Fusion of multiple 3D fetal sonography scans
Maryam Ahmed, Linacre College
Fusion of multiple 3D fetal sonography scans can improve the diagnostic field of view and anatomical definition over standard scans as demonstrated in a series of technical and clinical publications by our group. We are interested in the extension of both the methodology and application for fetal medicine applications.
Specifically the short project will involve:
- Working with clinicians on defining scanning protocols for fetal scanning and acquiring clinical datasets (a minimum of 10 studies).
- Evaluation of the existing fusion methodology software (which was developed for cardiology applications).
- Based on (2) extending the current methodology to include automatic identification of good scans for fusion.
If time permits, evaluation of fusion datasets of the same subject by two sonographers to be compared with single conventional scans. The hypothesis is that fusion leads to ultrasound volumes that are invariant to the sonographer.
3D Realtime Imaging for Ultrasound Enhanced Drug Delivery
Steve Bian, Linacre College
Successful treatment of tumours is often difficult because chemotherapy drugs cannot penetrate into the core of tumours. This is especially so for new and powerful drugs such as antibodies which are much larger in size than many older drugs. As a result, a reserve of cancer cells remains untreated and can regrow the tumour after treatment. One promising technique to improve drug penetration is to mix very small bubbles (called microbubbles) with drugs, and then burst the bubbles with a blast of ultrasound. This has been shown to create shockwaves which can push drugs into and throughout tumours. While the benefits of this approach are clear, the processes taking place within tumours exposed to microbubbles, ultrasound and drugs need to be better understood before this method can be used in patients.
The goal of this project is to develop an instrument which will enable the study of ultrasound enhanced drug delivery in near real-time and in 3 dimensions (3D). This will enable direct visualisation of the enhancement process, and provide a better understanding of the mechanisms responsible for enhanced drug penetration, understanding that is crucial for patient safety, drug efficacy, and regulatory approval.
Identifying changes in connectivity in brain networks related to a risk gene for Alzheimer’s Disease.
Giles Colclough, Magdalen College
Non-invasive brain imaging has made valuable contributions to the understanding of Alzheimer's disease (AD) pathology and holds great diagnostic potential. In particular, recent work has investigated the role played by a risk gene for AD brain activity. However, the study of changes in brain anatomy and physiology associated with AD using brain imaging has mainly focused on finding the specific brain areas affected by the disease. This project will investigate the role played by a risk gene for AD on the networks of brain function. Features within these networks have the potential to provide much more sensitive markers of AD. Thereby allowing diagnosis of the disease years ahead of what is currently possible, and allowing for timely therapeutic intervention.
An affordable upper limb prosthesis for India (PURAK)
Vikranth Nagaraja, St Hilda’s College
The innovative design of an affordable prosthesis is just the start of a programme to introduce a new device to enhance the abilities of the poorest amputees in India. Engineering innovation alone is insufficient to overcome the barriers to sustainable adoption of new technology. This project aims to evaluate and maximise the advantage of a new design of upper limb prosthesis over existing solutions. This will require basic experimental work, data from a linked market research project, and also extensive interaction with the clinical providers of upper limb prostheses to the poor in India.
Using Mind Reading to Personalise Treatment in Post- Traumatic Stress Disorder
Kate Niehaus, Trinity College
Psychological traumatic events, such as war or road traffic accidents, are widespread. A small but significant proportion of survivors develop post-traumatic stress disorder (PTSD). Distressing, sensory-based involuntary memories of trauma (henceforth ‘flashbacks’) are the hallmark symptom of PTSD. This project aims to use functional neuroimaging data, in particular Functional MRI (FMRI) and Magnetoencephalography (MEG), to ‘read the mind’ of an individual, to understand the neural basis of flashback memory formation and recall. This will help us determine: i) from neural response when traumatic film footage is being viewed (i.e. memory encoding), to predict if they will cause subsequent flashbacks; ii) from neural response during the experience of having a flashback (i.e. memory recall), which images from the trauma film are being visualized.
The specific nature of the spatio-temporal features leveraged by the brain decoder will provide new clinical insight into the neurological mechanisms involved in flashbacks. Ultimately, this approach could be used to understand and diagnose different phenotypes, i.e. to determine individual PTSD patients who may be most ‘at risk’ of PTSD after trauma. Intriguingly it could also provide the first objective index of ‘seeing’ another person’s flashbacks to help better target treatment interventions.
Learning to Count and Detect Cells with Partial Annotation
Thomas Nketia, Kellog College
In high through-put screening there is a need to detect and count cells of a particular type, for example detecting cancerous cells in histopathology images whilst ignoring other non-cancerous cells. This can be achieved automatically using computer vision and machine learning. However, in order to train the system a user must first provide ground truth annotation of the cells, e.g. by `dotting' the cells of interest in multiple samples. This can be quite onerous.
The purpose of this project is to relieve the burden of complete annotation, and investigate machine learning methods where a user provides only partial annotation, for example by annotating some examples of cells of interest and some examples of those that are not.
Data from existing studies will be used to evaluate the new methods which will be compared against machine learning method(s) using complete annotations.
Delivering siRNA and mRNA to cells with shock waves
Sandra Nwokeoha, Pembroke College
Many diseases result from the faulty production of proteins by cells. Short sequences of nucleic acid can be designed to specifically ‘knock-down’ this faulty production and reverse the disease. Unfortunately delivering nucleic acid into the cells so that it can contact its target is an inefficient process. Here we will investigate the use of shock waves (which are already employed for several other medical applications) as a mechanical force to temporarily open up cell membranes in order to enhance transfer of the nucleotide into diseased cells. If successful this technique would result in more effective therapies with lower side-effects.
Circadian assays for abnormality detection in mental and physical health
George Qian, Brasenose College
This project is connected with the newly established Sleep, Circadian and Neuroscience Institute (SCNi) which is funded by the Wellcome Trust. As a collaboration between Neuroscience, Genetics, Psychiatry, and the Institute of Biomedical Engineering, week seek to establish pathways for circadian rhythm disorders from the genetic mechanisms up to the social behaviours which contribute to and result from such disorders. The project fits within the larger aim of providing the infrastructure to explore the relationship between sleep/circadian rhythm disruption and a sub-set of neuropsychiatric disorders and drive forward translation studies for the improvement of health. Although the initial focus will be on a sub-set of neuropsychiatric disorders, ultimately we hope that the expertise developed within the SCNi will be applied to additional psychiatric problems and encompass a broad range of neurodegenerative disease. This study outlines an integrated approach to define some of the specific mechanisms that give rise to sleep/circadian rhythm disruption in neuropsychiatric disorders and to use this information for the provision of evidence-based approaches for the improvement of health.
As sleep and circadian physiology are dependent upon all the major neurotransmitter systems, it should be no surprise that sleep complaints are reported in more than 80% of patients with either depression or schizophrenia and that sleep abnormalities are common in patients with chronic pain, Alzheimer's and Parkinson's disease. Interestingly, sleep and circadian rhythm disruption often precedes the symptoms associated with these conditions. Disruption of sleep/circadian control will in-turn have wide-spread effects, ranging across all aspects of neural and neuroendocrine function including impaired cognition, emotions, metabolic abnormalities, reduced immunity and elevated risks of cancer and coronary heart disease. This spectrum of pathologies, which frequently arise from sleep/circadian disruption, are also co-morbid with many brain disorders. Yet these co-morbid pathologies are rarely linked to the disruption of sleep. Furthermore, sleep/circadian rhythm disruption will lead to abnormal light exposure and atypical patterns of social behaviour that will feedback to further destabilize sleep/circadian physiology and exacerbate an abnormal pattern of neurotransmitter release within the brain.
Initial summer project work will explore sleep structure, actigraphy, and cardiovascular phase using several large databases (comprising 100’s of patients) of scored clinical data. Signal processing, feature selection and statistical machine learning techniques will be applied to extract clinically useful characteristics for assessing sleep structure and disease symptom severity. Clinical applications will include a variety of sleep related illnesses and behavioural abnormalities such as borderline mid disorder, attention deficit hyperactivity disorder (ADHD), and schizophrenia.
The long range research for a DPhil will involve analysis of social networking behaviour, actigraphy, geolocation, cardiovascular signal phase, and sleep structure. We also hope to connect the behaviours to gene expression and phenotypic categories.
Find the brain chemical ‘fingerprints’ of mental illness
Nick Palmius, Wolfson College
The diagnosis and mental illness and subsequently assessing how successful treatment is being is very challenging. This is largely because mental illnesses such as depression have a wide range of subjective symptoms. In this project we are looking at how we can measure brain chemistry using a non-invasive Magnetic Resonance scanner and using this information to search for the unique ‘fingerprints’ of mental illness.
Improving the accuracy of pharmacokinetic modelling in DCE-MRI using patient-specific parameter estimation and motion correction
André Hallack Miranda Pureza, St Cross College
The short project focuses on some of the important aspects involved in the parameter estimation of pharmacokinetic (PK) modelling of DCE-MRI in patients with colorectal cancer undergoing combined chemo-radiotherapy (CRT). The study will involve assessing the effects of motion on some of the patient-specific parameter estimations required for PK modelling.
Electromagnetic acoustic imaging of lesions formed in tissue by therapeutic ultrasound
Visa Suomi,St Hilda’s College
High intensity focused ultrasound (HIFU) is non-invasive method of destroying cancerous tissue with no significant side-effects. One barrier to its wider spread use is the need for accurate monitoring of which region of tissue has been successfully treated. In this project a novel imaging modality developed at Oxford University, Electromagnetic Acoustic (EMA) imaging, will be evaluated to assess if it can monitor the treatment of tissue by HIFU. It will involve fundamental experiments using gel phantoms followed by evaluation using ex vivo bovine liver.
Mapping of Ultrasound-Enhanced Drug Delivery by Passive Acoustic Mapping
Iulia Popescu, St Cross College
In drug-based cancer therapy, there is presently a clinical need to identify methods that make it possible for the clinician to know in which parts of the tumour a drug has been successfully delivered. Ultrasound can enable targeting of drugs, and the mechanisms by which it so does can be monitored remotely using a conventional diagnostic ultrasound array. The aim of this project is to determine whether a new technique developed within Oxford, called Passive Acoustic Mapping, can enable the clinician to visualize the tumours or tumour regions in which a drug has been successfully delivered in real time.
Non-contact vital sign monitoring using webcams
Jon Daly, St John’s College
We have been developing novel methods for non-contact vital sign monitoring using a webcam. The webcam measures the light reflected from one or more regions of interest in the patient’s face. Previous work by others has shown that photoplethysmogram (PPG) signals can be remotely acquired from the human face with normal ambient light as the source. We have patented a novel method of identifying spectral components in the PPG image signal which correspond to the heart rate and to the respiratory rate. We are also extending the method to obtain measurements of blood oxygen saturation (SpO2) from the webcam image.
We have large amounts of data already collected from a completed study carried out in the Oxford Kidney Unit. The initial work on the project will require the student to reproduce the results which we have obtained from one or two patients, whose data we have extensively analysed in our current work.
There is a major unmet need to identify a methodology that will enable imminent hypoglycaemia to be detected in a reliable manner. Hypoglycaemia, a well-known side effect of glucose lowering treatments in patients with diabetes mellitus, is often a limiting factor to achieving good glycaemic control. We wish to determine whether hypoglycaemia detection can be enhanced by using data fusion techniques to combine glucose data with non-invasive measurements of vital signs (such as heart rate, skin temperature, respiratory rate and oxygen saturation) that may be influenced by low glucose levels.
The objective of the feasibility study underpinning this short project is to capture changes in value and variability of heart rate, respiratory rate, oxygen saturation, blood pressure, skin temperature and body movement as hypoglycaemia is induced (blood glucose will be lowered to less than 2.2 mmol/l) in 10 volunteers. A high-resolution video camera (Point Gray, Richmont, Canada) will be used to film the volunteer’s forehead throughout the Insulin Tolerance Test (ITT). The data acquired will be used to derive possible skin colour changes during hypoglycaemia, as well as changes in heart rate, respiratory rate and oxygen saturation. Each of the 10 volunteers will have their vital signs double-monitored throughout the ITT.
Numerical model of shock wave interactions with cells
Dongli Li, St Anne’s College
Shock waves are used therapeutically in medicine to fragment kidney stones. However, they also have potential to destroy cancer tumours and to open up cell membranes to allow therapeutic agents to be delivered. Here we will use state-of-the-art computational tools to understand how cells respond to shock waves and assess the process by which they can be destroyed or temporarily opened for therapy. This will let appropriate shock wave parameters be determined to result in the desired therapeutic effect.
Patient-specific prediction of abdominal aortic aneurysm rupture: Translating Models to the Clinic.
Jack Hornsby, Wolfson College
Abdominal aortic aneurysms (AAAs) are localised dilatations of the abdominal aorta. Although most AAAs are asymptomatic, they may rupture, and in that case the outcome is usually catastrophic: the overall risk of death due to rupture is 80% to 90%. Overall, this disease accounts for 1-2% of all deaths in the UK. Up to date, there is no satisfactory methodology to assess whether a detected aneurysm is at risk of rupture and simplistic size correlations are used: currently the decision on whether to intervene is based solely on the diameter of an aneurysm reaching a critical size (5.5cm). This approach fails to identify small aneurysms with high risk of rupture and those larger aneurysms which should be managed conservatively due to the risk of intervention exceeding their risk of rupture. Distinguishing those aneurysms most at risk of rupture will yield significant improvements in patient healthcare and corresponds to a pressing and unmet clinical need. In fact, more sophisticated modelling approaches to predict aneurysm rupture are currently available. There is a need to translate these models into clinical practise. This short project, which links bioengineers, clinicians with industry provides the foundations for realisation of this aim.
Predicting Physiology Patterns in Pregnancy (4P study)
Rebecca Pullon, St Hugh’s College
An urgent need to develop an evidence-based, national, Modified Obstetric Early Warning Score (MEOWS) was highlighted in the two most recent Confidential Enquiries into Maternal Deaths in the UK. An essential prerequisite to developing such an early warning system is knowledge of the normal distributions of vital signs (blood pressure, temperature, respiratory rate, heart rate and oxygen saturation) in “low-risk” pregnant women. The 4P Study aims to obtain this longitudinal data, from <14 weeks’ gestation to 2 weeks after delivery, from women participating in the Oxford arm of the INTERBIO-21st Study.
Maternal vital signs are now being recorded at each ultrasound scan visit. All routinely collected intra-partum and post-partum measurements are being added to the dataset. Each participant is also being trained in the use of home monitoring equipment to provide a daily dataset for the 2 weeks after delivery.
The student will develop pattern analysis methods to track the changes in physiology during pregnancy, and for the first two weeks after delivery, for the whole population and for individual women. This will require both univariate analysis (each variable in turn) and multivariate analysis based on visualisation techniques. In addition, the student will develop and test algorithms to extract respiratory rate from the accelerometer data recorded with a smartphone at each visit and subsequently in the home.
Quantifying collateral flow pathways in the brain
Flora Kennedy McConnell, St John’s College
Your brain needs blood all the time to continue functioning. Even a short interruption can lead to fainting and a longer term blockage can result in a stroke, which kills brain tissue and often causes serious disability. Blood flows through a number of different routes to the brain and even when one is blocked it can get round another way (what we call collateral flow). However, this is not always very robust and can sometimes go wrong. What we would like to do is to measure how well blood is travelling around the brain, which pathways the blood is travelling through and see whether this information can help to identify which stroke patients are likely to be most at risk of a secondary stroke. We have a way of doing this at the moment, but we need to make it more accurate. To do this we will construct a mathematical model of blood flow and the measurement technique that will enable doctors to get more useful and trustworthy information.
Text Prediction for Therapeutic Interfaces
Neil Dhir, Wolfson College
Human computer interaction for patients with severely degraded motor skills including but not limited to ALS, MS, Stroke, and other sufferers is impossible using traditional user interfaces. Companies, techniques, and specialized interfaces have been developed to serve these communities. At the core of all is prediction, in particular prediction of either an intended mouse-like input or an intended text continuation input (not unlike T9 or the text prediction used on a modern smart-phone). There is both formal and practical room to improve the predictive performance of such systems. This project has two key components: one a mathematical component, namely, research to improve the predictive performance of natural language text models; second, engineering research to integrate such models into clinically relevant therapeutic computer interface systems.
A review of a subset of available text-prediction language models has been performed. Research will include understanding, leveraging, and improving these models for therapeutic applications and more. In particular we propose to investigate ways to include contextual information extracted from past discourse to improve conversation-specific prediction (e.g. using past email communication to improve predictive accuracy in an on-going conversation). This is a fundamental machine learning research task formalized as discrete conditional distribution estimation with side knowledge.
Preliminary contact has also been made with both a commercial vendor of therapeutic computer interface software (Applied Human Factors http://newsite.ahf-net.com/) and a single ALS patient who uses said software. A second, applied component of this research project would be to develop the software infrastructure required to translate the research from the first component into a real-world clinical application.