MRes Bioengineering students work on their research project throughout the year.   You can apply for one of the projects listed below, or contact your preferred supervisor to discuss a different project

You must name at least one project or potential supervisor in your personal statement when you apply.

Applications will be considered in three rounds. We encourage you to apply in Round 1 or 2 for the best chance to be considered for your preferred project. If you apply in round 3, please consider including a second or third choice project in your application, as some projects may already have been allocated.

Visit our How do I apply? page for full instructions and deadlines.

Projects available for 2024-25

Dr Chiu Fan Lee

Contact details: c.lee@imperial.ac.uk

Title Description
Modelling tissue regeneration as fluid flow: Implication for wound healing If you scratch a layer of epithelial tissue, the epithelial cells will have the remarkable ability to proliferate in a coordinated manner to repopulate the scratched area, to the extent that the tissue almost regenerates to the original form. This extraordinary property of regeneration is far from fully understood. Recent advances have revealed that the cells at the edge of the wound are not the only active players in the healing process, rather, cells deep in the bulk exhibit active motility forces [1], and large scale swirling patterns [2]. The highly dynamic nature of tissue under regeneration suggests the possibility of viewing wound healing as a collective movement of a fluid under self-generated stress (due to cell motility) and self-generated pressure (due to cell division). This perspective has recently been investigated using computer simulations which were able to capture some of the salient features of experimental observations [3]. To render the model more realistic, a number of the model assumptions need to be connected to actual biological processes. Further, a comprehensive verification of the model requires derivable predictions that can be validated by experiments. In this project, we will first construct a cell proliferation model by incorporating elements of cell adhesion, motility and proliferation common to epithelial cells. We will then employ simulation methods to study how tissue regeneration would proceed for different types of tissue damages. For instance, we will investigate how the shape of the finger-like protrusions from the edges of the wound varies with the shape of the wound. This project will lead to an in-depth understanding of the biophysical mechanisms behind wound healing, and will equip the student with the computational skills to model cell proliferation in diverse contexts.References1. Trepat, X., et al., Physical forces during collective cell migration. Nature Physics, 2009. 5(6): p. 426-430.2. Angelini, T., et al., Glass-like dynamics of collective cell migration. Proceedings of the National Academy of Sciences, 2011. 108(12): p. 4714-4719.3. Basan, M., et al., Alignment of cellular motility forces with tissue flow as a mechanism for efficient wound healing. Proceedings of the National Academy of Sciences, 2013. 110(7): p. 2452-2459.
Cytoplasmic organisation through phase separation Biological cells organise their contents in distinct compartments called organelles, typically enclosed by a lipid membrane that forms a physical barrier and controls molecular exchanges with the surrounding cytosol. Recently an intriguing class of organelles lacking a membrane is being studied intensely. Membrane-less organelles have attracted an intense interest from the biology community as they are present in many organisms from yeast to mammal cells, and are critical for multiple biological functions. For example, P granules are involved in the asymmetric division of the Caenorhabditis elegans embryo, and stress granules assemble during environmental stress and protect cytoplasmic RNA from degradation. Strong experimental evidence indicates that membrane-less organelles are assembled via liquid–liquid phase separation, a common phenomenon in everyday life responsible for example for oil drop formation in water. Under the equilibrium condition phase separation is well understood. However cells are driven away from equilibrium by multiple energy consuming processes such as ATP-driven protein phosphorylation, which can potentially affect the phase-separating behavior of membrane-less constituents. In this project, we will study how these energy-consuming processes affect the dynamics of phase separation in the context of the cell cytoplasm.
Biophysical modelling of the pathogenesis of Alzheimer's disease Many human diseases are characterised by the formation of amyloid fibrils  linear aggregates of abnormally folded proteins  among which Alzheimers disease (AD) is a prevalent and particularly morbid one that affects us all. A common signature of amyloid-related pathogenesis is the gradual replacement of healthy tissue by aggregates of amyloid fibrils (e.g., in the form of amyloid plaque (AP) in AD), resulting in the degradation of the functioning of the tissue. Amyloid fibrils are insoluble biopolymers that are robust against proteolysis, and their presence in the forms of extracellular AP and intracellular neurofibrillary tangles are the defining histopathologic features of AD. However, mounting evidence has indicated that proteins in the monomeric form and oligomeric form, instead of proteins in the fibrillar form, are predominantly responsible for cell death. This finding thus raises the conundrum of the role of amyloid fibrils in amyloid pathogenesis: Since monomers have a tendency to self-assemble into amyloid fibrils, fibril formation should be a good way to sequester monomers and oligomers in the system by locking them up into the fibrillar form. Unfortunately, evidence points to the contrary, as demonstrated by the facts that injecting fibrils into transgenic mice induces the onset of amyloid pathogenesis and the well documented cases of Creutzfeldt-Jakob disease resulting from ingesting the misfolded form of prion proteins. In other words, there is a disconnect with the histopathologic characterisation of AD and our understanding of the cell-death inducing mechanism at the molecular level. Physically, amyloid fibrilisation and AP formation are well-described by the physics of aggregation [1]. Separately, there are well-developed methods to model toxicity-induced cell death. This project thus aims to combine these two distinct fields to further our understanding of Alzheimers disease pathogenesis. [1] Hong L, Lee C F and Huang Y J 2016 Statistical Mechanics and Kinetics of Amyloid Fibrillation. To appear in Biophysics and biochemistry of protein aggregation, edited by J.-M. Yuan and H.-X. Zhou (World Scientific). E-print: arxiv:1609.01569.
Modelling bone marrow stem cell dynamics in mouse Leukemia cells can be highly motile in the bone marrow environment.  However the exact type of motion executed by a Leukemia cell, be it diffusive, sub-diffusive or super-diffusive, remains unclear.  In this project, with the help of high-resolution and high-frequency imaging data, we will perform an in-depth study of the motility pattern of Leukemia cells in the bone marrow.  We will also incorporate the complex environment into account and investigate how Leukemia cells interact with the diverse components (e.g., bone cells and blood vessels).  The knowledge generated in this project will help us understand the dynamical behaviour of Leukemia cells in the bone marrow, and may provide key insight into how blood cancer develops from few initial malignant cells.Activities1. Image analysis of the trajectories of stem cells in the bone marrow and the changes in cell shapes2. Statistical analysis of how stem cell movement is affected by the bone marrow environment3. Statistical analysis of how the stem cell interacts with the diverse cell types in the bone marrow4. Particle simulation incorporating the movement and interactions of the stem cell as well as the various cell types in the bone marrow environment5. Mathematical modelling of the system by setting a set of partial differential equations6. Numerical analysis of the mathematical model

Dr David Labonte

Contact details: d.labonte@imperial.ac.uk

Title Description
3D shape analysis via flow fields Any two objects may differ in size and in shape. Size differences are typically obvious, but shape differences can be subtle, and challenging to quantify, because typical methods involve 2D measures such as characteristic lengths. Recent advances in computational methods have enabled detailed quantitative analyses on complex 3D shapes. In this project, you will learn and deploy one approach, based on a shape atlas. This approach is applicable to any set of 3D objects, and thus has substantial generality.Generating a shape atlas requires to estimate an anatomical model (i.e. template) as the mean of a set of input shapes. Subsequently, shape variation is quantified via the calculation of the deformation required to mold this mean shape onto a shape representing an individual from the population of relevant shapes. Mathematically, this approach represents deformation between shapes as the diffeomorphic transformation of flow fields. By means of a gradient descent optimisation scheme, the method is able to produce a statistical atlas of the population of shapes, so allowing both quantitative and detailed visual comparison of shape differences across objects.You will use this approach to investigate differences in shape across workers of a colony formed by social insects - leaf-cutter ants. Workers of this important pest species differ in weight by more than two orders of magnitude, and it is commonly speculated that this size-differences goes hand in hand with a task specialisation, leading to the definition of "worker castes". However, it remains entirely unclear whether such shape differences exist. The required workflow is established in the laboratory, and µCT scans of ant workers of different sizes are available. Your main task will be to process these scans using CT-segmentation software and Blender, to then deploy the developed workflow, and produce the shape Atlas, which will enable you to answer if ant workers of different castes differ only in size or also in shape. Recommended literature:Toussaint, N., Redhead, Y., Vidal-García, M., Lo Vercio, L., Liu, W., Fisher, E. M., ... & Green, J. B. (2021). A landmark-free morphometrics pipeline for high-resolution phenotyping: application to a mouse model of Down syndrome. Development, 148(18), dev188631.

Dr Guang Yang

Contact details: g.yang@imperial.ac.uk

Title Description
Synthetic Data Enhanced Medical Image Segmentation Multi-modal structural Magnetic Resonance Image (MRI) provides additional information and has been widely utilised for brain tumour diagnosis and treatment planning. While machine and deep learning are now widely used to interpret and evaluate MRI images, most available methods are dependent on entire sets of multi-modality data that are expensive and, in some cases, difficult to obtain in real clinical circumstances.In this project, we will develop novel deep generative models for multi-modality glioma MRI synthesis in order to solve the problems of incomplete multi-modal MRI acquisitions. The proposed medical datasetGAN method, in particular, will use an encoder-decoder architecture to map the input modalities into a common feature space, from which (1) the missing target modality(-ies) can be synthesised by the decoder, and (2) the jointly performed glioma segmentation can help the synthesis task to better focus on the tumour regions. The synthesis and segmentation tasks use the same common feature space, and multi-task learning improves both. Besides tumour and non-tumour regions will be synthesised separately to disentangle the confounding issues when GAN-based models may get confused. The validation will be performed against tumour sub-regions segmentation and tumour grading/classification accuracies. The outcome of this project will lead to potential conference presentations/publications and a further PhD project.  
Super-Resolved 3D Late Gadolinium Enhanced Cardiovascular MRI Aim:To develop novel AI powered data driven and MR physics guided super-resolution and blind denoising methods for cutting-edge cardiovascular MR techniques, potentially producing step-changes in routine clinical CMR. The AI models will be validated on prospectively scanned low-resolution data with degraded image quality.Background:Cardiovascular disease (CVD) the major cause of mortality globally with an estimated 17.9 million people died from CVD in 2019, representing 32% of all global deaths. Cardiovascular MRI diagnosis enabled by artificial intelligence (AI) has a promising future. However, the limited spatial resolution and inherent noise in MR data could affect the AI assisted diagnosis and analysis. Main goal:Cardiovascular MR (CMR) is one of the major clinical tools for diagnosis, prognosis, risk stratification, treatment planning and follow-up. 3D late gadolinium enhanced (LGE) CMR plays an important role in scar tissue detection in patients with atrial fibrillation (AF). However, LGE CMR technique suffers from limited spatial resolution and low signal-to-noise ratio (SNR). Increasing the acquired spatial resolution of LGE CMR is a major challenge and generally not recommended in practice, because it is expensive and time-consuming. Thus, super-resolution (SR) based post-processing becomes a promising option to increase the spatial resolution without increasing LGE CMR data time. In this study, we will design novel data and MR physics driven deep learning methods to boost the spatial resolution of the LGE CMR images while blindly suppressing image noise by utilising recently proposed Noise2Noise models. Retrospectively acquired data will firstly be used to train a general super-resolution model and prospectively acquired data will then be incorporated to fine tune the developed model for the super-resolution of really acquired low-resolution data with lower SNR. Experimental approach:Establish AI based strategies to i) super-resolve and denoise retrospectively acquired data, ii) perform trained model(s) on prospectively acquired data, iii) transfer learning using information from prospective data, and iv) validation against clinical quantifications (e.g., atrial scar). Outcome:Developing data driven and MR physics driven super-resolution and blind denoising method(s) for LGE CMR and comparing and quantifying clinical estimations using the developed AI powered method, e.g., estimation of atrial scar percentages before and after using our super-resolution methods.
Topological Imaging-Biomarker Discovery via Graph Neural Networks This MRes project aims to explore topological features that can act as crucial histological biomarkers for estimating disease progression and adjuvant therapies or surveillance schedules. The project is set to be implemented within the TRACERx Renal cohort, aiming to decipher the morphological characteristics and the composition of the Tumor Microenvironment (TME) that are linked with aggressive tumour behavior and the development of metastatic competence in clear cell Renal Cell Carcinoma (ccRCC).
Revolutionizing iPSC Research: AI-Powered Insights from Multicentred Microscopy Imaging This project aims to develop AI-based algorithms for the segmentation and analysis of iPSC morphology in POLG-related mitochondrial diseases, enhancing understanding and treatment strategies through advanced imaging and machine learning techniques.
Accelerated Cardiac Diffusion Tensor Imaging via Graph Neural Networks Based Multi-Modality Magnetic Resonance Restoration This MRes project aims to utilize innovative graph neural networks (GNNs) for multi-modality image restoration tasks including denoising, reconstruction and super-resolution. The primary goal is to enhance the quality and accelerate the process of cardiac diffusion tensor imaging (cDTI).
Advancing Explainable AI Models for Prognosing Lung Disease and Identifying Innovative Biomarkers via Feature Disentanglement This project is dedicated to exploring a variety of techniques to unravel the intricacies of the feature space within sophisticated AI models. By isolating and utilizing these disentangled features, we aim to create prognostic models for lung diseases that are more interpretable to clinicians and researchers. This will, in turn, empower us to identify the crucial image patterns that play a pivotal role in the decisions made by AI, which will facilitate the discovery of innovative biomarkers for lung disease prognosis.
Text Promptable Pulmonary Organ and Tissue Segmentation with Self-supervised Learning and Vision-Language Models This MRes project will exploit novel large-scale self-supervised learning and vision-language models based pulmonary organ and tissue segmentation methods from Lung CT images, e.g., pulmonary airway, vessel, COVID, fibrosis, nodule, etc., which facilitates the diagnosis and prognosis of pulmonary diseases.

Dr Joseph van Batenburg-Sherwood

Contact details: jvbsherwood@imperial.ac.uk

Title Description
A new method to measure the secretion of aqueous humour in mouse eyes The pressure in the eye is regulated by the flow of aqueous humour across resistive tissues. In glaucoma, the pressure in the eye increases and expands the eye, irreversibly damaging the retinal cells. In order to better understand the way in which the pressure is altered, we need to be able to measure both the rate at which the fluid is secreted and the resistance of the tissues. We developed the world-leading technology for the measurement of tissue resistance in mouse eyes, iPerfusion, but there are presently no accurate ways to measure the rate of aqueous humour secretion. The rates of this flow are extremely low, at around 100 nl/min.We have conceived a novel method to accurately measure this flow rate by cannulating the eye with two needles, exchanging a fluorescent marker and evaluating dilution using droplet microfluidics and optical methods. By comparing the initial concentration and how the concentration reduces over time, we will be able to quantify the dilution rate, which will be equal to the rate of aqueous humour secretion. The project will use analytical methods, computer-aided design, microfluidics, optics, and possibly computational fluid dynamics to first develop an in vitro model of the system. The next step will be to build a prototype and test it on real eyes, using a known flow rate to evaluate the accuracy and repeatability of the measurement. Please get in touch if you have any questions

Dr Juan Alvaro Gallego

Contact details: juan-alvaro.gallego@imperial.ac.uk

Title Description
On the universality of neural network models of motor behaviour Artificial neural networks have emerged as powerful models in computational neuroscience: not only can they perform many laboratory tasks that we study in animals but, more intriguingly, the activity they produce during these “behaviours” is remarkably similar to that of their biological counterparts (e.g., [1,2]).
A network’s activity is influenced by a number of variables, including its architecture and how the model is trained. Thus, an important question at the intersection of neuroscience and artificial intelligence research is how different classes of models perform the same task. Previous studies have tried to understand how different neural network architectures perform a common series of cognitive tasks [3], but this question remains mostly unexplored in the case of motor tasks —motor tasks are quite different in that movements are continuous variables that need to be controlled by the brain, rather than discrete, often categorical quantities that need to be selected.
Your goal in this project is to train a variety of models to perform a complex motor task in which monkeys reached to a series of targets that were presented during movement execution [4] and compare the dynamics produced by the different types of networks. Since brains arguably evolved to control movement [5], establishing the universality –or lack of– of the solutions found by different networks performing the same motor tasks would be an important contribution to neuroscience. Moreover, identifying any similarities and differences across classes of models is crucial to understand how much they can tell us about biological brains.

[1] Sussillo et al Nature Neurosci 18:1025–33, 2015
[2] Yang et al Nature Neurosci 22:297–306, 2019
[3] Maheswaranathan, Williams et al NeurIPS 2019
[4] Glaser et al. Nature Communications 9(1):1–14, 2018
[5] Wolpert et al Trends Cogn Sci 5:487–94, 2001
Understanding the neural basis for different types of learning Animals including humans can rapidly adjust their behaviour in the presence of perturbations. These includes counteracting mechanical perturbations to the limb [1] (such as when running into water), or distorted feedback about the state of the limb [2] (such as when reaching into a pond). Yet, despite considerable effort, the neural basis for this rapid learning or motor adaptation remains largely elusive.
Recent theoretical and computational advances in understanding the coordinated activity of neural populations [3] are helping to cast light into this fascinating problem. For example, our group has shown that, in contrast to learning a new skill through substantial practice, this form of learning may not need “plastic changes” in the brain [4,5]. Moreover, while motor regions of the brain seem to be key for counteracting mechanical perturbations [4,6], learning to adapt to distorted feedback may be driven by higher brain regions [6]. In this project, you will systematically compare these two forms of adaptation by multi-area neural population recordings from monkeys performing these two classic experiments on different days (data from [3]). These analyses will include understanding changes in neural dynamics as animals plan and execute these movements, as well as a comparison between motor regions of the brain. Your results will shed light in the neural basis for rapid learning and also inform the field of brain-computer interfaces, which necessitates more flexible “decoding algorithms”.

[1] Shadmehr and Mussa-Ivaldi. J Neurosci 1990
[2] Krakauer et al. Nature Neurosci 1999
[3] Gallego et al. Neuron 2017
[4] Perich, Gallego & Miller. Neuron 2018
[5] Feulner et al. Nature Communications 2022
[6] Sun, O’Shea et al Nature 2022

Dr Jun Ishihara

Contact details: j.ishihara@imperial.ac.uk

Title Description
Enhancing the effect of COVID-19 vaccine by prolonged skin tissue retention of antigen Lab-based project for those with synthetic biology or biomaterials experience Our new laboratory focuses on a protein engineering based translational immunology for drug development. We have recently developed a new method to engineer protein that can retain at the skin tissue through fusing extracellular matrix binding domain to the protein. Skin tissue has many antigen-presenting cells, including dendritic cells. In theory, the prolonged tissue retention of the COVID-19 antigen in the skin tissue would increase the vaccination effect. We will design the protein, produce, and purify. We will characterise the protein and compare with non-tissue retaining COVID-19 antigen. We take protein engineering approaches without biomaterials to make advanced vaccines for the future use. To be clear, you will not be exposed to the virus. We want to offer 1 lab-based project for dedicated students who can do full days in the lab to work in this area.
Engineering anti-cancer cytokine for effective cancer immunotherapy Lab-based project for those with synthetic biology or biomaterials experience Our new laboratory focuses on protein engineering based translational immunology for drug development. We have recently developed a new method to engineer protein that can deliver cancer drugs to the cancer tissue. We will try producing cancer targeted cytokines. We will design the protein, produce, and purify. We will characterise the protein. We take protein engineering approaches for drug development. We want to offer 1 lab-based project for dedicated students who can do full days in the lab to work in this area.
Engineering of serum albumin fusion proteins with enhanced therapeutic potential to treat cytokine release syndrome  The lymph nodes are extremely important target for autoimmune disease development. We hypothesise that lymph node delivery of anti-inflammatory cytokines may be a way to find a new therapeutics in these diseases. We use the protein engineering approach to develop new molecules to suppress inflammation.
Investigating tumour-immune cell interactions for cancer immunotherapy using capillary-on-chip models  

Dr Laki Pantazis

Contact details: p.pantazis@imperial.ac.uk

Title Description

Generation of ChemiGenEPi1.0-Biosensor

The Pantazis Lab recently developed GenEPi, the first genetically-encoded green fluorescent reporter for mechanical stimuli based on Piezo1. This ion channel is a pivotal molecule in cellular mechanosensation across biological scales, and its discoverers were awarded the 2021 Nobel Prize in Physiology or Medicine.

The project aims to design, develop and pilot test the first generation of quantitative chemigenetic biosensors for mechanical forces based on pressure-sensitive ion channels, such as the above mentioned Piezo1. This collaborative project (ChemiGenEPi1.0 Biosensor) brings together the expertise of the Howard Hughes Medical Institute/Janelia Research Campus and Imperial College London to advance the next generation of quantitative biosensors, focusing on cellular mechanosensation and investigating how cells sense and interpret mechanical forces across biological scales. The collaboration capitalises on the distinctive strengths of two world-leading laboratories. Dr Schreiter’s lab contributes unique expertise in developing novel chemigenetic biosensors, while Dr Pantazis’ lab boasts a proven track record in advanced optical precision imaging, notably introducing the first genetically-encoded biosensor for the pressure-sensitive ion channel Piezo1. 

This initiative fosters innovation with broad scientific impact and potential for translational research outputs and new drug screening pipelines.he experimental steps will involve designing strategies for generating such sensors followed by molecular cloning of GenEPi variants. Eventually, the variants will be characterised in cells. 

The interested student will be trained extensively in standard molecular and cellular lab practice. In addition, the student will be exposed to sophisticated and advanced optical imaging techniques present in the Imperial College London and Leica Microssytems Imaging Hub used, a UK-unique centre of excellence in optical imaging that is directed by Dr Pantazis.

Dr Lance Rane

Contact details: lance.rane14@imperial.ac.uk

Title Description
Data synthesis for wearables-based motion systems: Musculoskeletal modelling At our spinout Biomex (biomexlabs.co.uk) we use motion data in real time to infer gait characteristics for the guidance of therapy. However, data are often much easier to obtain from healthy controls than from patients. Can techniques of data synthesis be used to make up the shortfall?One issue with data from body-worn sensors is that data augmentation - a crucial technique for successful application of machine learning - can be very difficult. How would the data change if the subject were a little taller, or heavier, or walked a little faster? These questions are difficult because the relationships in each case are complex and non-linear. They are best answered by the use of musculoskeletal models - full mathematical descriptions of the human body in motion. This project aims to use data from IMUs to feed existing musculoskeletal models to produce simulations of the original human motion. In model-space, perturbations may be applied - eg to walking speed - and the simulation re-run to regenerate the sensor data. This provides an infinite data generator, through a combination of tweaks to important parameters - such as speed, leg length, height etc. Final evaluation of performance will be made by using the resulting data to train existing neural network classifiers, and testing their performance against baseline.
Data synthesis for wearables-based motion systems: Generative modelling At our spinout Biomex (biomexlabs.co.uk) we use motion data in real time to infer gait characteristics for the guidance of therapy. However, data are often much easier to obtain from healthy controls than from patients. Can techniques of data synthesis be used to make up the shortfall?In this project, students will use existing motion datasets from healthy controls and patients. Applying advanced generative modelling techniques (variational autoencoder, GAN, diffusion modelling) they will seek to create representations of the data that disentangle pathology features. This will enable the simulation of pathology in healthy subjects, and vice-versa, providing a potent new method of performing artificial data synthesis. It could also provide new insight into the characterisation of movement disorders and musculoskeletal conditions that affect movement. Final evaluation will be made by using the newly generated data to re-train and test existing neural network classifier models.

Dr Majid Taghavi

Contact details: m.taghavi@imperial.ac.uk

Title Description
Multimodal haptic device Haptic stimulation is an innovative technology that leverages the sense of touch to convey information to users. Traditionally, this has been achieved through vibrations to deliver notifications. However, in this project, our aim is to develop a novel wearable technology capable of delivering a diverse range of subtle mechanical stimulations, including tapping and sliding, to the wearer. This device could be used for sending complex physical information to individuals with visual, hearing, or mobility impairments, providing cutaneous sensory re-education, or therapeutic guidance to the patient during an activity.
Interactive robotic skin  Human can examine structures or areas for size, shape, consistency, texture, and other characteristics simply by moving their fingers and palms over the surfaces. Dexterous movement, flexible and soft architecture, and a variety of sensing capabilities on the skin enable this palpation. In this project, we aim to advance this concept by developing a cutting-edge soft robotic material. This material will incorporate embedded electroactive artificial muscles and tactile sensors, creating an interactive composite working like a robotic skin. By employing this innovative technology, it will be possible to scan the body and accurately detect variations in stiffness, opening up new possibilities for medical diagnostics and interventions
Soft-rigid hybrid robotic material  Soft robotics built from highly compliant materials suggests safe and adaptable technologies for many applications including human-machine interactions (e.g. surgical robots, implants, and wearables). In practical uses, it is usually needed to integrate dissimilar materials, such as soft with rigid and electrodes with non-conductors, in the architecture of soft robots for applying mechanical support and electrical connections. Inspired by nature, where dissimilar materials are found with progressive compositional changing (e.g. muscle-tendon-bone), in this research we will develop a novel robotic material with a smooth transition from soft to rigid, embedding a unique capability of interacting with the environment. This soft-rigid hybrid muscle will benefit from a reliable connection to rigid bodies, withstanding diverse mechanical deformation and robust electrical connections.
Variable stiffness soft robotic material Variable stiffness materials provide a core technology for wearables and assistive robotics. They are a class of smart materials with the capability of stiffness change under external stimulation. Within this project, we will develop different types of electroactive ionic elastomers for variable stiffness applications. The intrinsic mechanical properties of the ionic elastomer and their response to the applied electric field will also be characterised. We will then design a proof of concept robotic component to demonstrate its use in wearables and implants.
Nylon thread artificial muscle Coiled polymer artificial muscles offer a new type of thermo-active soft actuation technology with high potential to be used in wearables and assistive robotics. A twisted and coiled sewing thread such as silver-coated Nylon can generate large stroke and high-power density when actuated by ohmic heating. They however suffer from the lack of high bandwidth controllability because of the slow cooling process. In this project, we aim to tackle this challenge by implementing a multi-physical approach and introducing intelligent structures to extend their applicability in interactive textile technologies.  
Social soft robot Social robots are robots that interact and communicate with humans, usually to provide some kind of social or service function. They are becoming increasingly common in settings such as healthcare, education, and the home. In this project, you will develop a new type of social robot that is made of soft material and can change its shape and characteristics using soft actuators to communicate with people. For example, the robot could change its shape to communicate different emotions, such as smiles, anger, or fear. 

Dr Pedro Ballester

Contact details: p.ballester@imperial.ac.uk

Ballester group page - https://ballestergroup.github.io/

Title Description
The impact of high-dimensional datasets on AI for precision oncology Introduction:Precision oncology aims at prescribing the most suitable anti-cancer treatment to a patient to ensure maximum efficacy and minimum side effect. By analysing pharmaco-omics data, we can identify individual variability in drug response by studying the molecular profiles of the patient's tumours. Furthermore, clinical features such as age, co-morbidities and the like can influence how patients respond to treatment. Artificial Intelligence (AI), and more concretely its most developed subarea Machine Learning (ML), holds tremendous promise for precision oncology. Recently published ML studies show that, at least in some cases, it is possible to predict which patients are resistant, or responsive, to treatment from their molecular and/or clinical features. However, a major challenge is the very large number of features describing each patient.About the supervisor:Dr Pedro Ballester has a 10-year track record on this research topic. His later paper on this topic was on the application of AI to patient response prediction to drug treatments from multi-omics tumour profiles (https://onlinelibrary.wiley.com/doi/10.1002/advs.202201501). His group has hence developed infrastructure around the data sources I will employ. For example, the Genomic Data Commons (GDC) from the US National Cancer Institute (NCI).Research plan:The first stage will be to master scripts to integrate clinical and omics data from the GDC. Here we already combined the clinical responses annotated with the RECIST standard as two types, namely “Responder” (including “Complete Response” and “Partial Response”) and “Non-Responder” (including “Stable Disease” and “Clinical Progressive Disease”).   Then we will select problem instances with various dimensionalities as case studies. That is, cohorts of patients with the same cancer type, whose tumours were profiled with the same omics technology and administered the same drug. Tumours were comprehensively profiled (CNA, DNA methylation, miRNA and mRNA)Lastly, we will investigate the application of ML techniques to identify the most predictive features among the hundreds to thousands of them. Among the techniques to be assessed, we will look at integrating ML algorithms with feature selection schemes such as OMC, Boruta or RFE. Particular attention will be paid to rigorously estimate classification performance, some methods for this are already implemented here: https://cran.r-project.org/web/packages/MXM/ -  (Ballester group page - https://ballestergroup.github.io/)
Optimal design of virtual screening benchmarks from in vitro screening data Introduction:Virtual screening (VS) has become an important source of small-molecule drug leads. A benchmark is needed to identify the VS method/s that will perform best prospectively for that therapeutic target. Benchmarks are also needed to find the optimal settings for the selected VS method/model. A VS benchmark is a library with two classes of molecules: those whose activity for the target is above a given threshold - actives as the positive class- and those with weaker or no activity at all - inactives as the negative class. Among them, a screened library is one whose molecules have been screened in the same centre and using same assay/s, e.g. the results of high-throughput screening (HTS) a compound library against the considered target.Unfortunately, HTS data have been used for VS benchmarking in a convenient yet unrealistic manner (e.g. generating benchmarks with much smaller chemical diversity than HTS). The question is how useful are these HTS-derived datasets as VS benchmarks with respect to the ground truth represented by the unadulterated HTS datasets.About the supervisor:Dr Pedro Ballester has over 17 years of experience in this research area. His last papers in this area have shown the potential of Artificial Intelligence (AI) for structure-based drug design:https://wires.onlinelibrary.wiley.com/doi/abs/10.1002/wcms.1478https://academic.oup.com/bib/article-abstract/22/3/bbaa095/5855396Research plan:The student will start by learning about these data types as well as existing VS benchmarks (e.g. MUV) and VS methods (e.g. USR, Smina). Then, s/he will be applying each VS method to rank the HTS-derived benchmark molecules in order to assess its performance on the associated target. This process will also be carried out for other VS methods, unadulterated HTS benchmarks and targets. The results will be employed to investigate to which extent the filters used to select molecules for a VS benchmark make it unrealistic. This is crucial for the development and selection of VS methods.About the candidate:This project is suitable for a student who is keen to learn about molecular modelling in the context of early drug design. Python programming is required. Contact: p.ballester@imperial.ac.uk -  (Ballester group page - https://ballestergroup.github.io/)
Optimal design of simulated-library benchmarks for virtual screening Introduction:Virtual screening (VS) has become an important source of small-molecule drug leads. A benchmark is needed to identify the VS method/s that will perform best prospectively for that therapeutic target. Benchmarks are also needed to find the optimal settings for the selected VS method/model. A VS benchmark is a library with two classes of molecules: those whose activity for the target is above a given threshold - actives as the positive class- and those with weaker or no activity at all - inactives as the negative class. Among them, a screened library is one whose molecules have been screened in the same centre and using same assay/s, e.g. the results of high-throughput screening (HTS) a compound library against the considered target.In the absence of publicly-available HTS data, one has to resort to define a simulated-library (SimL) benchmark. Here one gathers actives from the literature and generates decoys from each of them so that these molecules collectively simulate a screened HTS library. This is the most common type of VS benchmark, where decoys are often required to have dissimilar chemical structures to that of their active to make them even less likely to be also active (activity is a rare event and hence randomly-chosen molecules are already unlikely to be active). Furthermore, an active and their property-matched decoys have similar physico-chemical properties so that they cannot be trivially discriminated by one of these properties, as this does not happen prospectively either. One popular example of such benchmarks is DUD-E (http://dude.docking.org/).The question is how useful are these SimL benchmarks for VS with respect to the ground truth represented by the entire HTS datasets (in the targets for which these are available). We expect this study to reveal which are active-to-inactive ratios and decoy schemes best anticipate performance on HTS datasets. This will be useful to decide best SimL benchmarking in absence of HTS data.About the supervisor:Dr Pedro Ballester has over 17 years of experience in this research area. His last papers in this area have shown the potential of Artificial Intelligence (AI) for structure-based drug design:https://wires.onlinelibrary.wiley.com/doi/abs/10.1002/wcms.1478https://academic.oup.com/bib/article-abstract/22/3/bbaa095/5855396Research plan:The student will start by learning about these data types as well as existing VS benchmarks (e.g. MUV) and VS methods (e.g. USR, Smina). Then, s/he will be applying each VS method to rank the DUD-E benchmark molecules in order to assess its performance on the associated target. This process will also be carried out for other VS methods, unadulterated HTS benchmarks and targets. The results will be employed to investigate to which extent the design factors for a retrospective VS benchmark make it unrealistic. This is crucial for the development and selection of VS methods.About the candidate:This project is suitable for a student who is keen to learn about molecular modelling in the context of early drug design. Python programming is required. -  (Ballester group page - https://ballestergroup.github.io/)
What makes supervised learning hard? IntroductionSupervised learning comprises classification and regression, where the goal is to build models able to predict the class and the numerical value, respectively, with which a data instance is labelled. Each data instance is characterised by a set of features. The trained model returns the predicted label of the instance from its features. Supervised learning is a large branch of machine learning.There are many factors that make supervised learning hard: a higher number of features, a low proportion of informative features, class imbalance, feature correlation, feature noise, etc. However, the impact of these factors on a given problem is uncertain, as we do not know the functional form of the data generator (our datasets only contain noisy measurements of such process).Research planIn this project, datasets will be generated with a given functional form, so we will know the data generator (which features and how are combined to determine the label of each instance). The student will evaluate a battery of algorithms to find out which ones are most suitable for a given harness factor. Several data generators will be considered, so that we can better understand which algorithms are most suitable for a problem instance from its characteristics. We also aim at exploring the practical limits of machine learning algorithms and investigating strategies to overcome them.About the candidateThis project is suitable for a student who is keen to dive deep into machine learning concepts and algorithms. Python programming is required.Referenceshttps://link.springer.com/article/10.1007/s42979-021-00592-x -   (Ballester group page - https://ballestergroup.github.io/)

Dr Reiko Tanaka

Contact details: r.tanaka@imperial.ac.uk

Title Description
Systems biology approach for mechanistic understanding of paediatric asthma exacerbations  Asthma is the most common chronic disease of childhood, affecting up to 10% of children in Westernised societies and 200,000,000 individuals worldwide. Many factors indicate the importance of the microbiome in asthma. Asthma is rare in rural societies, and its prevalence has been increasing markedly in the developing world as populations become urbanised. Exacerbations of asthma are often precipitated by otherwise trivial viral infections. Our studies have shown that the normal human airways contain a characteristic microbiome that is altered in children and adults with the illness. Asthmatic airways contain an excess of pathogens (which may damage the airways) and also lack particular commensal species that may be necessary for normal airway functions. This project will take a systems biology approach, by combining experiments with primary bronchial epithelial cells, in silico modelling, and clinical data analysis, to elucidate the effects of the airway bacterial microbiome in asthma, and the role of epithelium barrier integrity in disease initiation and control. We already have a preliminary mathematical model that will be used to quantify the dynamic interactions among pathogen, commensals at the airway surface, the airway barrier and the immune system, preliminary data from in vitro experiments, and clinical data to be analysed. The student(s) will conduct several computational methods to identify the model structures and model parameters. 
Systems biology approach for cancer immunotherapy: Dynamical mechanisms to turn hot tumours into cold tumours Immunotherapy is a promising treatment for cancers. Several immunotherapy treatments have already achieved clinical success. For example, checkpoint inhibitors (CPIs) targeting PD1 and CTLA4 achieved up to 11% of cure rates in advanced melanoma. However, its treatment response rates remain low for a majority of cancers. This project aims to apply a systems biology approach to increase the efficacy of CPI therapies. CPI therapy achieves cancer regression by preventing exhaustion of T cells that pre-exists in the tumours but does not exhibit strong antitumour immunity if tumours have few immune cells. One way to improve the CPI efficacy is to convert immune-excluded cold tumours into immune-infiltrated hot tumours, by enhancing infiltration of tumour-specific immune cells into the tumour tissue. Ishihara et al have recently engineered a tumour-homing anti-tumour cytokine IL-12 (Nat Biomed Eng 2020) that turns immunologically cold tumours into hot ones, and demonstrated a high antitumor efficacy of a combination therapy using IL-12 and CPI for cold tumors. However, the actual dynamical mechanisms by which IL-12 converts cold tumours to hot tumours and how IL-12 acts in combination with CPI therapy is still elusive. This project aims to investigate the cellular mechanism of immune cells infiltration during the combined therapy. The student will develop and analyse a computational model of immune cells infiltration based on the results from in vitro and in vivo experiments, to suggest the best timing of adding CPI theraty to achieve the most anti-tumour efficacy.
Learning from noisy labels by EczemaNet Eczema is the most common form of skin disease. The eczema severity is currently assessed by trained clinical staff. However, the subjective nature of assessing these disease signs could be a source of significant inter and intra-observer variability. Our group recently developed EczemaNet, a prototype novel computer vision pipeline for automated evaluation of eczema severity from camera images (Pan, Hurault et al. 2020). EczemaNet uses a CNN that first detects areas of AD from camera images and then makes probabilistic predictions of the severity of seven clinical signs of AD. EczemaNet was trained using the camera images with eczema severity scores provided by trained clinical staff. However, the severity scores may have inter-rater variability, resulting in noisy labels. This project aims to further improve EczemaNet by achieving a better generalisation capability in the presence of noisy labels.  The student is expected to review, implement, and validate the image analysis pipeline using off-the-shelf deep-learning and image analysis techniques and software packages to improve the prediction accuracy in the presence of noisy labels.
Development of automated evaluation tool for clinical signs of atopic dermatitis using machine learning image analysis Machine learning methods and their application to image processing have shown a rapid progress over the last decades. Automated classification of melanoma malignancy has already achieved a good progress (Esteva et al. 2017, Nature), thanks to the large number of available images (129,450 images used), as well as advances in characterisation of melanoma severity. This project aims to develop an automated evaluation tool for clinical signs of atopic eczema, a chronic skin disease, by applying machine learning methods to the images of the lesional skin sites. It will enable the daily monitoring of the disease symptoms by patients by themselves, without coming to a clinic. Strong programming skills and a good understanding of statistics are required.
Automatic quantification of fungal burdens in histology images using deep neural networks Invasive fungal infections are commonly treated by antifungal drugs, which are under the growing threat of antifungal resistance. New antifungal drugs that overcome antifungal resistance are being sought actively. Effectiveness of new antifungal drugs is often tested in vitro by comparing fungal growth rates estimated from time-course optical density (OD) data of fungal burden in different treatment conditions. However, OD is an indirect measure of fungal burden, with potential bias in the evaluated growth rates and artefacts in the data such as multiple scattering. Various methods to improve bacterial growth rate estimation from OD have been proposed and subsequently been made available to the wider bacterial community. However, these methods require converting OD to cell counts in a process called calibration, which is not routinely available for fungi. 
Uncovering dynamical interactions in the altered microbiota of atopic dermatitis skin towards designing live therapeutics  Atopic dermatitis/eczema (AD) is a devastating and very common chronic skin disease affecting 15-30% of children worldwide. The ultimate aim of this project is to design live therapeutics for AD. Healthy skin is habitated by a rich, balanced diversity of microbes, which help protect our body from invading pathogens and infections. However, this balance is thrown off on AD skin with a microbiome dominated by staphylococci, primarily opportunistic pathogen, S. aureus (SA). SA release peptides (via Agr quorum sensing (QS) system) that kill competitor microbes and damage the skin barrier, exacerbating AD symptoms. Other friendly staphylococci, such as S. epidermidis (SE) and S. hominis (SH), are an integral part of the healthy skin microbiome and appear to co-exist with SA on AD skin, although they are also armed through their own Agr QS systems. How does SA win the battle against SE and SH? Can we find a way to stop SA winning the battle and improve AD symptoms? This project aims to answer these two questions. The student will develop a simple mathematical model of the interspecies interactions [2], and fit the model to the experimental data to unveil the key interactions between SA, SE and SH.

Dr Rodrigo Ledesma

Contact details: r.ledesma-amaro@imperial.ac.uk

Title Description
Bioengineering yeast for the production of fuels, vitamins, antioxidants and colorants This project aims at hacking the metabolism of a yeast cell in order to produce commercially relevant compounds, specifically carotenoids. Carotenoids are used as colorants, antioxidants and as vitamin precursors and they are normally extracted from plants in a low efficient process that makes the products very expensive. During this project, cutting edge synthetic biology techniques (such as Golden Gate DNA assembly or CRISPR-Cas9) will be used for engineering the cells in a reliable and efficient manner towards the production of the desired compounds. The development of novel, more efficient bioprocesses will help us to move towards more environmentally friendly industrial setups. Please, do not hesitate to ask Rodrigo Ledesma-Amaro (r.ledesma-amaro@imperial.ac.uk) for further information.Skills: The student will develop skills and knowledge in molecular biology, genetic engineering, bioprocesses, metabolic engineering, analytical techniques, microbiology, synthetic biology, etc. Recommended for students with interests in synthetic biology and cellular engineering, especially for those who want to develop a career either in academia or in industry.
Development of a novel CRISPR-based synthetic biology method to control metabolism  CRISPR is the genome engineering technique that has recently revolutionized synthetic biology and medicine. This project aims at creating a novel method based on CRISPR technology to achieve metabolic control. The control of metabolism is an essential bioengineering problem that applies not only to microbial cell engineering for the biotechnological production of industrially relevant compounds, such as fuels and chemicals, but also to tackle diseases related to unbalanced metabolic states. The proposed method will be developed in microbial cells (yeast) and will be applied to relevant metabolic pathways for industry and health. Please, do not hesitate to ask Rodrigo Ledesma-Amaro (r.ledesma-amaro@imperial.ac.uk) for further information.Skills: The student will develop skills and knowledge in molecular biology, genetic engineering, bioprocesses, metabolic engineering, analytical techniques, microbiology, synthetic biology, etc. Recommended for students with interests in synthetic biology and cellular engineering, especially for those who want to develop a career either in academia or in industry.
Synthetic biology and metabolic engineering for microbial biotechnology and bioengineering Microorganisms are important for both industrial bioprocesses and biomedicine (i.e. gut or skin microbiota). The lab is interesting in a wide array of organisms, from yeast (S. cerevisiea and Y. lipolytica), fungus (A. gossypii) and bacteria (E. coli and Acetobacter) to complex microbial consortia (human and industrial microbiota).The manipulation and optimization of microbial metabolic pathways are the keys to biotechnology and a bio-based economy. we are highly interested in hacking metabolism using synthetic biology tools to create new properties and enhanced behaviors in microbial cells. The engineering strategies are not only designed to produce new high-value products or higher amount of pre-existing products but also to facilitate the downstream and upstream parts of the bioprocesses.
Deciphering the codon usage code and its role in metabolism  with applications in synthetic biology The DNA codes for all the heritable information required to form life. Due to the degeneration of the genetic code, different DNAs can code for the same proteins, this is possible because several codons (groups of 3 nucleotides) can be translated into the same amino acid. This property emerges in all living systems and there are many theories that justify this mechanism. One of these theories, yet to be explored, is that the codons represent an additional level of regulation. This project will explore variations in codon usages in specific metabolic pathways or conditions in order to identify novel regulatory elements. This project could lead to important biological insights that can be used to understand diseasese and to improve synthetic biology approaches. 
Investigating Anaerobic Growth in Strict Aerobe Yarrowia lipolytica Engineering Yarrowia lipolytica to make it able to grow in anaerobic or microaerobic conditions
Design and implementation of novel synthetic circuits in yeast using CRISPR An important challenge in Synthetic Biology is the development of complex synthetic circuits to translate a mix of inputs into defined outputs given a cellular context. Many barriers have made this goal difficult to reach. First, standardized circuit modules (or building blocks) that can be easily connected and tested have been lacking. Second, synthetic circuits are not truly orthogonal to the cellular context making the performance of complex circuit designs difficult to predict. Thanks to its multiplexing and modularity abilities, CRISPR is a promising platform to design next generation circuits. Using programmable single-guiding RNAs, CRISPR allows to easily link input to output by regulation at the transcription level using a single protein, such as dCas9. This also considerably reduces the genetic load of the circuit in the host cell. Using novel state-of-the-art techniques like golden-gate cloning and computational methods, design and synthesis of complex circuits can now be faster than ever before. This project will involve computational and wet lab aspects. In the computational part, mathematical methods are applied to design and predict performances of CRISPRi-based circuits. In the wet lab part, the designs are then tested experimentally.
Deciphering the codon usage code and its role in metabolism  with applications in synthetic biology The DNA codes for all the heritable information required to form life. Due to the degeneration of the genetic code, different DNAs can code for the same proteins, this is possible because several codons (groups of 3 nucleotides) can be translated into the same amino acid. This property emerges in all living systems and there are many theories that justify this mechanism. One of these theories, yet to be explored, is that the codons represent an additional level of regulation. This project will explore variations in codon usages in specific metabolic pathways or conditions in order to identify novel regulatory elements. This project could lead to important biological insights that can be used to understand diseasese and to improve synthetic biology approaches. 
Synthetic protein design for food applications Our current food systems is not sustainable and there is a clear need to produce alternative sources of proteins for human nutrition. This project aims to use synthetic biology and computational tools to explore the potential of designed proteins to become the food of the future. This project will have a main computational component that can be followed by an experimental validation if the student wants.This project can be open to more than one student with different objectives.

Dr Shlomi Haar Millo

Contact details: s.haar@imperial.ac.uk

Title Description
Behavioural Biomarkers for Deep brain stimulation Deep brain stimulation (DBS) is a routine treatment for patients with Parkinson's disease (PD) which improves their motor symptoms and as a result their function and quality of life. While DBS is an effective therapy, it is still not clear how and why it works and therefore there are many open questions as to how it can work better. This research project is part of a program that aims for a better understanding of the effects of DBS on the system level (in addition to improving symptoms) and a search for biomarkers to improve DBS delivery. In this research project, you will help to collect behavioural (movement sensor) and neural (EEG) recordings from PD patients with implanted DBS electrodes. You will analyze the behavioural data to study the effects of the DBS parameters on their body movement, and develop behavioural digital biomarkers to improve DBS delivery.
Motion tracking in the home for Dementia care The UK DRI Care Research & Technology centre, works to create new technologies that will deliver the highest quality dementia care in the home. We develop a range of devices that allow us to track a person’s behaviour and health at home, to understand an individual’s behaviour and predict when problems might arise. In this project we will explore the potential of using camera based 3D skeletal body tracking for continuous monitoring in the home. You will work on setting an of-the-self depth camera with single-board computer (e.g. Raspberry Pi) and skeletal tracking SDK as a standalone device that can work in the home and record body movement. You will then test and validate the data quality and reliability during daily life activities in our living lab. This is a domestic studio flat with passive movement sensors, providing a bridge to research ‘in the wild’. Lastly, you will work with our software engineers to integrate this data stream into our healthy home database. 
Motor control fingerprints of neurological conditions In the UK DRI Care Research and Technology Centre we are collecting a multimodal dataset of movement, brain activity, physiology and cognition from a diverse cohort of participants with neurological conditions (including dementia, Parkinson’s disease, motor neurone disease, and traumatic brain injury). This is done in our Living Lab which is a domestic studio flat with passive movement sensors, providing a bridge to research ‘in the wild’. In this project, you will support the data collection effort (and will get to work with the different neurological patient groups), and lead a sub-project of the data analysis. This will look into defining, coding, and analysing specific features of the subjects’ movements which are motor control fingerprints of neurological conditions. This will be done both with a data-driven ML approach and in a clinically guided hypothesis-driven approach. 
Real-World Motor Learning in Embodied Virtual Reality A key challenge in neuroscience, neurology and neurorehabilitation is to measure and train motor control and learning in free-behaving real-life tasks. We recently demonstrated the feasibility of studying real-world neuroscience using wearable technologies and data-driven approaches to uncover neural mechanisms of learning. We also developed an embodied virtual-reality (EVR) setup, which allows us to study motor control and learning in a controlled-real-world learning environment. In this project, you will use our EVR setup to induce perturbations aimed to manipulate motor learning mechanisms. You will record subjects’ movement (with body sensor networks) and brain activity (with mobile EEG) while performing a motor learning task with visual perturbations in the VR. This will force subjects to use different learning mechanisms and in your analysis, you will work to map the behavioural changes induced by the perturbations and changes in the brain activity.

Dr Sophie Morse

Contact details: sophie.morse11@imperial.ac.uk

Title Description
Do immune cells enter the brain when delivering drugs across the blood-brain barrier with focused ultrasound? Focused ultrasound and microbubbles is the only technique that can non-invasively, locally and temporarily open the blood-brain barrier to allow drugs into the brain. The way this technology works is by first injecting clinically-approved microbubbles and drugs into the bloodstream. When the microbubbles reach the area where the ultrasound is targeted in the brain, these bubbles oscillate, mechanically stimulating the blood vessels, allowing the barrier to open so that drugs can get reach the brain. This technology has generated a lot of excitement as the blood-brain barrier currently prevents over 98% of drugs from entering the brain. Therefore, this technique can allow drugs that have previously failed clinical trials due to this barrier, to be tested again. But do immune cells also infiltrate from the blood into the brain, together with the drug, while the blood-brain barrier is open? In some situations immune cell infiltration would want to be avoided, for safety reasons. However, in other cases, such as when treating brain tumours, the presence of peripheral immune cells inside tumours could really help. Our own immune cells are actually best suited at tackling the complex heterogeneity of brain tumours. In this project, you will explore with sectioning, staining and fluorescence microscopy whether different types of immune cells are extravasating into the brain when the blood-brain barrier is opened with focused ultrasound. Ultimately, the aim is to find out with which ultrasound parameters you would get immune cells entering the brain, to either choose parameters for a safer drug delivery profile without immune cell presence, or to harness the power of our own immune cells to treat brain tumours.
Non-invasive manipulation and imaging of the brains immune system Our brain has its own dedicated immune system and rapid response team: microglia. These cells actively survey the brain, clearing away toxins and pathogens. The ability to temporarily stimulate microglia has generated much excitement, due to its potential to treat brain diseases. For example, stimulating microglia can help clear away the amyloid-beta plaques that build up in Alzheimers disease.  Focused ultrasound is a non-invasive and targeted technology that can stimulate microglia in any region of the brain. However, how ultrasound is stimulating these crucially important cells is unknown. This project aims to visualise whether focused ultrasound stimulates PIEZO1 mechanically sensitive ion channels in microglia to better understand the mechanism of this stimulation (expertise in Dr Morses group). A genetically-encoded fluorescent reporter based on PIEZO1, GenEPi, developed in Dr Pantaziss group, will be used to visualise whether ultrasound is stimulating these ion channels, that play multiple roles in the activation of microglia. The student will design a setup to simultaneously image the activity of PIEZO1 with confocal microscopy while performing ultrasound stimulation, which will be tested in a microglial cell line. These results will provide invaluable insight into the mechanism of how focused ultrasound stimulates microglia, allowing ultrasound treatments to be optimised to achieve improved beneficial therapeutic effects for the treatment of neurological disorders, such as Alzheimers disease. 
Can focused ultrasound stimulate the activity of the brains innate immune cells?  Microglia are the innate immune cells of our brain. They actively survey the brain and clear away toxins and pathogens. The ability to temporarily stimulate microglia has the potential to help treat brain diseases, such as brain tumours, Alzheimers disease and Parkinsons disease. For example, stimulating microglia can help clear away amyloid-beta plaques that accumulate in Alzheimers disease brains.  Focused ultrasound is a non-invasive technique that has been used to stimulate neurons. However, we have discovered that the activity of microglia can also be modulated depending on the way ultrasound is emitted. In this project, the level of microglia activation will be assessed for a variety of different ultrasound exposure parameters. Staining for various markers of microglia and their activation will be used to visualise (with microscopy) changes in their morphology and activation level. This project will lead to key advances in the unexplored territory of how the activity of the innate immune cells in our brain can be non-invasively controlled with ultrasound. 
Can focused ultrasound stimulate astrocytes in our brain?  Astrocytes, together with endothelial cells, are the gatekeepers of the brains main security system “ the blood-brain barrier. They control which substances enter and exit the brain, and keep harmful substances out. Astrocytes also regulate blood flow to transport nutrients to neurons depending on their energy needs. These cells are at the forefront of our thought processes: they regulate our synapses, and recycle and secrete neurotransmitters. The ability to temporarily stimulate astrocytes has the potential to help treat brain diseases, such as Alzheimers disease and Parkinsons disease. Focused ultrasound is a non-invasive and targeted technique that has been used to stimulate neurons. However, we have discovered that the activity of astrocytes can also be modulated depending on the way ultrasound is emitted. In this project, the level of astrocyte activation will be assessed for a variety of different ultrasound exposure parameters. Staining for various markers of astrocytes and their activation will be used to visualise (with microscopy) changes in their morphology and activation level. This project will lead to key advances in the unexplored territory of how the activity of the brains gatekeepers can be non-invasively modulated with ultrasound to treat brain diseases. 
Can focused ultrasound delay Alzheimer's disease?  Focused ultrasound is a technology that has very recently shown to restore cognitive function in Alzheimer's disease patients. This is a non-invasive technology that can be focused onto specific regions of the brain. One theory is that this technology can restore cognition by stimulating the innate immune cells of the brain as well as neuronal function and health. In this project you will explore 1) whether this same technology can be used to delay Alzheimer's as well as restore cognition and 2) delve into exploring the mechanisms behind why these effects are observed. This will involve working with mouse brain tissue, sectioning, imaging, staining and fluorescence microscopy.  
Can focused ultrasound delay brain ageing?  Focused ultrasound is a technology that has very recently shown to restore cognitive function in Alzheimer's disease mice and patients. This is a non-invasive technology that can be focused onto specific regions of the brain. One theory is that this technology can restore cognition by stimulating the innate immune cells of the brain as well as neuronal function and health. In this project you will explore 1) whether this same technology can be used to delay age-related cognitive decline, as well as restore cognition in Alzheimer's disease, and 2) delve into exploring the mechanisms behind why these effects are observed. This will involve working with mouse brain tissue, sectioning, imaging, staining and fluorescence microscopy.  
Can ultrasound help prevent organ transplant rejection?  Immune cells triggering inflammatory responses can lead to the rejection of organ transplants. Recently, ultrasound has been shown to have an anti-inflammatory effect on immune cells, such as macrophages. We here propose to investigate how ultrasound can be used best to lead to the release of anti-inflammatory cytokines. CD4+/CD8+ T cells, purified T-regulatory cells and macrophages will be cultured in vitro and multiplexed cytokine assays will be performed following ultrasound treatment. These findings will be translated and tested on organ transplants of hearts, livers and lungs currently being done at the Technical University of Munich (TUM). This project is in collaboration with Dr Konrad Fischer from TUM. Cell culture skills are preferable, any experience with cytokine assays desirable. 

Dr Spyros Masouros

Contact details: s.masouros04@imperial.ac.uk

Title Description
Bone failure at high loading rates This project is heavily lab based. It involves using existing machines in the lab and a dual x-ray system that can capture x-rays during testing in order to quantify the initiation and progression of bone failure under multiple loading modes and at physiological loading rates (ie load not slowly, but quickly).This test will be the first of its kind. The data can be used to develop material models of skeletal damage, which are critical for the prediction of injury in widely used (mostly by the automotive injury) current human body models.If you want to combine this with computational modelling, then we can develop finite element (FE) models of the conducted tests and assess whether current damage models are able to capture the fracture initiation and progression we see in the lab. You will need background in FEA, which means having taken an FEA course previously or take BACSA offered in the spring term.
Project on injury biomechanics / orthopaedic and trauma surgery / optimal rehabilitation post injury Generic title. Please contact me during office hours to discuss options

Dr Sylvain Ladame

Contact details: s.ladame@imperial.ac.uk

Title Description
Engineering denaturing hydrogel for efficient and automated recovery of RNA biomarkers from blood Circulating cell-free nucleic acids (cfNAs) in blood have recently emerged as clinically useful and minimally invasive diagnostic and predictive tools for a broad range of pathologies, including cancer and prenatal disorders. Among them, non-coding microRNAs (or miRNAs) are frequently found to be upregulated or down-regulated in body fluids and have great potential as novel blood-based fingerprints, e.g. for the early detection of cancer. However, because of their complexity and cost, current gold-standards for microRNA detection (e.g RT-qPCR) are unsuitable for point-of-care testing and can only be carried out by trained professionals in equipped laboratories. One of their main shortcomings is their inability to provide clear answers from whole blood without heavy processing which, in the absence of standardised procedures, is a major source of errors and variation between analyses.Herein, we are proposing to engineer and test a broad range of denaturing hydrogels capable of efficiently extracting miRNAs from whole blood, focusing on their ability to i) release miRNAs bounds to proteins and/or trapped in exosomes and ii) isolate those small RNA oligonucleotides by size exclusion. Engineered hydrogels covering a broad range of viscosities and mesh-sizes (e.g. through chemical modification and variation in concentration of the monomers/fibres) will be tested.
Paper-based lateral flow assay for early prediction of preterm birth 15 million preterm babies are born every year, with preterm birth (PTB) the largest cause of death of children under five worldwide. Birth before 26 weeks associates with 80% mortality, 25% severe handicap, and 75% overall morbidity in survivors. The strongest risk factor for PTB is previous PTB.[2] However the majority of PTBs are to women with no identifiable risk factors and occur in first pregnancies. So an urgent need exists to develop a simple, economic test which identifies preterm labour (PTL) risk in all settings and particularly in low-risk women early in pregnancy. Current predictive tests include cervical length scans and fetal fibronectin tests, which are usually conducted during late 2nd or 3rd trimester, at which point it becomes difficult to prevent PTB.[3] Timely medical interventions such as progesterone and steroids could help prevent PTB and reduce the risk of health issues in new-borns, justifying the need for a new, non-invasive and low-cost test based on the detection of highly specific blood biomarkers for earlier prediction of PTB and point-of-care monitoring of response to treatment.Having already demonstrated that probes engineered in-house can detect endogenous concentrations of miRNA in solution, we will next carry out similar sensing experiments on lateral flow (LF) paper strips, using an optical readout. Briefly, a lateral flow nitrocellulose membrane will be used as a low-cost platform for both sensing and detection. It will be functionalised in-house with a test line of biotinylated PNA catch-probe. Sequence-specific immobilisation of the only miRNA of interest followed by on-chip reaction between the catch probe and the reporter probe will result in the formation of a fluorescent dye that can be visualised optically.
Hydrogel-coated microneedle arrays for early diagnosis of skin cancer Minimally-invasive technologies that can sample and detect cell-free nucleic acid biomarkers from liquid biopsies have recently emerged as clinically useful for early diagnosis of a broad range of pathologies, including cancer. Although blood has been so far the most commonly interrogated body fluid, skin interstitial fluid has been mostly overlooked despite containing the same broad variety of molecular biomarkers originating from cells and surrounding blood capillaries. Minimally-invasive technologies have emerged as a method to sample this fluid in a pain-free manner and often take the form of microneedle patches. Herein, we will develop microneedles that are coated with an alginate-peptide nucleic acid hybrid material for sequence-specific sampling, isolation and detection of nucleic acid biomarkers from skin interstitial fluid. This platform technology will also enable for the first time the detection of specific nucleic acid biomarkers either on the patch itself or in solution after light-triggered release from the hydrogel. Considering the emergence of cell-free nucleic acids in bodily fluids as clinically informative biomarkers, platform technologies that can detect them in an automated and minimally invasive fashion have great potential for personalized diagnosis and longitudinal monitoring of patient-specific disease progression.

Dr Yap

Contact details: c.yap@imperial.ac.uk

Title Description
Deep Learning Simulation of Vascular Fluid Mechanics In this project, we will use deep learning to do computational fluid dynamics simulations of blood vessels.Computational simulations of flow in blood vessels, including those in the heart, neck, brain, torso and extremeties, has helped improved our understanding of disease physiology and mechanobiology, such as how abnormal flow shear stresses caused atherosclerosis or aneurysm, and helped us determine surgical options, such as how to connect great veins to great arteries in the Fontan patients. However, computational simulations are slow and this prevented clinical adoption. Recently, physics informed neural networks (PINN) has shown great promise of replacing traditional computational flow simulation algorithms. A PINN calculates flow velocities and pressures by imposing Navier-Stokes governing equation as constraints on the network outputs. PINN is highly versatile, but still requires long training time whenever a new vascular case is encountered. Here, we propose a new strategy of pre-training PINN over a range of geometric cases, by parameterizing the geometry. Once trained, the goal is that algorithm can then generate flow results almost instantaneously, and can then be incorporated in real time to the clinical setting. Working with a PhD student or a research fellow, student will investigate various strategies to achieve this goal. 
Importance of the Forces of the Heart During Embryonic Heart Development Congenital Heart Malformations affects about 1% of pregnancies, and can be devastating. There are evidence that abnormal biomechanical forces and subsequent abnormal mechano-biology expressions could be responsible for the malformations. For example, fetal aortic stenosis can prevent normal contractions of the heart, and the lack of deformational stimuli may lead to underdevelopment of the ventricle. In this project, we will perform analysis of 4D microscopy images of zebrafish embryonic hearts, and use a finite element model to characterize the stresses and strains of the myocardium.  We will study the normal zebrafish embryonic heart and compare it to disease models, and correlate the mechanical environment to their biological expressions, to understand the importance of proper mechanical stimuli on embryonic heart growth. The project will be co-advised by Dr. Yap on the mechanics part, and Dr. Vermot on the biology part.
Deep Learning Segmentation of the Fetal Heart to Improve Congenital Malformation Detection Echocardiography screening of fetal hearts is a standard of care, and is routinely performed during a pregnancy. However, mass screening programmes only detect 50% of congenital heart malformations. Being surprised at birth with cardiac malformation result in less time for clinicians to plan for and provide treatment, resulting in higher mortality risks, and in greater risks for neurological maldevelopment, due to prolonged duration where the malformation is not treated and continue to compromise brain oxygen transport. Here, we pursue a strategy to improve detection with deep learning of echocardiography images, which involves using a deep learning algorithm on 4D echo images to extract sizes, shapes and motions of cardiac structures, and subsequently, using unusual size, shape and motion information to lead to malformation detection.In this project, the student will work with a team, to investigate a subset of this grand strategy, where we will develop, train and validate a deep learning algorithm to segment cardiac chambers from clinical echocardiography images of fetal hearts (both healthy and malformed), We will test the effectiveness of combinations of advanced strategies, including a variational cardiac shape encorder, an adversarial network, and a multi-scale matching approach, to perform the segmentation.
Deep Learning Simulation of Cardiac Myocardial Mechanics Finite Element computational simulations of myocardial tissue biomechanics has been very useful in the past to understand the heart's function during health and disease, and to predict how the heart will remodel in disease. However, the simulatinos are slow and clinicians find it hard to adopt the technique. Recently, physics informed neural networks (PINN) has shown great promise of replacing traditional biomechanics simulations. In this project, the student will work with a team to develop and test a Physics Informed Neural Network (PINN) that emulates the finite element simulations of the heart. PINN calculates tissue deformations and stresses by imposing the governing equation of tissue deformations and active tension as constraints on the network outputs. We will test the effectiveness of PINN with a cohort of 4D echocardiography images, and compare them to traditionally simulated finite element models. We will test a new strategy of pre-training PINN over a range of geometric cases, by parameterizing the geometry. Once trained, the goal is that algorithm can then generate flow results almost instantaneously, and can then be incorporated in real time to the clinical setting.
Computational Biomechanical Modelling of Heart Failure  Heart failure is the #1 killer in the world. During heart failure, the heart remodels (changes it shape and tissue architecture), but the effects of this process on the biomechanical function of the heart is not well-understood. Our preliminary data suggest that such remodeling changes reduces the biomechanical efficiency of the heart, leading further deterioration of function, in a downward spiral for cardiac health. In this project we will test this idea. If this proves true, we can recommend to doctors to focus on using drugs to prevent cardiac remodelling . We will use MRI images of human hearts in health and during cardiomyopathy disease, and use an existing finite element computational code to model the heart's function. We will vary various various parameters describing remodelling changes, and test their effects on the heart's function. The student will learn image processing and computational modelling of soft tissues, and work with members from the school of medicine and Bioengineering to ensure effective guidance. 

Professor Anil Bharath

Contact details: a.bharath@imperial.ac.uk

Title Description
Generative Modelling for Biomedical Time Series Score matching has a relatively simple definition, but has lots of implications in developing computational techniques for several applications: the synthesis of examples of medical data that preserves privacy of individuals, the learning of representations, and even more mundane things such as testing of software for the purposes of industry regulation. Recent developments in synthesizing data examples has led to improvements in generating, for example, plausible synthetic faces that do not represent actual individuals. The aim of this project is to apply these ideas of generative modelling to biomedical data, focussing specifically for time series (signals). 
Temporal GANs and de-identified datasets Generative Adversarial Networks are being explored possible solutions to privacy-preserving challenges in the context of data driven healthcare. In essence, the trained generator of a GAN is able to draw samples from complex probability density functions without having to have an explicit expression for that PDF: samples of real data serve the purpose. Recent developments in synthesizing image data examples has led to improvements in generating, for example, plausible synthetic faces that do not represent actual individuals. The aim of this project is to apply these ideas of generative modelling to biomedical data, focussing specifically for time series (signals).  Two key challenges are: i) the concept of $k$ anonymity, applied to high-dimensional data vectors, such as those representing time series. The goal, here, is to consider the distance of generated samples from _any_ data samples that are present in the real data, and to do so in a way that provides a convincing argument for privacy preservation along the lines of $k$ anonymity. ii) the ability to achieve conditional generation of synthetic data that preserves distances of sub-populations in the dataset.
Learning from Mostly Unlabelled Data  Although data-driven machine learning is now being used to build components of AI systems for medical diagnostics, there is the need to find ways of being able to learn representations that are well suited to new forms of imaging data for which ground truth does not exist. This problem is not fully addressed in the AI field, where it is often assumed that data are readily available with ground truth, and that the (data, label) tuple can be made sufficient. But as new ways of imaging or measurement become available, the "data" part of that tuple can be subtly or radically different. Subtle differences can probably be handled by transfer learning, but radical ones require new approaches to learning, such as **contrastive learning**. This project will investigate contrastive learning in the context of medical imaging data from a variety of sources.

Professor Darryl Overby

Contact details: d.overby@imperial.ac.uk

Title Description
Organ-on-chip models for glaucoma This project will develop an organ-on-chip model for glaucoma, focussing on rebuilding the microenvironment of the inner wall endothelium of Schlemm's canal, which is a principle regulator of intraocular pressure. Disruption of this endothelium contributes to glaucoma.A PhD student is currently working on this project. The MSc/MRes student will interact with the PhD student to contribute to specific developments, including examining the effect of fluid shear stress on nitric oxide production, examining the effect of substrate stiffness, examining the effect of dexamethasone (which induces glaucoma like symptoms by disrupting the mechanobiology of the inner wall of Schlemm's canal).
The altered mechanobiology of Schlemm's canal cells in glaucoma Schlemm's canal endothelial cells regulate pressure in the eye. Specifically, they create a barrier to fluid drainage from the eye that preserves eye pressure. If the cells become too stiff, then barrier resistance increases and pressure rises as occurs in glaucoma (Overby et al., PNAS 2014). Decreasing cell stiffness is associated with reduced eye pressure. Understanding the factors regulating Schlemm's canal cell stiffness is thus critical to understand the pathogenesis of glaucoma, a disease associated with elevated eye pressure that affects 60 million people worldwide.This project will examine the mechanobiology of Schlemm's canal cells in steroid-induced model that mimics glaucoma. A key feature of Schlemm's canal cell mechanobiology is that, like vascular endothelial cells, they produce nitric oxide (a bioactive soluble gas) in response to shear stress, and nitric oxide is a potent regulator of cell stiffness. In this cell culture project, the student will work with a post-doc to expose human Schlemm's canal cells to controlled hydrodynamic shear stress and measure nitric oxide production using an established biochemical (Griess) assay. The student will specifically examine whether exposure to dexamethasone (a corticosteroid that stiffens cells and incudes glaucoma-like symptoms) reduces the ability of Schlemm's canal cell to produce nitric oxide in response to shear stress.
The mechanobiology of cell volume change and pore formation A key determinant of the biomechanical properties of living cells is cell volume, which is in turn controlled by intracellular water content. By osmotically drawing water out of the cell, the cell deswells and its stiffness increases. Likewise, increasing the intracellular water content by swelling the cell leads to a decrease in cell stiffness.In collaboration with researchers at the Institute for Light and Matter (CNRS, Lyon), we are investigating the relationship between cell volume, intracellular water content and cellular biomechanical properties. Cells are cultured within a microfluidic chamber that allows precise optical measurement of cell volume whilst perturbing the osmolarity of the surrounding media. A senior PhD student is currently optimising on this setup, which will be used by a second incoming PhD student who will assess cell mechanical properties.In this project, the UG student will collaborate with other lab members to further develop the microfluidic approach to measure cell volume whilst perturbing media osmolarity. Of particular importance, we have noticed that deswelling the cell leads to the formation of openings or pores through the cell body, which may be relevant for the transport of fluids, solutes or other cells across the endothelial barrier. Working with a senior lab member, the UG student will use this setup to investigate the hypothesis that pore formation involves similar membrane proteins as involved in vesicle fusion in other cell types (e.g., in the neurological synapse).This project has opportunities for multiple students collaborating together.
Micropatterning living cells to measure endothelial barrier function Endothelial cells serve as gatekeepers that control the passage of fluids, macromolecules and cells between the blood and all other tissues of the body. The transport properties of endothelial cells are therefore critical for health and disturbed transport properties leads to disease such as oedema, cancer, glaucoma and atherosclerosis.In this project, the student will work closely with a senior PhD student to develop a micropatterning system to assess endothelial barrier function. Micropatterning is a way to precisely control the location and shape where cells adhere to a surface using microfabrication. The microfab techniques have already been largely sorted out by a PhD student and postdoc.The Overby lab has been awarded a new grant to apply these microfabrication approaches to better understand glaucoma, a blinding eye disease associated with elevated eye pressure. Elevated eye pressure is caused by increased barrier function of Schlemm's canal endothelial cells in the eye.This project works with a team of researchers, including postdocs and PhD students, to test the hypothesis that engineering the microenvironment of Schlemm's canal endothelial cells is key to regulation of barrier function. Factors to investigate include cell shape, substrate compliance, co-culture with neighbouring subendothelial cells, and mechanical forces associated with stretch, shear and transendothelial flow. Our grant was funded based on a plausible design strategy for a microfabricated platform that can recapitulate all of these key factors.Successful completion of this project would introduce a novel screening platform that can be used to test the applicability of novel anti-glaucoma drugs that more successfully lower eye pressure to prevent vision loss.
An high-throughput assay for single-cell contractility In this project, the student work with senior lab members to develop a microfabricated platform to measure the contractile forces of individual cells. The goal is to apply this approach to blood-outgrowth derived smooth muscle cells (i.e, SMCs that can be isolated from patient's blood). We aim to compare the response of BO-SMCs isolated from diabetic vs non-diabetic donors, across a range of different contractile stimulants.The project will adapt a design previously published in the literature. Briefly this uses a thin hydrogel layer upon which adhesive islands are created using micropatterning (see prior project). Cells adhere only to the adhesive islands, and by measuring the shape change of the island in response to a stimulus, it is possible to calculate the cellular contractile force.The student will collaborate with several senior lab members who are experienced with microfabrication, and a senior research fellow in NHLI who has collected BO-SMCs from various patients.
Organ-on-chip approaches to visualise fluid flow through native tissues The Overby lab is developing innovative organ-on-chip technologies to preserve the viability and function of living tissues outside of the body. Briefly, resected tumour tissue removed during surgery is placed within our device, where it is kept alive by nutrient perfusion. The ultimate goal is ex vivo screening and analysis, as is useful for personalised medicine (e.g., finding out which drugs best kill cancer cells within a patient’s own tumour tissue).A key question in the success of this technology is visualising how flow passes through the tissue. This is important because flow is what is providing nutrients and oxygen to maintain tissue viability, and hence regions that get more (or less) flow have more (or less) viability.In this project, the student will work with senior lab members in the Overby and Tang labs to measure the distribution of fluid flow through perfused tissues. This project leverages the super-resolution capabilities of the Tang lab that can measure the spatial location of microbubbles with extreme accuracy using ultrasound. By perfusing tissue with these microbubbles, we can in principle map the spatial distribution of these microbubbles and how they change over time, revealing the distribution of flow and nutrient delivery.There is room in this project for multiple students, who will work with an experienced team of researchers between the Overby and Tang labs, including post-docs and PhD students. This work is currently funded by the CRUK, NC3Rs and has a various applications in cancer, microphysiological systems and imaging.
organ-on-chip approaches to visualise cancer metastasis in living tissue slices Metastasis is the spread of cancer cells throughout the body, from the primary tumour to sites in the brain, liver, lung or elsewhere. Metastasis is responsible for the large majority of cancer deaths, and so understanding metastasis is key to developing better therapies that target and stop cancer spread.Most models of metastasis involve in vivo studies, mostly using mice, where cancer cells are injected intravenously. The cancer cells then extravasate into healthy tissues to colonise a secondary tumour. However, because the studies are done in the living animal, it is difficult to follow the dynamics of how cancer cells leave the blood vessel ('extravasate'), colonise and survive within the surrounding tissue.The Overby lab is develop next-generation organ-on-chip approaches to study cancer cell metastasis using various different approaches, including bottom-up and top-down organ-on-chip designs as well as tissue slice cultures. The advantage of the latter two approaches is that the native tissue microenvironment is fully preserved, rather than trying to build it from the "bottom-up" as in conventional organ-on-chip designs.In this project, the student will work with senior lab researchers to develop new ways of studying cancer metastasis in living tissues using tissue slice models. In this model, living tissue is cut into thin sections, which are cultured within a controlled environment. Cancer cells can be labelled fluorescently and added to the slice, where there dynamics (migration, proliferation, viability) can be measured over time. The student will contribute to the design and development of the culture device, including microfabrication and microfluidics, and validation (e.g., measuring tissue viability and function over time). This device will allow novel hypotheses to be tested, for example, why some cancer cells prefer to metastasise to specific tissues (brain vs. liver), which can be tested by comparing the spatiotemporal dynamics of the same cancer cells within different tissues.

Professor Holger Krapp

Contact details: h.g.krapp@imperial.ac.uk

Title Description
Electrophysiological characterization of optic flow processing interneurons in flying insects. The visual system in many animals and humans contributes to state change estimations by analysing panoramic retinal image shifts known as optic flow. Earlier studies in blowflies have revealed the underlying neuronal mechanism which is believed to complement state change estimation based on mechanosensory/inertial systems. This project aims to support the idea that optic flow processing in the visual system of flying insects is tuned to control species-specific natural modes of motion which are determined by the animal's flight dynamics. Experimental evidence in support for this mode sensing hypothesis requires a comparative study of the receptive field properties of motion sensitive interneurons in the visual systems of dipteran flies other than blowflies as well as species that belong to other orders of flying insects. Hoverflies, which show distinctly different flight patterns than blowflies and perform differently in behavioural gaze stabilization experiments, would be an ideal candidate species. Preliminary experiments have shown that hoverflies, too, employ motion sensitive interneurons in their visual system. But only a few of those interneurons have been studied so far. This project requires the dissection of flying insects for extracellular recordings from motion sensitive interneurons during visual stimulation. The neuronal responses will be analysed using customized MATLAB/Python programmes which reveal the cell's receptive field organization. From the electrophysiological results the preference of individually identifiable interneurons for specific self-motion components (state changes) will be derived. The principles underlying state change estimates in flying insects will inform the design of energy-efficient flight control architectures for autonomous aerial systems.  

Professor Jimmy Moore

Contact details: james.moore.jr@imperial.ac.uk

Title Description
Lymph node physiological mass transport Many important immune processes occur in lymph nodes, but we actually know little about their structure. This is important because it determines patterns of mass transport which assure that antigens are presented to the appropriate cells.  Based on a combination of high-resolution imaging protocols that quantify the structure of human lymph nodes, we will construct models of mass transport of chemokines and antigens in collagen bundle conduits, parenchyma and blood vessels.
Lymph Node Implant for Breast Cancer-Related Lymphoedema A large percentage of breast cancer patients who undergo lymph node resection develop an incurable swelling of the arm called lymphoedema. We are developing an implant to replace the fluid delivery characteristics of lymph nodes. We have developed a lymph node implant that releases a growth factor to regenerate the damaged lymphatic vessels. We would like to visualize the release and flow of the growth factor once it is implanted in the tissue. We can do this using a microfluidic chip that simulates fluid flow like in tissue. The goal of this project is to use a fluorescence microscope to live image the microfluidic device as the fluorescent-labelled growth factor flows through it over time. This will help us understand where the growth factor travels to within the surgical site. 
Compression/vacuum device for treating lympoedema Patients suffering from incurable lymphoedema use compression bandages to limit limb swelling.  Our flow modelling studies have shown that application of compression actually shuts down lymphatic pumping.  We have therefore designed a device to enhance lymphatic pumping with the application of oscillatory pressures.  We aim to test the device on volunteers and then progress towards trying it on patients.
Stem Cell Injection Device for Minimising Shear-Induced Cell Lysis While there is great potential in using cells as part of therapeutic strategies for many diseases, these strategies are limited by the survivability of cells during the injection process.  We have designed a combination hydrogel and syringe injection system that aims to maximise cell viability.  The project will involve making different hydrogel formulations and testing their mechanical properties.  This information will be used to determine the details of the syringe design, which will be tested computationally and experimentally.
Infrared imaging device for quantifying lymphoedema Around 20% of breast cancer patients develop an incurable swelling of the arm following cancer treatment.  This condition (lymphoedema) is painful, debilitating, and leads to depression in many victims.  One factor that limits the ability to develop new treatments is the lack of a reliable diagnostic.  We have developed a rough prototype of a device that uses 3 infrared cameras to generate a 3D point cloud of the arm, and wish to deploy it clinically.  The project will involve image analysis, programming, and designing a user interface.  The goal is to have a device ready for clinical application by the end of the project.
MRI phantom for testing lymphatic vessel imaging sequences There are no means for measuring pressure, flow or diameter reliably in any lymphatic vessel without surgical exposure.  While the vessel diameters are on the order of MRI spatial resolution, their unique flow dynamics profile offers opportunities to develop diffusion-based imaging sequences to distinguish them from blood vessels and interstitial fluid movement.  We aim to develop a phantom that reproduces these flow patterns to aid in the development of better imaging sequences.

Professor Pantelis Georgiou

Contact details: pantelis@imperial.ac.uk

Title Description
Design of large-scale chemical sensing platform for diagnostics At the Centre for Bio-Inspired Technology, we have developed Lacewing, a handheld platform able to perform nucleic acid detection in under half an hour. Our approach relies on Lab-on-Chip technology, combining microfluidics, molecular biology and microchip technology to form a smart cartridge. The microchip is implemented in standard semiconductor technology and integrates thousands of label-free electrochemical sensors which detect the release of protons associated with nucleic acid amplification. The sensors suffer from non-idealities including noise and temporal drift which compromise the readout and therefore the accuracy. The molecular assay is based on loop-mediated isothermal amplification (LAMP) which is an equivalent reaction to PCR offering rapid time-to-result and higher specificity. In this project, the student will focus on implementing an efficient temperature regulation method based on a PID controller to maintain the reaction temperature at 63oC during the nucleic amplification. The student will apply to the new methods and data processing algorithms to a new very large array of 59,000 sensors (on a 8 mm x 12 mm microchip called TITAN), considering the limitations in terms of data bandwidth. If the student is interested they will have the opportunity to run DNA experiments in the laboratory. The project is novel and a good outcome will lead to a conference publication. We expect the student to have basic knowledge in hardware design and programming (python and Matlab). The student will be joining a multi-disciplinary team of enthusiastic researchers and will be encouraged to get involved with several fields and learn new skills. This project is suitable for students wanting to do lab based testing of biochemical reactions (DNA based), characterization of sensors and development of lab-on-chip platforms. Required expertise: Instrumentation, Matlab programming, wet-lab skills, understanding of DNA. Skills on small volume testing is desirable.

Professor Rylie Green

Contact details: rylie.green@imperial.ac.uk

Title Description
A bioelectronic implant for cancer treatment This project revolves around aiding in the development of a device for the selective delivery of chemotherapy directly to the site of non-operable brain tumors (glioblastoma multiforme). This device consists of a conductive polymer-based material that can used as an electrically controlled drug delivery system. The goal of this project is to evaluate the drug release profiles for multiple different molecules that are analogs to those commonly used in chemotherapy. Parameters such as molecule size, charge, and stability will be investigated. Characterization of the drug release profiles will be accomplished through chemical, electrochemical, and spectroscopic techniques. 
Printing of flexible polymer bioelectronics The overall goal of this project is to investigate the feasibility of fabricating well defined patterns of conducting polymer-based bioelectronics through printing (inkjet or melt electrospin writing). This technique takes advantage of the viscous liquid phase dispersion of the conductive polymer in solvent to enable printing through a small diameter nozzle. Use of thermal processes will be investigated as methods to control viscosity and printing tolerances. Students with robotics interests will have an opportunity to build a bespoke printer which can be controlled through CAD file geometries and used to create 3D implants from the extruded material. 
Living Bionics: Stimulation to drive neural network development Electrical stimulation has been demonstrated to induce directional neurite growth in various cell types, both human and non-human using biphasic stimulation. This research project aims to evaluate a range of sinusoid stimulation frequencies to drive activity, growth and release of neurotransmitters of developing neurons using a cell stimulation rig made in house. 
Spinal cord bridge Nerve regeneration in an injured spinal cord is often restricted. One possible reason may be the lack of topographical signals from the material constructs to provide contact guidance to invading cells or re-growing axons. This research project aims to evaluate randomly oriented or aligned collagen fibers coated on cuff electrodes to study device topographical effects on astrocyte behavior and neurite outgrowth respectively, using electrical regimes. 
Biofunctionalising electrodes through conductive hydrogel coatings Peripheral nerves cuffs can have there biointegration improved through the use of conductive hydrogel coatings. One of our long-term goals is to improve the long term performance of these electrodes. This projects will investigate reducing inflammation at the site of implantation by the incorporation and release of anti-inflammatory mediators from the hydrogel coatings. Different mediators will be investigated along with different release methods and how these impact the performance of the cuff electrodes. 
A bioelectronic implant for cancer treatment This project revolves around aiding in the development of a device for the selective delivery of chemotherapy directly to the site of non-operable brain tumors (glioblastoma multiforme). This device consists of a conductive polymer-based material that can used as an electrically controlled drug delivery system. The goal of this project is to evaluate the drug release profiles for multiple different molecules that are analogs to those commonly used in chemotherapy. Parameters such as molecule size, charge, and stability will be investigated. Characterization of the drug release profiles will be accomplished through chemical, electrochemical, and spectroscopic techniques. 
In ear monitoring of EEG This project is a funded collaboration with EEE and international partners. It seeks to deliver accurate EEG and other biosignal information through eletronics mounted in an ear bud. The project will focus on the design and characterisation of dry wearable electrodes and the identification of biosignals through processing. The areas of focus will span biomaterials, electronics, signal processing and neural netoworks.