Background: Images are an essential aspect of monitoring and prediction in clinical management in a wide range of diseases in particular in retinal diseases, such as diabetic retinopathy (DR) and retinal vascular occlusive (RVO) disease. Evaluation of retinal disease typically focuses on assessing the number or size of lesions, but this ignores their spatial context and the patient’s clinical information, which makes existing models less accurate for clinical use.
This project will develop new spatio-temporal statistical monitoring and predictive models for eye diseases by identifying, synthesizing and testing suitable models from a family of novel hierarchical models that we have developed in Liverpool. One of the most promising new avenues of monitoring retinal disease is to have a personalised computerised monitoring tool, but we have yet to identify a suitable statistical spatio-temporal and discriminant approaches to combine imaging data with clinical data.
This studentship will focus on development of statistical longitudinal and spatial models for pathological lesions found in wide field imaging (WFI) data. The student will work under supervision of a statistician (Dr Gabriela Czanner, primary supervisor), closely with the imaging group of (Dr Yalin Zheng, co-supervisor) and retinal consultants (Prof Simon Harding, co-supervisor; Amira Stylianides, Savita Madhusudhan and senior grader David Parry from the Liverpool Ophthalmic Reading Centre). It will develop statistical monitoring models to relate the lesions to diseases such as capillary nonperfusion lesions and will develop statistical methods to find and validate biomarkers for management using existing large datasets.
1) Statistical analytical validation: To develop spatio-temporal hierarchical statistical models for pathological lesions imaging biomarkers measured via automated image grading for DR and RVO diseases. Hierarchical models will be extended into discriminatory tools to give probability of disease progression. They will be implemented in a statistical software (in R).
2) Qualification: To demonstrate how the imaging biomarkers and clinical measurements (e.g. blood glucose) are associated with clinical endpoints and to derive a personalised monitoring tool. The statistical methods for validation (sensitivity, specificity and time dependent ROC analyses) of the monitoring tools will be proposed and developed.
3) Utilisation: To explore a statistical strategy for best implementation of the personalised tool for patient monitoring visits. The confidence intervals of disease progression will be derived for each patient, which is needed for understanding of the risks and uncertainties.
Training and Support: This PhD project is suitable for an ambitious individual wishing to develop a portfolio in statistical research with strong elements of understanding the biomedical imaging and clinical research. The candidate will obtain skills in statistical modelling, statistical software development, statistical models for image analysis, data collection, understanding the cutting edge imaging techniques including WFI, optical coherence tomography, angiography and writing clinical papers. The successful candidate will need to have suitable experience in statistical modelling and computing. The student will some necessary experience in image grading, and the management of DR and RVO. The candidate will have access to an industrial collaboration and academic collaborators in the NetwORC UK (Moorfields/UCL, Queens Belfast and Wisconsin).
The Institute of Ageing and Chronic Disease is fully committed to promoting gender equality in all activities. We offer a supportive working environment with flexible family support for all our staff and students and applications for part-time study are encouraged. The Institute holds a silver Athena SWAN award in recognition of on-going commitment to ensuring that the Athena SWAN principles are embedded in its activities and strategic initiatives.
Enquiries to: Gabriela Czanner [email protected]
To apply: please send your CV, a covering letter, names of two referees, to [email protected]
with a copy to [email protected]
The successful candidate should have, or expect to have an Honours Degree at 2.1 or above (or equivalent) in Statistics or Biostatistics. It is essential to have good background knowledge in mathematics, statistics, computer programming (e.g., Matlab, STATA or R) plus a proactive approach to their work. Candidates whose first language is not English should have an IELTS score of 6.5 or equivalent.
The successful candidate will be provided with state-of-the-art resources for computing, and support for research, training courses and conferences, as well as tuition fees (at Home/EU rate) and a monthly stipend for 3 years.
(1) Bowman FD and Waller LA, Modelling of Cardiac imaging data with spatial correlation,” Stat Med 2004;23(6): 965–985.
(2) Fleming AD, Philip S, Goatman KA, Prescott GJ, Sharp PF, Olson JA. The evidence for automated grading in diabetic retinopathy screening. Curr Diabetes Rev 2011;7(4):246-52.
(3) ) Hughes D, Komarek A, Czanner G, Garcia-Finana M. Dynamic longitudinal discriminant analysis using multiple longitudinal markers. Stat Methods Med Res. 2016