernel. Most EPT studies have used conventional “gold-standard” MRI pulse sequences dedicated for electromagnetic tissue properties mapping to generate an optimal transceive phase for EPT, which limits the potential clinical applications of EPT techniques. Recently, our research group have developed phase-based EPT from multi-modality MRI using multi-echo gradient-echo (ME-GRE) acquisition pulse sequences in the brain, aiming to broaden the clinical applicability of EPT. 

Multi-parametric mapping (MPM) MRI is a widely used multi-contrast quantitative MRI technique. MPM uses ME-GRE sequences, enabling phase-based EPT from MPM MRI although this has not been attempted previously as most MPM studies do not save the phase data. This project aims to optimise phase-based EPT to reconstruct brain tissue conductivity from clinical MPM MRI acquisitions where the phase data has been saved. We will apply the optimised MPM-EPT methods to MPM MRI acquisitions in diseases such as Parkinson’s. The overall goal is to establish methods to broaden the clinical applicability of EPT. 

Experience in MATLAB is desirable. 

Work Plan: 

• Weeks 1-2. You will familiarize yourself with the existing EPT reconstruction techniques in the literature and the EPT algorithms developed by our research group. You will choose a suitable method for phase-based EPT for MPM MRI. 
• Weeks 3. You will test the chosen EPT method on synthetic brain data (from the recent EPT reconstruction challenge) and learn to generate synthetic brain data that exactly mimic the in-vivo MPM MRI by adding appropriate noise etc. 
• Weeks 4-6. You will test and optimise the EPT reconstruction algorithm on the synthetic MPM brain data. 
• Weeks 7-8. You will apply the optimised EPT method on in-vivo MPM MRI data acquired previously in a clinical study. 

Tele-operated robotic device for MRI-guided minimally-invasive surgery

Supervisor: Dr Ziyan Guo

Project Aim: To design and test a novel robotic manipulator capable of operating inside a magnetic resonance imaging (MRI) array and maintaining minimal electromagnetic interference (EMI).

Project Summary: Magnetic Resonance Imaging (MRI) is a widely used imaging technique that provides high-contrast visualisation of soft tissues without exposing patients to ionising radiation. Its ability to guide intraoperative procedures, such as brain surgery and cardiovascular interventions, makes it highly valuable in clinical settings. However, two key challenges limit its routine use for surgical guidance: the presence of a strong magnetic field and the restricted space within the MRI bore. These constraints prevent the use of conventional robotic surgical tools, as ferromagnetic materials are strictly prohibited in the MRI environment. As a result, patients often need to be moved in and out of the scanner for imaging updates and surgical interventions, making procedures more complex and time-consuming. 

At present, very few MR-safe robotic systems are designed to function within the confined space of an MRI scanner without compromising imaging quality. This project aims to develop a teleoperated robotic manipulator specifically for MRI-guided minimally invasive procedures. The system will be compact enough to fit within the MRI imaging array while maintaining compatibility with the imaging environment, ensuring minimal electromagnetic interference. 

The project will involve computer-aided design (CAD) modelling, 3D printing, and positional control implementation using MATLAB. 

Work Plan: 

Transfer learning for MRI quantitative susceptibility mapping in the head and neck 

Supervisor: Dr Matthew Cherukara

Project Aim: To develop and test a deep learning model to optimize MRI quantitative susceptibility mapping in the head and neck, by applying transfer learning techniques to an existing model that was trained on brain data, using synthetic and in-vivo data. 

Project Summary: Quantitative susceptibility mapping (QSM) is an MRI technique that uses the magnitude and phase components of the complex MRI signal to probe the magnetic properties of tissue. Susceptibility is clinically useful as it is closely related to tissue composition, such as iron deposition, myelin content and blood oxygenation, which change early in disease such as brain neurodegeneration or cancer in the body.  

Recovering distributions of magnetic susceptibility from MRI-measured maps of magnetic field requires solving an ill-posed inverse problem. Recently, several deep learning methods have been proposed which estimate a susceptibility distribution from the field map. These methods typically follow a 3D U-Net architecture, and are trained on either synthetic data or in-vivo data from healthy participants. 

QSM is most commonly applied to brain imaging, and to date, almost all published deep learning models for QSM have been trained exclusively on brain data. Recently, however, our research group has investigated using QSM in other anatomical regions, including the head and neck. Given the broader range of susceptibility values outside the brain, as well as other differences with head and neck MRI such as the presence of fat and other tissues, it is not optimal to apply a model trained on brain data to estimating susceptibility in other regions. 

The aim of the project is to improve QSM in the head-and-neck by applying transfer learning techniques to optimize an existing deep learning model (trained on brain data) for QSM in the head-and-neck region. This will be done using synthetic MRI data and validated on head-and-neck data from healthy controls and head-and-neck cancer patients acquired as part of an ongoing clinical trial. 

Experience in MATLAB or Python, as well as with machine learning, is desirable. 

Work Plan: 

• Weeks 1-2. You will familiarize yourself with the basics of QSM reconstruction and the currently available deep learning QSM models for the brain. You will identify and implement a suitable pre-trained brain QSM model to use as a starting point for the transfer learning process. 
• Weeks 3-4. You will use the simple forward model for QSM (https://doi.org/10.1002/cmr.b.20034), together with data augmentation by deformations and other techniques, to generate synthetic head and neck QSM data for retraining the deep-learning brain QSM model. You will review the literature to learn about and select appropriate transfer-learning techniques and implementations. 
• Weeks 5-6. You will use your synthetic QSM data and implement a selected transfer-learning technique to retrain the deep learning model for head-and-neck QSM, testing it on previously acquired in-vivo head-and-neck MRI data from healthy participants (https://doi.org/10.1002/mrm.28377). 
• Weeks 7-8. You will apply the retrained model to estimate QSMs from head-and-neck cancer patient data, comparing the results with those from state-of-the-art (non-deep-learning) QSM reconstruction methods we have developed. 

Understanding the pathways determining evolution of childhood leukaemia

Supervisor: Prof Ben Hall

Project Aim: Extend new software to infer networks controlling cancer development 

Project Summary: B cell precursor acute lymphoblastic leukaemia (BCP-ALL) is a leading cause of paediatric cancer deaths, and one example of a cancer that arises through several mutational events. At present, 20% of children with BCP-ALL will die either from relapsed disease or complications of treatment, and those surviving into adulthood may suffer both long-term secondary medical complications and impaired life-chances because of the disruption to their education.

A key driver of BCP-ALL is the ETV6-RUNX1 fusion in utero, resulting in cells entering a pre-leukaemic state where targets of RUNX1 are repressed. Less clear is the mechanism by which secondary and tertiary hits determine the development of the cancer. Whilst there is a list of commonly altered genes, predicting how mutations both drive and limit carcinogenesis is not possible, and experimental verification of different pathways is not tractable due to the combinatorial explosion of different mutation combinations. 

Figure 1. Illustrative diagram showing branching outcomes for different orderings of four gene mutations. The naïve cell is mutated four times, in different orders all eventually leading to a population with the same genotype (left). Some orders will however not allow carcinogenesis (right) - this is represented here as green nodes (compatible with carcinogenesis), or blue nodes (incompatible).

This is either due to evolutionary disadvantage or the impact of downstream effectors when activated in specific orders (such as epigenetic switches). As more genes become mutated, the number of possible gene orderings rises with the factorial of the number of genes. 

In this project you will develop software and modelling tools to automatically search the wider literature and develop gene regulatory networks that allow us to simulate the impact of different mutation orders. Working from existing code, you will add extensions to the code that allow us to efficiently search downstream effectors and identify the underlying mechanism of cancer progression and evolution. 

Work Plan:

Working from prototype code from the Hall group, you will develop and test new functions for inferring networks from large databases 

• Week 1-2: Install tools, familiarise yourself with code/literature 
• Week 3-6: Write & test code 
• Week 6-8: Finalise code, develop report on work 

Teaching and Learning Support Summer Studentship

These placements are intended to allow undergraduate students to undertake paid work this summer with academics to develop teaching and learning materials, content or activities for use in MPBE courses.

Available projects:

Eligibility

Open to all undergraduate students in Medical Physics and Biomedical Engineering, including iBSc and those in their final year (due to graduate this summer).

Renumeration

Each studentship is accompanied by a stipend of £3,500 for an 8-week project (pro-rated for shorter projects), typically commencing at the start of July. Training will be provided appropriate to the requirements of the project. In addition, up to £200 may be requested for materials and consumable items.

Submission Instructions

Submit your application

1. You are required to complete the above application form.
2. You are required to submit a cover letter (no longer than one page) and specify at the top which are your 2 or 3 projects in order of preference. In your cover letter, you should address how you meet the required attributes, your prior experience in research and/or teaching and why you are interested in the chosen project/s.
3. You must also submit a one-page CV.

DEADLINE: Sunday 6 April 2025, 23:59

If you have any questions, please email medphys.teaching@ucl.ac.uk