Precision Medicine Doctoral Training Programme

Use of artificial intelligence for community-based detection of glaucoma

Precision Medicine Project - Use of artificial intelligence for community-based detection of glaucoma

Supervisor(s): Prof Baljean Dhillon,  Dr Emma Pead,  Dr Andrew Tatham & Prof Niall Strong (Glasgow Caledonian University)
Centre/Institute: Centre for Clinical Brain Sciences

Background

Worldwide, glaucoma is the leading cause of irreversible blindness. As sight lost from glaucoma cannot be restored, early diagnosis and treatment is key but this is challenging as most people with glaucoma are not aware they are affected until substantial vision has been lost.  The diagnosis of glaucoma relies on the detection of changes to the optic nerve, visible on retinal photographs; however, changes can be subtle and there is large variation in normal appearance. This leads to a large number of false positive referrals and patients with glaucoma being missed.  A recent review by the UK National Screening Committee advised against population screening for glaucoma due to the inadequacy of current screening tests.1  Artificial intelligence (AI) has the potential to improve accuracy of glaucoma detection, which may make screening feasible.2

Aims

The project will utilise the Scottish Clinical Optometry and Ophthalmology Network e-research (SCONe) collaboration, a research repository developed by academic partners at Glasgow Caledonian University and University of Edinburgh.  In Scotland, over 1 million retinal images are captured by optometrists each year, providing a rich population-based resource for research. SCONe aims to utilise images from optometry to build a curated dataset, incorporating retinal photographs and optical coherence tomography (OCT) images linked to clinical information. This is a growing, longitudinal resource, ideal for developing AI tools to improve clinical decision making.

The main aim of this project is to use SCONe to develop AI algorithms that can be used in optometric practices to aid identification of glaucoma.  The longitudinal nature of the SCONe dataset, combined with hospital and blind registration data, will also be used to address the question of whether features on retinal photographs can be used to better predict risk in glaucoma.

Glaucoma-related AI studies have already examined the ability of convolutional neural networks (CNNs) to detect glaucoma from optic disc photos.2  A major challenge has been establishing the ground truth, as the diagnosis of glaucoma is challenging.  Google Health used consensus grading of images by a panel of experts to establish the diagnosis, and found an AI algorithm was better at detecting glaucoma than single glaucoma specialists.2  However, it is not clear how well this algorithm would perform in the general population, and in the real world, optic disc images are interpreted in the light of other clinical information.  A better gold standard for determining glaucoma is change in optic disc appearance over time3; however, datasets with sufficient longitudinal data are rare. The longitudinal design of SCONe provides an opportunity to evaluate AI against this ground truth on a large scale in the general population, also allowing incorporation of other clinical information.

Training Outcomes

The student will receive state-of-the-art training in the core disciplines of image analysis, computational modelling, statistical methods, and data science while gaining expert knowledge in the context of ophthalmic disease. This highly interdisciplinary approach is well aligned with the “T-shaped researcher” training requirements identified as key in the DTP in Precision Medicine. The student will develop the essential soft and domain-specific skills necessary to design and implement novel quantitative and computational methods that could solve challenging problems across the entire spectrum of ophthalmology/optometry both in academic and industrial settings.

References

  1. Public Health England. The UK NSC recommendations on glaucoma screening in adults, December 2019. Accessed online at: https://legacyscreening.phe.org.uk/glaucoma
  2. Phene S, et al. Deep Learning and Glaucoma Specialists: The Relative Importance of Optic Disc Features to Predict Glaucoma Referral in Fundus Photographs. Ophthalmology 2019;126(12):1627-1639.
  3. Medeiros FA, Zangwill LM, Bowd C, Sample PA, Weinreb RN. Use of progressive glaucomatous optic disk change as the reference standard for evaluation of diagnostic tests in glaucoma. Am J Ophthalmol. 2005 Jun;139(6):1010-8.

Apply Now

Click here to Apply Now

  • The deadline for 21/22 applications is Thursday 7th January 2021.
  • Applicants must apply to a specific project, ensure you include details of the project you are applying to in Section 4 of your application. You should contact the primary supervisor prior to making your application.  
  • As you are applying to a specific project, you are not required to submit a Research Proposal as part of your application. 
  • Please ensure you upload as many of the requested documents as possible, including a CV, at the time of submitting your application.