Aim of the PhD Project:
Ophthalmology has reached a point where gene vectors, small drug molecules, and stem cells are considered as hopeful treatments for degenerative diseases that cause blindness. For example, gene therapies for Choroideremia appear effective, while implanted stem cells show potential for retina regeneration in Age-Related Macular Degeneration (AMD). These therapeutics need to be delivered to specific retinal layers for their promise to be fulfilled.
Precise delivery of novel therapies will be guided by Intraoperative Optical Coherence Tomography (iOCT) microscopes, a modality microscopically imaging not only the retinal fundus, but also the “hidden” subretinal layers where therapeutics should be delivered. This PhD project concerns computationally augmenting the capabilities of iOCT imaging through incorporation of real-time AI in the acquisition and processing pathways towards the creation of a robust navigation system that guides therapy implantation in vitreoretinal surgery.
Project Description / Background:
Due to their requirement for real-time imaging, iOCT systems provide axial/slice images that sacrifice resolution for increased frame rates. Therefore, they are not yet utilised for interventions that require subretinal visualisation. Further, the images’ 2D/slice nature makes orientation/localisation of delivery sites within the 3D subretinal volume challenging for the surgeon. Within the context of the proposed PhD Project, the student, of a computer science or computer engineering background, will develop algorithms that computationally enhance the image reconstruction process therefore improving the capabilities of iOCT devices without risking CE-marking violations of the hardware.
This project builds on a many year-long collaboration with Moorfields Eye Hospital, the existence of a clinical protocol for data acquisition and annotation, and access to a clinical-grade iOCT microscope (Zeiss Rescan/Lumera). The PhD project entails addressing the following challenges:
– Improve iOCT image resolution: Use high resolution pre-operative 3D OCT to improve the low axial and lateral resolution of 2D iOCT images and clearly delineate retinal layers. This improvement will increase the usefulness of iOCT for vitreoretinal interventions. Preliminary results using deep learning based super-resolution algorithms demonstrate the potential of this approach and seed this research component.
– Assist in target localisation: Augment the iOCT device to direct its scanning to regions of the retina that have been pre-operatively selected as therapeutics-delivery sites. This will reduce the time of surgery by removing the need for the clinician to perform a laborious slice-by-slice iOCT scanning and site re-annotation during surgery. The team has already conducted preliminary research that uses retinal vessels as landmarks that enable the registration of iOCT slices and pre-operatively acquired OCT volumes, and this objective represents a natural progression.
– Enable guided therapy delivery: Develop an iOCT-based navigation framework that detect and tracks therapy-delivery targets, overall guiding the clinician to the retinal sites identified for intervention.
Such innovations are necessary to increase the value of iOCT in vitreoretinal surgery, and are stepping stones towards the goal of microprecise delivery of sight-restoring therapies.