PhD studentships available

Combining deep learning to mechanistic modelling to automate the interpretation of clinical retinal imaging

Application deadline: Tuesday 30 July 2019

Project description

Increasing diabetes incidence and an aging population have placed an unprecedented burden on clinical ophthalmology. With an anticipated 60% increase in demand for services over the next twenty years, this will necessitate new ways of working. Equally, a range of new and improved retinal imaging technologies have been developed that have the potential to improve both patient monitoring and diagnosis, but which require in-depth expert interpretation. To address these challenges, this project will combine mechanistic, biophysical modelling of the retina with machine learning tools to develop a software platform to assist ophthalmologists with the interpretation of image data from wide-field, colour retinal photography and optical coherence tomography angiography (OCT-A). The platform will also be used for treatment planning and optimisation. 

This PhD project draws together mechanistic biophysical modelling, machine learning and clinical ophthalmology. It integrates with a larger project that aims to meet a range of clinical challenges in ophthalmology, including 1) assisting ophthalmologists in the diagnosis and monitoring of retinal diseases, in an era of increasing disease incidence and limited healthcare funding; 2) developing machine learning image analysis approaches that do not require extensive, time-intensive labelling input by experts (and which can be adapted to a wide range of retinal imaging modalities); 3) providing a platform for personalised treatment planning, optimisation and monitoring of response to treatment. 

We have developed a mathematical model of human retinal vascular structure and fluid dynamics (blood flow, oxygenation), based on our REANIMATE framework (d’Esposito et al, Nature Biomed Eng, 2018). Blood vessel networks are generated according to a physiologically-realistic growth process, parameterised by pro-angiogenic signalling (VEGF) and arterial oxygen delivery (Figure 1). This model will be extended to model pathology such as diabetic retinopathy (loss of capillary bed due to vascular occlusion, leading to associated revascularisation and leakage), haemorrhage and age-related macular degeneration. 

These synthetic vascular networks form the basis of our machine learning method for identifying features in real-world retinal images. Unpaired cycle-consistent adversarial networks such as the cycleGAN algorithm (Zhu et al, arXiv, 2018) enable mappings between domains to be performed without the need for ground- truth labels. We will use this novel approach to identify features in retinal photographs and OCT-A images, by creating a library of synthetic retinal structures. Subsequently, the algorithm will be trained to monitor the progression of disease in longitudinal measurements and to detect the earliest signs of disease. 

Eligibility

Normally, to be eligible for a full award a student must have no restrictions on how long they can stay in the UK and have been ordinarily resident in the UK for at least 3 years prior to the start of the studentship. Please check EPSRC website for full eligibility criteria: https://epsrc.ukri.org/skills/students/help/eligibility/ 

Application process

Eligible applicants should first contact Prof Simon Walker-Samuel (simon.walkersamuel@ucl.ac.uk). Please enclose a cover letter (including the names and contact details of two referees), one-page research statement and two pages CV. The supervisory team will arrange interviews for short-listed candidates. After interview, the successful candidate will be required to formally apply online via the UCL website. Regrettably, we are only able to contact candidates who are successful at the shortlisting stage. Thank you for your interest in this position.

Funding notes

The studentship will be 4 years in duration (or 6 years for part-time candidates), starting Autumn 2019. It will cover UK/ EU fees, a yearly stipend of £17,009 and a small allowance for consumables.

Discovering the biomechanical signatures of cancer cells in 3D tumouroid models

Application deadline: Sunday 18 August

Project description

This PhD will entail development of 3D models of solid tumours, which we have termed tumouroids. These tumouroid models are made of up native extra-cellular proteins, and are biomimetic in terms of matrix density, matrix composition, cell spatial positioning and formation of hypoxia gradients. Tumouroids comprise of a dense cancer mass embedded within ‘normal’ tissue stroma. This studentship aims to measure the biomechanical signatures across the tumour-stroma boundary to predict cancer invasiveness. We will use different cell line of cancer with low and high invasive capability and measure stiffness signatures as they develop across the tumour stroma boundary. In addition, we will study the formation of tumour associated collagen signatures and correlate these to mechanical and biological characteristics.

Eligibility

This funding is available for UK and EU passport holders. There is no minimum qualifying residence requirement for applicants from the EU. Please DO NOT enquire about this studentship if you are ineligible. Please refer to the following website for eligibility criteria: https://www.ucl.ac.uk/prospective-students/graduate/research-degrees/mechanical-engineering-mphil-phd

Application process

Eligible applicants should first contact Dr Emad Moeendarbary ( e.moeendarbary@ucl.ac.uk ). Please enclose a cover letter (including the names and contact details of two referees), one-page research statement and two pages CV. The supervisory team will arrange interviews for short-listed candidates. After interview, the successful candidate will be required to formally apply online via the UCL website. Regrettably, we are only able to contact candidates who are successful at the shortlisting stage. Thank you for your interest in this position.

Funding notes

Full tuition fees and stipend of up to £17,009 per annum for 4 years

More details and person specification

Development of machine learning mechanisms for better clinical trial design for rare disease

Application deadline: Wednesday 21 August

Project description

We have entered into a new era of genetic therapies, with the first approved gene therapy, Luxturna, for biallelic RPE65 retinal dystrophy and a number of active clinical trials for both gene-replacement and small molecule drugs targeting particular genetic mutations. One of the challenges to bringing these new therapies to trial is the high costs associated with long trials due to challenges at present is predicting degeneration rates based on previous history. The second challenge is needing large numbers of patients in rare disease case, whether large number of patients may not exist. 

This project will develop and apply time series machine learning (ML) tools to the existing cohorts of data labels. These models will look to combine the multiple factors into a single model to predict progression of each eye. Once this model is developed it can be applied other disease. The ideal result would develop a high specificity profile which can be used in designing cohorts for clinical trials. By better predicting natural disease course, trial cohorts can more accurately be developed and trial efficacy is determined. Through the combination of machine learning, computer vision and medicine, this project hopes to investigate the following questions. 

Ideal person specification 
• A good degree (2.1 or above; or equivalent EU/overseas degree) in computer science or have a strong computational background 
• Have previous experience in image processing, especially in context of medical images 
• Have significant understanding of Machine Learning techniques, specifically in relation to medical image classification 
• Good analytical/mathematical skills, preferably with some knowledge of statistical approaches 
• Good communication skills - especially in written English 
• Very strong work ethic, with the ability to think creatively and work within a team 

Duties and Responsibilities 
• Development of some image analysis routines to facilitate data labelling 
• Development of Machine Learning models, specifically General Adversarial Networks, Spare Data reconstruction 
• Work in collaboration with other researchers especially medical professionals 
• Prepare presentations, including text and images, for delivery by self and others. 
• Travel for collaboration and other meetings or conferences. 
• Prepare manuscripts for submission to peer-reviewed journals. 
• Contribute to the overall activities of the research team, department and be aware of UCL policies. 

Application process

Please contact Dr Adam Dubis (a.dubis@ucl.ac.uk) or Dr Waty Lilaonitkul (watjana.lilaonitkul.16@ucl.ac.uk) for more information 
Application is by CV and covering letter emailed to Hattie de Laine (h.delaine@ucl.ac.uk)