MSc Projects 2015-16

Here is a list of projects being offered for MSc students in the Department.

For details of deadlines and for registering submissions, see the Moodle page. Full instructions are available here.

Processing of fNIRS signals to enhance understanding of fundamental neural processes in the human brain

Supervisors: Ilias Tachtsidis and Joy Hirsch

Functional neural imaging is the leading technology for studying brain and behaviour relationships. However, acquired signals based on the oxygen content of blood may include spurious sources of noise and/or be influenced by factors other than neural activity. Separating the noise and neural from non-neural components of NIRS signals is an active area of signal processing development aimed at improving methods for understanding neural function. Noise reduction strategies based on models of the expected hemodynamic response function and structured task paradigms are common. Additionally NIRS signals are particularly vulnerable to global systemic effects due to surface blood flow that contributes to the signal as well as secondary effects related to physiological factors. Separation of task-specific neural effects from distributed global (systemic) effects is challenging due to correlations of the waveforms. Signal processing approaches to solve this problem are an important current and on-going topic of investigation.
We have high quality 98 channels fNIRS data from 10 healthy subjects performing a finger and thumb tapping task. The student task will be to use this data and develop signal processing techniques to separate local neural haemodynamic effects from global systemic haemodynamic effects in addition to investigating haemodynamic connectivity between brain areas. The student needs to be familiar with MatLab and have some previous experience with signal processing.
This project is part of a new collaboration between Yale (Professor Joy Hirsch) and UCL (Dr. Ilias Tachtsidis) in the application of fNIRS technologies to monitor the adult functional brain.

Student: (Project available)

Neural specializations for interpersonal interaction in a competitive poker task: an fNIRS investigation

Supervisors: Ilias Tachtsidis and Joy Hirsch

Dynamic interpersonal interaction is integral to typical human social behaviour. Although recent dyadic studies have investigated the paired neural correlates of communicative and cooperative tasks, little is known about the paired neural correlates that underlie competitive tasks requiring social interaction. Conventional neuroimaging techniques such as fMRI have identified key social regions in single brains specifically activated by competing with a human rather than a computer opponent, particularly the temporal-parietal junction (TPJ). A goal of our study is to extend these findings to interpersonal neural correlates between two brains in an ecologically valid context. We have function near-infrared spectroscopy (fNIRS) data from several pairs of subjects playing a simplified poker game. In this game, one player was first randomly dealt either a low or high card. This player then had the opportunity to bet or fold. The opposing player was required to bet or fold in response to the action of the first player. The next trial then began with the opposite player receiving either a low or high card. This simplified poker game was played in two different conditions. In condition one, two subjects played the poker game against each other. In the condition two, the same subjects simultaneously played against matched computer opponents.
The student will be utilising signal processing techniques with the fNIRS data to investigate interbrain coherence during the human vs human and human vs computer conditions. In particular the student will be looking at synchronicity of the haemodynamic signals and investigate how that changes at different conditions. The project is computational and will be suitable to students that are familiar with MatLab and have some knowledge of signal processing.
This project is part of a new collaboration between Yale (Professor Joy Hirsch) and UCL (Dr. Ilias Tachtsidis) in the application of fNIRS technologies to monitor the adult functional brain.

Student: (Project available)

Exploration of the implementation of combined magnetic resonance and optical methods for assessment of brain metabolism and haemodynamics: application to a preclinical model of birth asphyxia.

Supervisors: Pardis Kaynezhad and Ilias Tachtsidis

Perinatal cerebral hypoxic-ischemia (HI) is a condition resulting from reduced oxygen delivery or/and blood flow occurring either in utero or during delivery. It occurs in 1 to 2 per 1000 live births and can result in physical or sensoreal handicap or fatality. Although there are significant advances in the treatment of asphyxiated babies, little is known about the effects of these neuroprotection strategies on brain blood perfusion and metabolism following HI. This project is part of an exciting collaboration between the Perinatal-Brain Magnetic-Resonance Group at UCH and the Biomedical Optics Research Laboratory at UCL that aims to deliver novel measurements to investigate the effects of brain neuroprotection through combination of magnetic resonance and optical (near infrared) technologies. The main aims of this MSc project are to (1) assist in experimental data collection; and (2) to analyse imaging, perfusion and metabolic data, from the optical instruments, obtained from piglet brains in-vivo after HI. The student will learn how to operate the optical instrument and use novel software tools that will allow quantification of the optical measurements and then will focus on analysing those in conjunction with the magnetic resonance imaging & spectroscopy measurements. The larger scope of the analysis is to investigate the benefits of implementing treatment in birth asphyxiated infants. This project would be suitable for a student with an interest in optical technologies, physiology/pathophysiology, brain tissue biochemistry; and will involve data collection, data processing and some statistical analysis.

Student: (Project available)

Spatially Resolved Spectroscopy to Monitor Neonatal Brain Injury.

Supervisors: Gemma Bale and Ilias Tachtsidis

Neonatal brain injury, in particular hypoxic-ischaemic encephalopathy, occurs in 1 in 1000 births and can lead to severe neurodevelopmental impairment or even death. There is an urgent need to monitor the changes in cerebral oxygenation in the first days of life to assess the injury, detect those infants at risk and help direct treatment. To address this challenge our group is focussed on developing instruments and methods to monitor cerebral changes at the cotside. Near-infrared spectroscopy (NIRS) is an optical technique that can monitor changes in brain tissue oxygen levels non-invasively. Infrared light is directed into the head and light that is reflected back to the surface will be attenuated by different chromophores (light absorbing molecules) such as oxygenated and deoxygenated haemoglobin according to their concentration. In this way we are able to monitor changes in these molecules. Using multiple detectors at different distances from the light source, it is possible to measure the difference in attenuation at different distances and work out the level of oxygenation absolutely; this is called spatially resolved spectroscopy (SRS). This project will focus on implementing this method and assessing its use in monitoring brain oxygenation in newborn brain injury.
Our group has built a broadband NIRS system, called CYRIL, which has been used to monitor infants with brain injury in the neonatal intensive care unit at UCL Hospital (UCLH) over the past two years. We have studied over 50 infants and are continuing to collect data.
This project will involve implementing SRS algorithms to measure oxygenation on the optical data from these infants, testing the robustness and reproducibility of the method, and using the results to assess its potential to identify the severity of brain injury. There is also scope to conduct studies on phantoms or adult volunteers to aid in testing the algorithm. Further it might be necessary to modify the existing study procedure and/or probe design to improve the collection of data for SRS. It will require programming skills (preferably Matlab or similar) and enthusiasm to work in a multidisciplinary environment.
This project is a collaboration between the Biomedical Optics Research Laboratory in UCL and the Elizabeth Garrett Anderson (EGA) neonatal wing in UCLH.

Student: (Project available)

Measuring Cerebral Blood Flow to Monitor Neonatal Brain Injury

Supervisors: Gemma Bale and Ilias Tachtsidis

Neonatal brain injury, in particular hypoxic-ischaemic encephalopathy, occurs in 1 in 1000 births and can lead to severe neurodevelopmental impairment or even death. There is an urgent need to monitor the changes in cerebral oxygenation in the first days of life to assess the injury, detect those infants at risk and help direct treatment. To address this challenge our group is focussed on developing instruments and methods to monitor cerebral changes at the cotside. Near-infrared spectroscopy (NIRS) is an optical technique that can monitor changes in brain tissue oxygen levels non-invasively. Infrared light is directed into the head and light that is reflected back to the surface will be attenuated by different chromophores (light absorbing molecules) such as oxygenated and deoxygenated haemoglobin according to their concentration. In this way we are able to monitor changes in these molecules and we can use these changes combined with information from the systemic physiology to assess cerebral blood flow (CBF) and cerebral blood volume (CBV) which might provide insight into the state of brain injury. This project will focus on implementing the method to calculate CBF and CBV and assessing their use in monitoring brain injury in newborns.
Our group has built a broadband NIRS system, called CYRIL, which has been used to monitor infants with brain injury in the neonatal intensive care unit at UCL Hospital (UCLH) over the past two years. We have studied over 50 infants and are continuing to collect data.
This project will involve implementing methods to measure CBF and CBV from data from these infants and using the results to assess its potential to identify the severity of brain injury. This will involve using data from CYRIL combined with data from the systemic physiology to identify drops in the systemic oxygen saturation and use these to estimate CBV and CBF. These values can then be compared to the severity of the brain injury in the hope to find an early cotside biomarker. The project will require programming skills (preferably Matlab or similar) and enthusiasm to work in a multidisciplinary environment.
This project is a collaboration between the Biomedical Optics Research Laboratory in UCL and the Elizabeth Garrett Anderson (EGA) neonatal wing in UCLH.

Student: (Project available)

Using Near Infrared Spectroscopy to Assess Frontal Cortex Asymmetry in Neonatal Brain Injury

Supervisors: Gemma Bale and Ilias Tachtsidis

Neonatal brain injury, in particular hypoxic-ischaemic encephalopathy, occurs in 1 in 1000 births and can lead to severe neurodevelopmental impairment or even death. There is an urgent need to monitor the changes in cerebral oxygenation in the first days of life to assess the injury, detect those infants at risk and help direct treatment. To address this challenge our group is focussed on developing instruments and methods to monitor cerebral changes at the cotside. Near-infrared spectroscopy (NIRS) is an optical technique that can monitor changes in brain tissue oxygen levels non-invasively. Infrared light is directed into the head and light that is reflected back to the surface will be attenuated by different chromophores (light absorbing molecules) such as oxygenated and deoxygenated haemoglobin (HbO2 and HHb) according to their concentration. In this way we are able to monitor changes in these molecules. Another chromophore in the near infrared is oxidised cytochrome-c-oxidase (oxCCO) which is a marker of tissue metabolism and we are able to monitor this with broadband NIRS.
Our group has built a broadband NIRS system, called CYRIL, which has been used to monitor infants with brain injury in the neonatal intensive care unit at UCL Hospital (UCLH) over the past two years. We have studied over 50 infants and are continuing to collect data. The data is recorded from two channels over the left and right frontal cortex respectively. We are interested in assessing the differences between the two brain hemispheres. Previous work on healthy neonatal brain asymmetries has shown marked differences between the hemispheres; it will be interesting to see if this is the case in the injured brain.
This project will involve calculating the laterality between the two hemispheres for changes in HbO2, HHb and oxCCO during spontaneous events such as blood pressure increases. The project will require programming skills (preferably Matlab or similar) and enthusiasm to work in a multidisciplinary environment.
This project is a collaboration between the Biomedical Optics Research Laboratory in UCL and the Elizabeth Garrett Anderson (EGA) neonatal wing in UCLH.

Student: (Project available)

Automated optic radiation tractography for surgical planning in epilepsy

Supervisors: Sjoerd Vos and Gavin Winston

Epilepsy affects over 600,000 people in the UK. Drug therapy does not control seizures in 30-40% of patients. Epilepsy surgery is a cost effective treatment for medication refractory focal epilepsy, enhancing quality of life. Temporal lobe epilepsy is the most common form of epilepsy that proceeds to surgery, with anterior temporal lobe resection is an effective treatment. In this surgery, there is a risk of damage to the optic radiation - part of the white matter pathways in the visual system - which can result in a visual field deficit (VFD) in between 48% and 100% of patients.
Diffusion MRI is a unique in vivo MRI technique to reveal the local organisation of the white matter (WM), and can be used to to virtually reconstruct the WM pathways, called fibre tractography. Tractography of the optic radiation (OR) in surgical planning and during surgery reduces the severity of VFD without affecting surgical success rates. Currently, OR tractography is performed manually within an otherwise fully-automated processing pipeline. The aim of this project is to automate OR tractography using image segmentation of the start (lateral geniculate nucleus) and end (visual cortex) points of the tract. Preliminary work has demonstrated the feasibility of seed point determination but full tractography automation requires tractography-informed evaluation (e.g., by inclusion of anatomical priors) to ensure correct tract morphometry and to allow clinical translation.

Student: (Project available)

Radioisotope mapping using RadICAL

Supervisors: Robert Speller and Dan O’Flynn

RadICAL is a detector system designed for localising radioactive isotopes. Several versions have been designed and built and this project is to look at alternative approaches to the basic principle. This can be approached in several ways and will depend upon the skills of the student. A Monte Carlo modelling approach could be used or just experimental measurements. Different designs would need to be tested and this will require some building of equipment probably using the facilities available in Institute of Making. Testing would be undertaken in the Radiation Physics labs using radioactive isotopes.

Student: (Project available)

Novel collimation for spatial discrimination in x-ray diffraction

Supervisors: Robert Moss and Francesco Iacoviello

UCL has developed a new approach to X-ray diffraction which make use of a multi-element detector (a pixellated array). In the present setup the detector collects data for every point where X-rays are scattered from a sample. A real sample is likely to contain a region of interest surrounded by 'clutter'. A more meaningful result would be obtained if diffraction data could be collected for the region of interest only and the clutter could be ignored. The aim of this project is to design and build a collimator that can be used to mask the detector such that only data from a specific region within a sample are collected. The student will be encouraged to use a combination of modelling and CAD to develop a series of designs which can then be built using the rapid prototyping capability in the Make Space. The student will characterise the prototype designs using modelling and optical techniques.

Student: (Project available)

Ray trace modelling of X-ray diffraction

Supervisors: Robert Speller and Francesco Iacoviello

Tissue diffraction using X-rays can potentially play an important role in the diagnosis of disease. We are currently studying how this technique can best be applied in the detection of early breast cancer. However, to support our experimental studies we are developing modelling techniques. This project is to look at using Matlab or IDLto model different diffractometer designs. It requires an interest and some experience in computing. The eventual aim will be to see if we can reproduce our experimental results.

Student: (Project available)

X-ray diffractometer design for tissue evaluation in breast cancer surgery

Supervisors: Robert Moss and Robert Speller

Recurrence of breast cancer following removal of the primary tumour is still disappointingly high. It is thought to be due to inadequate removal of disease; particularly infiltrating and difficult to image strands of tumour cells. X-ray diffraction may hold the key to being able to identify tissue involvement that cannot be seen with normal techniques. This project is to look at different aspects of the design of a system that could be used to measure diffraction signals on tissue. The project will involve designing different collimators and making measurements of the sensitivity of the system. It is also hoped that the student will be able to evaluate the effect of the positioning of samples within the sensitivity volume and hence allow an optimised system to be developed. Experimental skills will be important and the ability to use Matlab (or similar) an advantage.

Student: (Project available)

Comparison of different X-ray imaging techniques in breast cancer

Supervisors: Robert Speller and Robert Moss

We have a database of ~100 X-ray images of breast cancer samples. These have been recorded in a variety of ways - conventional absorption imaging on a clinical system, absorption imaging on a laboratory system and using a laboratory phase contrast imaging system. We also have a sub-set of these images (~40) that have been ‘imaged’ using a novel X-ray diffraction technique. The aim of the project is to compare the results from all these techniques to identify which of the techniques provides the most information for the different stages of cancer development as seen in the 100 samples. For example, an initial stage in the project might look at finding features in the diffraction data that can be quantified. The student would then compare these metrics with the equivalent values obtained from the other imaging techniques on the same samples. The project requires some computing skills so that image analysis can be carried out and statistical tests developed.

Student: (Project available)

Development and evaluation of material for an undergraduate lab practical on mechanical properties of materials

Supervisors: Pilar Garcia Souto and Alan Cottenden

The understanding of the mechanical properties of materials is essential in many aspects of engineering, from construction of buildings and bridges to biomedical devices and implants. We need to know how strong a material is, how it reacts to applied forces under different conditions, and to be able to predict under which conditions it would fail. This is of essential to avoid serious personal or material consequences. Our department currently runs a 3-hour laboratory session for first-year biomedical engineering students (in term 2) where they analyze simple real data from compression and tension experiments. Data is acquired using an Instron E3000, a state-of-the-art linear-torsion dynamic test instrument that can perform a range of mechanical tests, such as compression, stretching, and bending. This allows our first-year students to relate experimental data with the theoretical behavior of the materials they are taught about in lectures. The aim of this project to refine, support, and assess the pedagogical efficiency of the laboratory session. Extra material needs to be developed, data collected, and evaluation methods established to accommodate an increased number of students in the lab. The student undertaking this project will learn how to use the Instron E3000, learn about mechanical properties of materials and gain experience of teaching, and trial the developed material with a group of students. The student needs to be able to perform data collection and analysis and have some knowledge of Matlab.

Student: (Project available)

Identification of core temperature using infrared imaging: comparison of manually and automatically extracted data

Supervisors: Pilar Garcia Souto and Adam Gibson

During the outbreak of infectious diseases (e.g. SARS in 2003, the Influenza A pandemic in 2009, and Ebola in 2014) core temperature screening was used to detect individuals with fever with the aim of isolating infected individuals, which could help to contain the spread of such infectious diseases. In high transit places such as airports and hospitals, infrared (IR) thermometry has been used for screening as it is relatively easy to use, quick, and non-invasive. This method estimates the core temperature from measurements of heat (in form of IR radiation) emitted from the skin. However IR devices are prone to significant error due to incorrect use by personnel who have had insufficient training. Typical mistakes are the incorrect identification of the area of interest within the body, or the erroneous interpretation of the measurements. The student undertaking this project will compare the manual and the automatic approach to region-of-interest identification and the application of models. Front and/or side views of human heads with an IR camera will be provided. The aims of the project are: 1) Generate appropriate models using the automatically extracted data; 2) Compare the efficiency of models using manually and automatically extracted data; 3) prepare a report of publication quality for possible submission to a journal. The project builds on two previous studies that can be made available to the student undertaking this project. Prior knowledge of computer programming and statistics would be useful.

Student: (Project available)

Neural Networks for Learning Brain Atrophy Patterns in Neurodegenerative Diseases

Supervisors: Marco Lorenzi and Tom Vercauteren

There are currently 850,000 people in the UK with dementia, for an overall cost of £26 billions a year, and at the present moment there are no effective treatments that can stop the spread of damage in the brain. There is therefore a pressing need for computational frameworks enabling the understanding of dementia along the whole disease time course. Quantification of brain atrophy from time series of brain’s magnetic resonance images (MRIs) has been shown to be a sensitive marker of the disease. Nevertheless, neuroimaging is a very challenging area, due to the high dimensionality and variability of the data. Novel and efficient computational methods are necessary to advance the current understanding and treatment of dementia. In this project we will explore machine learning approaches to extract and model the brain atrophy from time series of MRIs. In particular we will design, implement and test neural networks for the analysis of brain longitudinal changes in single patients and in clinical cohorts. The student will be followed during the different steps of the project: preprocessing of brain imaging data with standard neuroimaging tools, design of the neural networks architectures (with pen and paper), algorithm implementation and optimisation (typically Python or Matlab), and testing on clinical data. We envision that this approach will allow assessing disease severity for diagnostic and therapeutic purposes. The framework will be developed on publicly available large imaging datasets in dementia (ADNI), and the project will be hosted by the Translation Imaging Group (TIG) within the Centre for Medical Image Computing (CMIC). Skills Required: Machine Learning, Statistics, Calculus, Numerical Optimization, Programming (Python or Matlab)

Student: (Project available)

The use of proton radiography to optimise the HU-RSP curve for proton therapy patients

Supervisors: Paul Doolan and Adam Gibson

In the planning process for proton therapy patients are scanned using X-ray CT. This provides an anatomical dataset of the patient, but the units are representative of X-ray attenuation (Hounsfield units, HU). To plan the treatment these must be converted to units relevant to proton therapy (namely the relative stopping power, RSP). This is typically conducted using a single piecewise linear conversion curve (the HU-RSP curve). This curve is subject to a number of uncertainties: (i) there is not a one-to-one relationship between HU and RSP; (ii) this curve is based on literature tissue compositions for healthy adults and is not specific to the patient being treated; and (iii) the I-values of tissues are not well known. The most obvious and elegant solution would be to acquire a proton CT of the patient, so the dataset would not require conversion, but this is not currently possible. However, previous work has shown that a single proton radiograph in the direction of intended treatment can be used to optimise the curve. Previously, assumptions had been made such as assuming a straight line for the proton path. This project will investigate the use of Monte Carlo simulations together with real proton radiographic images, to optimise the calibration curve.
The steps of the work include:
• The generation of simulated proton radiographic images using Geant4.
• Extraction of relevant parameters from the simulations to allow for optimisation of the HU-RSP curve.
• Modification of an optimisation module previously-written in Matlab.
If successful, potential areas to expand into include:
• The generation of organ-specific calibration curves.
• Determination of the optimal number of beams required.

Student: (Project available)

Analysing the long-term impact of extreme prematurity: the EPICure@19 neuroimaging study Project 1

Supervisors: Andrew Melbourne and Sebastien Ourselin

Preterm birth is now the world's leading cause of infant mortality and research into the causes and implications of early birth will have a significant social and economic impact. In 1995, the UK began the EPICure study which is following a cohort of extremely-preterm born infants throughout their lives. In 2014/15 we began a major magnetic resonance neuroimaging study on 150 of these young adults and this data represents an opportunity to study brain structure, brain tissue composition (axonal and myelin density) and cerebral blood flow in an unstudied population.
This project will look specifically at estimating myelin density from this neuroimaging data and investigate whether this represents a long-term imaging biomarker of the effects of extreme prematurity.
This project would be most suitable for a student with a strong interest in medical imaging with prior experience in programming and numerical model-fitting. This project will give the student the chance to enhance their numerical programming skills, and learn specifically how the signal generated by MRI can be used to understand tissue composition.
See: http://cmictig.cs.ucl.ac.uk/research/projects/20-neonatal-imaging for more details.

Student: (Project available)

Analysing the long-term impact of extreme prematurity: the EPICure@19 neuroimaging study Project 2

Supervisors: Zach Eaton-Rosen and Andrew Melbourne

Preterm birth is now the world's leading cause of infant mortality and research into the causes and implications of early birth will have a significant social and economic impact. In 1995, the UK began the EPICure study which is following a cohort of extremely-preterm born infants throughout their lives. In 2014/15 we began a major magnetic resonance neuroimaging study on 150 of these young adults and this data represents an opportunity to study brain structure, brain tissue composition (axonal and myelin density) and cerebral blood flow in an unstudied population.
This project will look specifically at estimating properties of tissue microstructure from Diffusion Weighted MRI (e.g. the fractional anisotropy) from this neuroimaging data and investigate whether this represents a long-term imaging biomarker of the effects of extreme prematurity.
This project would be most suitable for a student with a strong interest in medical imaging with prior experience in programming and numerical model-fitting. This project will give the student the chance to enhance their numerical programming skills, and learn specifically how the signal generated by MRI can be used to understand tissue composition.
See: http://cmictig.cs.ucl.ac.uk/research/projects/20-neonatal-imaging for more details.

Student: (Project available)

Analysing the long-term impact of extreme prematurity: the EPICure@19 neuroimaging study Project 3

Supervisors: Eliza Orasanu and Andrew Melbourne

Preterm birth is now the world's leading cause of infant mortality and research into the causes and implications of early birth will have a significant social and economic impact. In 1995, the UK began the EPICure study which is following a cohort of extremely-preterm born infants throughout their lives. In 2014/15 we began a major magnetic resonance neuroimaging study on 150 of these young adults and this data represents an opportunity to study brain structure, brain tissue composition (axonal and myelin density) and cerebral blood flow in an unstudied population.
This project will look specifically at estimating brain tissue volume from MRI using computational segmentation routines and investigate whether this represents a long-term imaging biomarker of the effects of extreme prematurity.
This project would be most suitable for a student with a strong interest in medical imaging with prior experience in programming and numerical model-fitting. This project will give the student the chance to enhance their numerical programming skills, and learn specifically how the signal generated by MRI can be used to understand tissue composition.
See: http://cmictig.cs.ucl.ac.uk/research/projects/20-neonatal-imaging for more details.

Student: (Project available)

Testing the robustness of atlas based synthetic CT methods to different MR properties

Supervisors: Jorge Cardoso, Ninon Burgos, Jamie McClelland, Antje Knopf and Maria Schmidt

We have previously developed a method of generating synthetic CT scans from a database of MR scans using a multi-atlas and label fusion based approach. The synthetic CT scans will be used to plan radiotherapy for prostate and gynae patients without the need for acquiring a planning CT scan. Our initial work on this used high quality MR images from a diagnostic scanner. However, the goal is to apply this method to MR scans that will come from the hybrid MR-Linac system currently under development, which is expected to have compromised image quality. The aim of this project will be to test how well these methods work with MR images of different contrast, lower resolution or lower signal to noise ratio, and to determine the best way of adapting or modifying the current approach to work with this data. This project is suitable for a student with an interest in Radiotherapy and image processing. The student should have good computing skills and experience with Matlab.

Student: (Project available)

Combining atlas based methods for auto-delineation of organs at risk and generating synthetic CT scans

Supervisors: Jamie McClelland, Ninon Burgos, Jorge Cardoso and Antje Knopf

We have previously developed methods for generating automatic delineations of organs at risk and synthetic CT scans for use in planning Radiotherapy. Both of these methods use the same multi-atlas and label fusion based approach, but have been implemented as separate methods. The aim of this project is to combine these methods into a single unified approach that will generate both a synthetic CT scan and the auto-delineations of the organs at risk, given an MR scan from a new patient. This project is suitable for a student with an interest in Radiotherapy and image processing. The student should have good computing skills and experience with Matlab.

Student: (Project available)

Assessing the Biomechanical Properties of Carotid Arteries in Spontaneous Coronary Artery Dissection

Supervisors: Baris Kanber , Kumar Ramnarine, Ferran Prados

Spontaneous Coronary Artery Dissection (SCAD) is a serious medical condition that can lead to a heart attack and sudden death. SCAD often occurs without any warning signs and people who develop SCAD are often otherwise healthy, not having the typical risk factors for heart disease such as high blood pressure or diabetes. The aim of this project will be to test the hypothesis that patients who have SCAD have biomechanically more flexible and mobile carotid arteries than patients who do not. The clinical implication is that if the hypothesis is found to be true, this may lead to a biomarker that can help identify people who might be at risk of developing SCAD in the future. Such a biomarker could enable preventative measures to be taken, thus saving lives.
The student will use and develop computational methods to analyse ultrasound image sequences of carotid arteries of patients with and without SCAD. The student will then employ statistical methods to determine whether the flexibility and mobility of carotid arteries differ between the two groups. Receiver Operating Characteristic curves or other tools then may be employed to determine the sensitivity and specificity of any potential biomarkers for predicting the presence of SCAD, while taking into account other factors that may affect the biomechanical properties of carotid arteries.
The following skills will be required/developed throughout the course of the project:
Literature searching and reviewing, image analysis, computational skills, statistics, attention to detail.

Student: (Project available)

Automated Photoreceptor Cell Identification in Adaptive Optics Images

Supervisors: Christos Bergeles , Adam Dubis

The modern ophthalmic clinic is changing at a rapid speed, with the addition of pharmaceuticals to treat eye disorders and ever increasing reliance on imaging to guide care. With these advances comes the need for better understanding of the pathological changes of diseases on a cellular level as well as improved non-invasive tools for identifying the best candidates for given therapies and monitoring the efficacy of those therapies. Ophthalmic adaptive optics (AO) is a technique to compensate for the eye's aberrations and provide nearly diffraction-limited resolution. The result is the ability to visualize the living retina with cellular resolution, and identify, quantify, and analyse the condition of its photoreceptors. AO is an unquestionably powerful research tool, with multiple companies currently developing commercial devices. The major hold up to adaptation of the technology is the time and skill required to analyse and interpret images. Imaging abilities and image quality are constantly improving, but software analysis is still lacking.
This M.Sc. project will be a collaboration between the UCL Institute of Ophthalmology, Moorfields Eye Hospital, and the UCL Translational Imaging Group, with the goal to develop a reliable automated cell identification algorithm. The tasks in this M.Sc. are: (1) Understand the state of the art in automated algorithms for retinal cell segmentation in adaptive optics images;
(2) Implement and characterise the performance of an existing algorithm or propose a novel image processing methodology;
(3) Assist in experimental data gathering and deploy the algorithm in a clinical setting.
The student will learn about the state-of-the-art technology of adaptive optics, will get a thorough understanding of cell segmentation techniques in microscopy images, and will contribute towards the clinical deployment of patient-critical automated cell-segmentation technologies.
This project is suitable for a student with an interest in microscopy/optical technologies, computer vision, and computer science/software engineering.

Student: (Project available)

Characterization of energy-resolved single-photon-counting x-ray detectors for application in phase-contrast imaging

Supervisors: Marco Endrizzi , Alberto Astolfo

Single-photon-counting detectors offer unique capabilities for x-ray imaging in terms of noise performance and energy resolution. This project involves a complete characterization of a detector based on this technology encompassing classical parameters such as noise and spatial resolution, but also more sophisticated and method-specific ones like signal spill-out between adjacent pixels and detector response as a function of energy. The parameters extracted with this analysis will enable the quantitative interpretation of the acquired data and the inclusion of the realistic detector performance into simulation software. Some familiarity with the basics of data acquisition and analysis (e.g. Matlab, ImageJ) will be useful.

Student: (Project available)

Optimisation of data reconstruction for laboratory-based X-ray phase-contrast imaging using indirect conversion detectors

Supervisors: Marco Endrizzi , Paul Diemoz

Large area, indirect conversion detectors are an attracting option for the translation of X-ray phase-contrast imaging into a standard laboratory tool. One of the prototypes available in the X-ray phase-contrast lab is based on such detector technology and, due to the shape of its point spread function, it requires the use of a “line-skipped” design mask. In practice, this means that only every second column of the detector is used and half of the data collected are simply not used. This project is focussed on the inclusion of this information into the standard reconstruction algorithms with the aim to provide images of higher quality at no additional cost. The student will develop an understanding of the basic principles of image formation and data collection. Existing software will be used to simulate the experimental set-up and, once an optimized data processing strategy is identified, it will be tested on experimental data. Familiarity with programming (e.g. Matlab) will be helpful.

Student: (Project available)

Comparison of two x-ray phase-contrast techniques through simulations of image contrast and signal-to-noise ratio.

Supervisors: Paul Diemoz , Marco Endrizzi

X-ray phase-contrast imaging (XPCi) techniques have emerged in recent years, which are based on exploiting interference and refraction effects, instead of x-ray absorption, to generate image contrast. The strong research interest they have attracted is due to their ability to increase the image contrast, which enables the detection of features invisible to conventional x-ray methods.
The simplest XPCi technique, free-space propagation (FSP), is based on moving the detector further away from the sample, so as to produce interference fringes that can be recorded by the detector. The XPCi group at UCL, instead, has been developing and investigating a more advanced method, called edge illumination (EI), which requires the additional use of two absorption masks before the sample and the detector, respectively.
The aim of this project is to quantify the improvement in contrast and signal-to-noise ratio provided by the EI technique compared to FSP, for a variety of different acquisition parameters (ex. feature size, source size, x-ray energy, etc.), and to determine under which conditions the advantages of EI are largest. The comparison will be based on existing simulation software developed within the group. If time allows, some experimental data will be also made available to validate the results of the simulations.
The student will have the opportunity to learn how to use a matlab simulation code, understand the different metrics defining the image quality, analyze simulation data and ultimately compare simulated and experimental results.

Student: (Project available)

Wave and ray-optics approaches to the modelling of x-ray phase contrast imaging

Supervisors: Paul Diemoz , Alessandro Olivo

X-ray phase-contrast imaging (XPCi) allows the generation of images with highly improved contrast compared to conventional x-ray imaging techniques. While the latters are based on measuring the attenuation of a photon beam when passing through different parts of the sample, XPCi exploits the interference/refraction effects experienced by the photons. Our group developed a new implementation of XPCi, the edge illumination (EI) technique, which was proven to work efficiently with standard x-ray sources and laboratory equipment. It has therefore great potential for applications in many fields of x-ray investigation, such as materials science, biomedical and clinical imaging.
There are two basic ways to model x-ray phase contrast imaging. The wave-optics approach, based on Fresnel/Kirchoff diffraction integrals, is very rigorous but time consuming. The ray-tracing approach, instead, offers a substantial simplification, however at the cost of accuracy. The phase contrast group at UCL has developed software to simulate phase contrast images following both approaches. The student will be provided with this software and will use it to investigate how the two models compare in different experimental situations, and establish under what set of conditions ray optics can be considered a satisfying approximation. The student will gain skills in simulation methods, data analysis, image analysis and familiarize with the basic concepts of optics. A reasonably sound mathematical background is required.

Student: (Project available)

Optimization of experimental parameters for x-ray imaging with edge-illumination

Supervisors: Peter Modregger , Alessandro Olivo

Modern x-ray imaging with the edge-illumination technique offers the exciting possibility to take advantage of the generally superior phase contrast over the traditional absorption contrast.
This project is aimed at optimizing specific experimental parameters for dose efficiency, which will be vital for the translation of edge-illumination into a tool for medical diagnostics. It requires the student to develop an understanding of image quality assessment, the physics of the image formation process, and the numerical simulation thereof. It is expected that the entirety of the project consists of numerical simulations.
Prior experience with Matlab or Python is required.
Knowledge in optics/imaging is useful.

Student: (Project available)

Evaluate different deconvolution methods for x-ray scattering with edge-illumination

Supervisors: Peter Modregger , Fabio Vittoria

Imaging the ultra-small angle x-ray scattering distribution with edge-illumination provides access to sub-pixel information, which implies the tantalizing opportunity to increase pixel sizes in diagnostic x-ray imaging and consequently to decrease dose significantly. In the post-detection processing the scattering signal is retrieved by deconvolution, which is a numerically challenging task. This project is aimed at comparing different methods for numerical deconvolution of the specific signals generated by x-ray imaging with edge-illumination. It is expected that the entirety of the project will consist of numerical simulations.
Prior experience with either Matlab or Python as well as numerical mathematics is useful.

Student: (Project available)

X-ray phase contrast imaging for detecting explosives: data collection and analysis

Supervisors: Alberto Astolfo , Alessandro Olivo

UCL, in collaboration with Nikon, is developing an X-ray phase contrast imaging system to improve sensitivity and specificity in security scans by exploiting the unique advantages of the technique. The system will have a 15x25cm2 field of view that will be the larger available in the world for X-ray phase contrast imaging. This opens the possibility to test the technique in a new range of applications that can go beyond the security field (e.g. medical, pre-clinical, quality control).
The project consists on acquiring images of threat and non-threat materials, as well as of other samples, using a new setup, and then performing some analysis on the collected data. The system will be available at UCL between Nov 2015 and Jan 2016. Prior knowledge of IDL/Matlab and some familiarity with the principles of x-ray imaging are advisable.

Student: (Project available)

Exploring the use of iterative algorithms for x-ray phase contrast CT reconstructions

Supervisors: Charlotte Hagen, Anna Zamir

X-ray Phase Contrast imaging (XPCi) stands for a class of radiographic imaging techniques, which, in addition to x-ray attenuation, are sensitive to phase and refraction effects. XPCi techniques are especially important for the imaging of weakly attenuating biological samples and are investigated by an increasing number of groups worldwide, including the phase contrast group at UCL.
Edge Illumination (EI) XPCi - a novel method developed at UCL – can measure the refraction of x-rays as they pass an object. EI XPCi has recently been implemented as tomographic modality. By rotating the object over an angular range of at least 180 degrees and acquiring images at every rotation angle it is possible to reconstruct volumetric maps of the refractive index distribution within the object. While until now all tomographic reconstructions were carried out using filtered back projection, the next step is to explore the benefit of iterative algorithms for the reconstruction of these maps. The student would be given an experimental EI XPCi dataset on which different iterative reconstruction algorithms can be tested. The student would define metrics that can be used for a comparison and eventually decide which algorithm is most suited to the reconstruction problem at hand.
Besides being involved in the development of a new imaging method, the student would get familiar with the basic concepts of computed tomography and image reconstruction. The project requires advanced mathematics and experience in Matlab programming. Existing software packages (e.g. the ASTRA toolbox) can be used.

Student: (Project available)

Rigorous simulation of light focussed through layers of materials with differing refractive indices

Supervisors: Peter Munro , James Guggenheim

The focussing of light through layers of materials with differing refractive indices is important in numerous biomedical imaging techniques, including optical microscopy, photoacoustic tomography and optical coherence tomography. The aim of this project is to develop a toolbox using Matlab that is easily used by researchers in a variety of fields. This project will require understanding of both optics and computer programming using Matlab. This project will suit students with an interest in the simulation of physical phenomena with minimal approximations.

Student: (Project available)

Optimal simulation of X-ray propagation using wave optics

Supervisors: Peter Munro , Fabio Vittoria

Wave optics is the most general way to model the interaction of X-rays with tissue. The problem is that wave optics can be computationally intensive, which is why a ray optical is often employed. This project will look at some ways in which a wave model of X-ray propagation can be made as computationally efficient as possible, using mathematical rather than programming optimisations. This project will underpin new approaches to image restoration and quantification in X-ray phase imaging.

Student: (Project available)

GPU modelling of wave optics

Supervisors: Peter Munro , Fabio Vittoria

The simulation of the propagation of optical waves (including X-rays) requires that a large number of very similar integrals be performed. This problem is thus highly suited to parallelisation using a graphics processing unit, which may be able to compute thousands of such integrals in parallel. This project will seek to establish framework upon which a toolbox will be build. The first part of the project will be to determine the most appropriate programming tool to use and the second part will be to implement a proof of principle demonstration.

Student: (Project available)

Construction of phantoms which mimic both X-ray and optical scattering

Supervisors: Peter Munro , Peter Modregger

Light scattering is one of the factors which limit the depth at which optical imaging may be performed in tissue. Techniques which rely on the propagation of light deep into tissue (e.g. diffuse optical tomography and photoacoustic tomography) benefit immensely from mapping of the strength of optical scattering throughout the tissue being imaged. This project is aimed at developing phantoms to demonstrate that optical scattering can be correlated with the scattering of X-rays, which has only recently been demonstrated to be possible in the laboratory. This project will require a research into suitable materials and the development of a protocol to construct the phantoms.

Student: (Project available)

Integrating bio-imaging methods. A single non-toxic sample preparation protocol for electron, fluorescence, bio-raman microscopy and macroscopic imaging methods.

Supervisors: Sergio Bertazzo , Adrien Desjardins

The wide array of bio-physical requirements for biological imaging using different methods (such as electron and fluorescence microscopy, photoacoustics, among others) inadvertently creates obstacles to a full integration and correlation of all bioimaging methods. In this way, a synergistic and holistic characterization of bio-medical systems cannot be achieved. We propose to develop in this project a single novel protocol whereby we will be using non-toxic reagents and inorganic polymers to prepare biological samples to be imaged sequentially by all major bio-imaging methods, from the nano to the macro (meter) scale. The possibility to use a single non-toxic protocol for sample preparation will allow us to select macroscopic regions of interest within a sample and then further investigate each region by different methods, consolidating the information of biochemical composition and structure, including ultrastructure. This new protocol, which we discovered by serendipity, is a truly exciting way to prepare biological samples for imaging. It will open unforeseen opportunities to integrate the information provided by different imaging methods into a single comprehensive overview of biological systems, leading to unprecedented advances in biomedical research.
The student will participate in the development of the project, sample collection and preparation, experimental measurements, analysis of the data and interpretation. Initially the project will make extensive use of electron microscopes, with new techniques potentially added as the research work progresses. This is an experimental research project where laboratory experience is a plus but is not a requirement.
This project will be carried out in close collaboration with several groups within the Department of Medical Physics and Biomedical Engineering and the groups of Dr Inge K. Herrmann at the EMPA (Swiss Federal Laboratories for Materials Science and Technology), as well as other international research groups.

Student: (Project available)

Unusual biomineralization systems

Supervisors: Sergio Bertazzo , Susan Evans

Biomineralization is a fundamental process present in several animals and plants. The structures produced by this process surround us and are clearly recognized in the form of shells, bone and teeth. Today it is well known that the most common biomineralized structures are composed of silica (diatoms), calcium carbonate (mainly present in shells in general) and calcium phosphate (mainly present in teeth and bone).
The nanostructure of bones from mammals, birds and reptiles is also well known, a well-defined distribution of collagen fibres and calcium phosphate crystals. This nanostructure, visible with electron microscopy, provides a clear fingerprint of bone "material".
Interestingly, the same methods currently applied to the study of bone have not yet been extensively transferred to the study of other biomineralization systems, such as the osteoderms found in the skin of many reptiles, or even to special bones, such as the elements of the sclerotic ring that supports the eye in reptiles and birds.
In this project, we intend to use state-of-the-art electron microscopy techniques to study and characterize the nanostructure of unusual biomineralization systems and compare them with regular mammalian/bird/reptile bones. The new information about these systems will not only improve our understanding of how a wide variety of organisms produce hard tissues, but could also lead to the creation of new materials bioinspired by this data.
The student will participate in the development of the project, sample collection and preparation, experimental measurements, analysis of the data and interpretation. Initially the project will make extensive use of electron microscopes, with new techniques potentially added as the research work progresses. This is an experimental research project where laboratory experience is a plus but is not a requirement.

Student: (Project available)

Investigating the Mechanisms of Cancer Calcification Using State of the Art Physical Chemistry Characterisation.

Supervisors: Sergio Bertazzo , Alessandro Olivo

The project is related to the study of calcification present in tissue samples from breast cancer patients. Breast cancer is the commonest cancer in women in the UK. The calcified material found in cancerous breast tissue has never been thoroughly characterized and there is no study in the literature comparing its characteristics with bone or with regard to calcification degree, site of origin and type of cancer. In this project, we will use for the first time cutting-edge physical-chemistry characterisation techniques, such as focused ion beam (FIB) and electron microscopy (SEM-EDX and TEM), to determine crucial variations in the nature of the calcified material found in cancerous tissue. This innovative research will build on the comprehensive skill set of the applicant and take it in a new and exciting direction that focuses on truly medically important breakthroughs, such as the mechanisms for cancer calcification and novel diagnosis.
The student will participate in the development of the project, sample collection and preparation, experimental measurements, analysis of the data and interpretation. Initially the project will make extensive use of electron microscopes, with new techniques potentially added as the research work progresses. This is an experimental research project where laboratory experience is a plus but is not a requirement.
This project will be carried out in close collaboration with several groups within the Department of Medical Physics and Biomedical Engineering and with the group of Dr Luca Magnani from the Department of Surgery and Cancer at the Imperial Centre for Translational and Experimental Medicine, Imperial College London.

Student: (Project available)

How long can DNA survive in the fossil record?

Supervisors: Sergio Bertazzo , Susannah Maidment

This project is related to one of the most exciting scientific paradoxes existing today, related to the presence of organic material in fossils.
It is a common claim in the literature that proteins, particularly DNA, cannot be found in animal/plant remains over 100,000 years old. On the other hand, several independent research teams have recently presented evidence of soft tissue preservation over timescales of millions of years. To examine this paradox, we will determine whether there is an upper time limit for the preservation of organic materials in fossils. With this information, we will use state-of-the-art of micro-manipulation to extract organic material from fossils and sequence the possible protein/DNA remains. Finally, we will also try to understand the kinetics of degradation of proteins/DNA by simulating the conditions found in the fossilization process.
The student will participate in the development of the project, sample collection and preparation, experimental measurements, analysis of the data and interpretation. Initially the project will make extensive use of electron microscopes, with new techniques potentially added as the research work progresses. This is an experimental research project where laboratory experience is a plus but is not a requirement.
This project is a joint research effort with the groups of Dr Susannah Maidment of Imperial College London and Dr Inge Herrman of the Swiss Federal Laboratories for Materials Science and Technology. The project will also benefit from strong collaboration of the Principal Investigator with several other international groups.

Student: (Project available)

Ultrasound imaging probe - Apprentice piece design.

Supervisors: Rebecca Yerworth , Daniil Nikitichev & Danial Chitnis

Biomedical engineering students need to learn basic electronics during their first year. To improve their understanding and retention of the material, and demonstrate the context within Biomedical Engineering, we are proposing that they build a simplified medical instrument, following detailed instructions. One candidate for this is a low frequency ultrasound probe. The aim of this MSc project is to design and prototype such a device, with a focus is on creating drive/receive circuits that demonstrate use of op-amps, analogue filtering ect. This project is most suitable for a student who has experience of electronic engineering/medical electronics.

Student: (Project available)

Harnessing the Rainbow – the fabrication of multimodality probes for ultrasound and photoacoustic imaging

Supervisors: Sacha Noimark ,Richard Colchester

Optical ultrasound transmission shows increasing promise as tool to guide interventional surgical procedures and reduce the incidence of complications. Recent studies have indicated that strong candidates for efficient laser-­‐generated ultrasound are composite coatings on optical fibres comprised of an elastomeric host (PDMS) and distinct optical absorber. Ultrasound generation is achieved via the photoacoustic effect; absorption of pulsed laser excitation and rapid conversion to heat within the composite coating, results in a pressure rise that propagates as an acoustic wave. Comparable in terms of ultrasound pressures achieved, these devices demonstrate key advantages over current generation piezoelectric devices in terms of bandwidth, device miniaturisation and immunity to electromagnetic interference.
Coatings developed for optical ultrasound transmission include metals such as gold or chromium and a range of carbonaceous materials, such as carbon black, carbon nanofibres and carbon nanotubes. However, to extend these devices to multiple modalities, the consideration of wavelength dependent absorbers is crucial. The development of selective-­‐absorbing coatings could lead to the fabrication of integrated transmit-­‐receive devices with direct uses in photoacoustic imaging and optical coherence tomography.
This interdisciplinary project demonstrates a unique blend of Chemistry and Medical Physics, allowing students to develop a broad skill set. Coatings for ultrasound generation will be synthesised using an elastomeric component and an industrially-­‐engineered pigment, formulated to selectively absorb at particular wavelengths, with almost complete transparency in the IR. The project will allow for an element of blue-­‐sky thinking in terms of coating deposition strategies, with possibilities including microplotting, dip-­coating or electrospinning. Full characterisation of the coatings will be achieved, including transmission measurements, scanning electron microscopy and ultrasound generation efficiency. Once optimised, the fabricated devices will be used for multimodality imaging of phantoms and ex vivo tissue. The candidate must demonstrate a keen interest in learning a wide range of skills from device fabrication and material characterisation to data analysis and it is preferable that the candidate demonstrates basic skills in MATLAB and LabView.

Student: (Project available)

Computational modelling of cerebral autoregulation using multimodal clinical data from adult acute brain injury

Supervisors: Matthew Caldwell and Ilias Tachtsidis

Multimodal monitoring of critically ill brain injured patients provides information that can help clinicians tailor treatment decisions to individual patient needs. The data from such monitoring is complex and often difficult to interpret, reflecting the interplay of many physiological, pathophysiological and environmental factors. Computational models of brain physiology can identify clinically-relevant relationships between the measured variables and the underlying health state. In this project, an existing computational model will be used to simulate brain states based on anonymized clinical data, comparing the model predictions to measured variables in order to improve the quantification of cerebral autoregulation. The project would suit a student with an interest in computational modelling and data analysis.

Student: (Project available)

Using MRI Susceptibility Mapping to Investigate the Brain Iron Content of Tanzanian Sickle Cell Anaemia Patients

Supervisors: Karin Shmueli and Fenella Kirkham, Jamie Kawadler

Sickle Cell Anaemia (SCA) is a genetic disorder affecting the haemoglobin molecule. In SCA the red blood cells change to an abnormal and fragile ‘sickle’ shape making them unable to deliver enough oxygen to the body’s tissues. SCA has the highest prevalence in sub-saharan Africa and affects 1 in 2000 UK births. SCA patients have a lower iron-containing red blood cell count than healthy controls and are not orally supplemented with iron, therefore we hypothesise that the iron content in some iron-rich deep-brain regions is lower in non-transfused SCA patients than in healthy controls. Tissue magnetic susceptibility derived by MRI susceptibility mapping, a recently developed technique, has been shown to correlate strongly with tissue iron content. Our previous work using susceptibility mapping in a UK cohort has suggested that the iron content in some deep-brain regions is lower in SCA patients than in healthy controls. We now have a large set of susceptibility-weighted MR images from a group of SCA patients in Tanzania. Therefore, the aim of this project is to use MRI susceptibility mapping to investigate the brain iron content of SCA patients in Tanzania. It is clinically important to measure brain tissue iron content in SCA patients as this could inform their future treatment.
This project would be most suitable for you if you have a strong interest in medical imaging with prior experience in programming. You will use Matlab scripts and other software to process clinical susceptibility-weighted gradient-echo MRI images of the brain acquired in Tanzania. You will calculate susceptibility maps from phase images and use region-of-interest analysis to assess changes in tissue magnetic susceptibility values. This project will give you a chance to enhance your image processing and programming skills and learn how the MRI signal can be used to probe tissue composition and answer clinically relevant questions.

Student: (Project available)

Probabilistic Programming for Medical Imaging

Supervisors: Pankaj Daga and Tom Vercauteren

Description: Most medical image analysis problems are formulated as optimization problems where one is interested in finding the local maximum-a-posteriori (MAP) estimate of the posterior distribution. However, MAP estimates are not representative of Bayesian inference methods in general. This is because they are point estimates and do not use the full posterior distribution. The attractiveness of MAP estimates lies in the computational tractability as compared with the Bayesian alternative of characterising the full posterior distribution.
To address this, various inference schemes like Laplace approximation, Variational Bayes (VB), Expectation propagation (EP) and others have been designed to provide approximations to the full posterior distribution in a computationally tractable way. The proposed project aims to develop a software framework to perform the inference in the model agnostic way. We will use existing schemes to describe the probabilistic models (https://github.com/pymc-devs/pymc) and implement the inference schemes with a clear separation between model definition and inference.
Using the methods described above, we will develop models to tackle the problem of medical image registration. The initial aim is to develop probabilistic models for global image registration. We will apply these models for interventional brain MRI registration and registration of endoscopic video frames acquired during fetal intervention.
Skills required: Interest in Bayesian statistical methods. Background in numerical methods and Python programming.

Student: (Project available)

Disrupting Costly Eye Examinations in Developing Economies

Supervisors: Christos Bergeles and Luis C. Garcia-Peraza-Herrera

Given the extensive worldwide population suffering from a potentially blinding ophthalmic pathology, innovation on intraocular observation methods is imperative. In rural societies, patients that suffer from detectable and treatable diseases, such as cataracts or retinopathy of prematurity, would benefit from easily accessible digital ophthalmoscopes. Notably, 80% of blindness is preventable when detected early. Disruptions in the miniaturisation of lenses, electronics, and mechanical components, together with ubiquitous computing through microprocessors, can instigate developments in ophthalmoscopy and reshape a field substantially based on 20th century developments.
Retinal fundus imaging is a critical task in ophthalmoscopy, and, thus, the focus of recent device innovations. Examples of new approaches to fundoscopy mainly make use of smartphone technologies. Initial approaches in 2012 entailed observing the retina using a smartphone's camera and a handheld indirect ophthalmoscopy lens. A limiting factor of smartphone-based approaches, however, is that they do not account for the expense of the device itself. Thus, the reported costs are misleading and may amount to the cost of a hand-held commercial digital ophthalmoscope. Furthermore, the fact that 90% of blind people live in low-income countries is not considered. Their location should be examined together with the lack of smartphone penetration, e.g., only 37% of population in China, and only 19% in Kenya owns a smartphone. Contrary, truly widespread ophthalmoscopic screening will make use of ubiquitous technology, such as miniaturised inexpensive computers like the Raspberry Pi.
This M.Sc. project is about building exactly such a device: an inexpensive ophthalmoscope that is controlled by tiny computers and embedded electronics. It will consist of state-of-the-art liquid lenses and high-definition cameras, all embedded in a hand-held device. The student will be able to leverage substantial existing progress on such a device, and will be encourage to contribute his own research ideas to shape the project.
The student will learn about state-of-the-art technologies in imaging, and will be able to design and create electronic components. He/She will get the opportunity to deliver computationally efficient code that runs on Raspberry Pi and Arduino, so that all image processing can be performed without the need of an external computer or an expensive smartphone.
This project is suitable for a student with an interest in optical technologies, electronics, and software.

Student: (Project available)

Testing the effect of breathing variability on proton therapy plans and motion mitigation techniques

Supervisors: Jamie McClelland and Richard Amos

Proton therapy could be an effective way to target lung cancer whilst sparing normal lung tissue, but respiratory motion can cause large errors if not suitably accounted for. In recent years a variety of proton beam rescanning methods have been introduced to mitigate the effects of respiratory motion. However, most of these have been evaluated using simulations based on 4DCT datasets, which assume the breathing motion is the same from breath to breath and often suffer from image artefacts.
This project will use computational motion models that have been developed in CMIC that can realistically simulate breath to breath variations in the respiratory motion, to test the effect of these variations on proton therapy plans and the motion mitigation techniques. This may lead to modifications to existing proton beam planning and delivery techniques, or new techniques being developed that are more appropriate for variable breathing motion.
This project would be most suitable for a student with a strong interest in medical image computing, motion modelling, and proton beam radiotherapy.

Student: (Project available)

Optimisation and validation of a respiratory motion model for use on unsorted 4DCT data

Supervisors: Jamie McClelland and Richard Amos

4DCT datasets have become the clinical standard data used for planning radiotherapy to lung cancer patients. However, they are based on the assumption that there are no breath to breath variations in the respiratory motion, which can often occur, and when they do will lead to artefacts in the reconstructed 4DCT images. Additionally, the radiotherapy plans will only be optimised for the motion seen in the 4DCT images, which may not be very representative of the motion that can occur during other breaths.
Recently we have developed a generalised motion modelling framework in CMIC, which can fit a computational model of the respiratory motion and its breath to breath variations directly to the unsorted 4DCT data. This has ability to produce artefact free images which simulate the respiratory motion and its variability, and could lead to radiotherapy plans that are better suited to the actual motion that is likely to occur during treatment. This project will investigate the optimum model and modelling parameters to use when building the models from unsorted 4DCT data, and will validate these using several clinical datasets. This may lead to developing new methods of incorporating the motion variability into radiotherapy plans.
This project would be most suitable for a student with a strong interest in medical image computing, motion modelling, and radiotherapy.

Student: (Project available)

Instrumentation of handrails on a gait analysis treadmill

Supervisors: Julian Henty and Adam Gibson and Mathew Thornton (RNOH)

The Motor Learning Lab at the Royal National Orthopaedic Hospital (RNOH) is a state-of-the-art facility used for gait analysis. The lab contains a treadmill with handrails and a 180°surround screen, providing a walking environment with an interactive feel. This equipment allows functional assessments to be made by measuring forces between the feet and the floor, although many patients also use the handrails for support during walking. Despite the fact that considerable weight is often applied through the handrails the system currently has no means of measuring this. The proposal of this project is to make a minimal modification to the treadmill system to allow instrumentation of the handrails, providing a means of dynamic feedback to assist the reduction of handrail dependence.
The student would be expected to produce technical drawings for the design of metal adaptors, and to assemble a strain gauge array for connection to a laptop. It is envisaged that the modification could be tested with a patient within the time-frame of the project.

Student: (Project available)

Tool for testing ECG defibrillator synchronisation output in hospital monitoring equipment

Supervisors: Julian Henty and Dimtrios Airantzis and Daniel Simmonds (UCLH)

Modern hospital monitoring equipment has a defibrillator output socket to allow a connected defibrillator to provide synchronised cardioversion, which is used to stop an abnormally fast heart rate or cardiac arrhythmia by delivery of a therapeutic dose of electric current to the heart during the R-wave of the cardiac cycle. While it is part of the routine maintenance procedure to test this output, the supplied manufacturer cable requires the use of an oscilloscope to observe the output, which are not normally available to hospital engineers. The proposal of this project is to design and build a device to produce a simulated ECG signal, and to display the monitor output signal from the manufacturer cable. Such a device would be of great assistance to hospital biomedical engineers conducting planned preventative maintenance.
The student would be expected to design a simple microcontroller system, involving both hardware and software skills. It is envisaged that the system would be tested within the hospital environment.

Student: (Project available)

The effect of image reconstruction assumptions on lung volume estimates

Supervisors: Rebecca Yerworth, Richard Bayford

Electrical Impedance Tomography has the potential to fill a urgent clinical need in the management of Acute Respiratory Syndrome as it can provide real-time, long term, bedside monitoring of the spatial distribution of respiration, without the use of ionising radiation. However, estimates of lung volume may be affected by assumptions within the image reconstruction process, specifically those related to the chest shape.
The aim of this project is to assess different reconstruction algorithm methods, which seek to minimise artefacts due to incorrect shape, with respect to the resulting lung volume estimations. The project will involve, mathematical simulations and re-analysing previously collected clinical data, using Matlab.
It will suit a student who has selected MPHY3B27 Computing in Medicine or the M-level variant, MPHYMB27 or has equivalent previous experience.

Student: (Project available)

Machine learning techniques for the improvement of pathological organ segmentation

Supervisors: Maria A. Zuluaga, Michela Antonelli

Medical image segmentation plays a key role in quantifying physiological and pathological changes to the structures of the human body. As such, it is an important first step in many applications including diagnosis, the tracking of the response to treatments, and quantitative medical studies.
Many of the leading approaches for this task are based on multi-atlas segmentation (MAS), a process which involves propagating segmentation information from a database of pre-labelled atlas images to the space of the target. These approaches are therefore heavily reliant on image registration algorithms to achieve meaningful anatomical correspondence between the target image and those of the database. When registration algorithms fail due to large differences in morphology, this can introduce errors into MAS-based segmentations. Efforts to overcome this problem have thus far focused on label fusion methods and post-processing that can be used to adjust the final segmentation.
This project aims to explore the possibility of new approaches to these two tasks based on machine learning. The student will investigate various techniques to determine if and how they can be applied to the problems of label fusion and post-processing for MAS segmentation in cases where registration is confounded by large pathological or genetic variation in anatomy.
Required skills: Good computing skills and programming experience.
Desirable skills: Some experience with machine learning techniques.

Student: (Project available)

Experimental validation of a numerical model for Fabry-Perot ultrasound sensor

Supervisors: Rehman Ansari,Paul Beard

Photoacoustic imaging is an emerging biomedical imaging technique which uses short duration laser pulses to excite molecules of interest (chromophores) in biological tissue, which leads to emission of ultrasonic waves. These ultrasonic waves are recorded and analysed to produce 3D images. Photoacoustic imaging has many clinical applications including detection of cancer, characterization of atherosclerotic plaques, etc.
A highly sensitive and wideband ultrasound detector is essential to obtain good quality and high-resolution images. Our group has developed optical ultrasound sensors based on Fabry-Perot interferometer, which offers several advantages over conventional piezoelectric transducers. Our group has also developed a numerical model for these sensors to maximize their performance. The goal of this project is to experimentally validate the numerical model by measuring the performance of Fabry-Perot sensors, for a variety of different design parameters (spacer thickness, mirror reflectivities, etc.) and Gaussian beam waist diameters. This project will give student a chance to gain experience through hands-on experimental work in an optics laboratory and the opportunity to contribute to the development of a promising new medical imaging modality.

Student: (Project available)

Multi-time-point Atrophy Calculator

Supervisors: Ferran Prados, Manuel Jorge Cardoso

Brain atrophy measured using structural magnetic resonance imaging has been widely used as an imaging biomarker for disease diagnosis and tracking of pathologic progression in neurodegenerative diseases. Recently, it has been presented a generalized and extended formulation of the boundary shift integral (gBSI) using probabilistic segmentations to estimate anatomic changes between 2 time points. This method adaptively estimates a non-binary exclusive OR region of interest from probabilistic brain segmentations of the baseline and repeat scans to better localize and capture the brain atrophy. Despite it has been demonstrated theoretically the transivity of BSI (it should allow to work with multipe time-points), never we have had implemented a version that support this. Then, the purpose of this project is to extend the actual in-house software to allow compute multi-time-point atrophy using the different BSI methods (KNBSI, GBSI and PBSI) in different brain areas (Brain, hippocampus, ventricles,...). This extension will allow the student to explore different topics of the image processing like: segmentation, registration, intensity normalization, design and building pipelines using Nypipe.
The following skills will be required/developed throughout the course of the project:
Programming skills, self-motivation, commitment, and attention to detail.

Student: (Project available)

In silico Modelling of Tumour Induced Angiogenesis

Supervisors: Vasileios Vavourakis, Peter Wijeratne

The process of tumour induced angiogenesis (TIA) – whereby cancerous cells, in response to a hypoxic environment, stimulate the formation of new blood vessels from the existing vasculature – is a crucial step towards cancer growth and metastasis. However, because of the complexity of the biological processes involved, it is difficult to specify the mechano-chemical factors and mechanisms in TIA via experimentation only. In this regard, in silico models have been developed to corroborate the experimental findings, and elucidate further the mechanical and chemical processes of cancer growth and neo-vascularisation.
We have developed a dedicated multiscale Finite Element model, which allows the various physical effects influencing angiogenesis to be analysed separately and in concert. As such, the model can provide insight into the optimal delivery of therapeutic drugs, and in conjunction with medical imaging, potentially point towards novel biomarkers. The student will be directly involved in applying the model using in vitro and in vivo data to verify the model developments, validate the numerical predictions, and optimise the model parameters. A reasonably sound mathematical background is required, while good skills in C/C++ will be very useful.

Student: (Project available)