Volumetric 3D reconstruction of the retinal vascular system in OCT angiography
3D automated reconstruction and analysis of retinal vessels of healthy and diseased eyes using machine learning and artificial intelligence in optical coherence tomography angiography.
Factsheet
- Schools involved School of Engineering and Computer Science
- Institute(s) Institut für Optimierung und Datenanalyse IODA
- Funding organisation Others
- Duration (planned) 01.11.2022 - 31.07.2025
- Head of project Prof. Dr. Tiziano Ronchetti
- Project staff Sascha Ledermann
-
Partner
Prof. Dr. Justus G. Garweg, Swiss Eye Institute AG
Dr. med. Christof Hänsli, Berner Augenklinik
SWISS EYE INSTITUTE AG
Berner Augenklinik AG
INVENTUS BERN - Stiftung
Campus Stiftung Lindenhof Bern - Keywords Diseases of the macula, ophthalmologic medical image analysis, optical coherence tomography angiography, machine learning, artificial intelligence, 3D reconstruction of retinal vessels
Situation
A well-functioning structure of the vascular network is crucial for optimal blood flow to the retina of the human eye. Regular monitoring is important, both for the timely detection of pathological changes and for monitoring the effect of therapies. Although the "state-of-the-art" imaging technology of optical coherence tomography angiography (OCT-A) provides high-resolution 2D projections of the various vessel levels, projection artefacts occur and three-dimensionality is completely lacking, which makes diagnoses prone to error. At this point, volumetric reconstruction of the retinal vasculature may provide a practical diagnostic tool for clinical experts and the opportunity to develop new measures to quantify disease progression.
Course of action
The method is being developed on data sets from subjects with a healthy retina and its robustness will initially be tested on cases with minor vascular changes and relatively normal retinal anatomy, such as macular telangiectasia (MacTel2). In follow-up projects, the method will be further developed and improved on data sets from test subjects with slight anatomical changes without significant impairment of the retinal structure (diabetic retinopathy and retinal vascular occlusion) and finally on patients with severe vascular changes in the retinal architecture (age-related macular degeneration). By combining various techniques from machine learning (ML) and artificial intelligence (AI), a hybrid framework is being developed that enables both the extraction of detailed, high-resolution spatial information and the extraction of global correlations from the data.
Looking ahead
The aim of this project is to improve the visualization of retinal vessels in order to detect microvascular pathologies of the macula. This holds great potential for understanding the functioning of retinal vascular networks in the healthy eye and especially in pathological changes. Various diseases of the macula, such as macular telangiectasia (MacTel2), diabetic maculopathy and age-related macular degeneration (AMD), have specific vascular changes, the understanding of which is an active field of research. An improved non-invasive imaging of the retinal vessels allows early detection of disease progression or reactivation. This could then be used to guide therapy and thus improve treatment safety and the long-term prognosis for affected patients.
