Deep learning for semantic segmentation of 3D vasculature in diabetic retinopathy and healthy control participants enabling evaluation of vessel metrics

Poster presented as part of the Crick BioImage Analysis Symposium.

Optical coherence tomography angiography (OCT-A) is an imaging modality enabling 3D imaging of the human eye retinal vasculature in vivo. Classification of arteries and veins in retinal images is of high medical interest as these systems are divergently affected in many retinal vascular diseases that affect vision [1]. Arteries and veins are more challenging to distinguish in 3D OCT-A images than via 2D retinal photography [2]. Here, a 2-stage deep learning approach is presented for classification of arteries and veins in 3D. This enables the evaluation of vessel metrics in diabetic retinopathy and healthy control participant datasets.

References

1. Kram, M.K., et al., Retinal vessel diameters and cerebral small vessel disease: the Rotterdam Scan Study. Brain, 2006. 129(Pt 1): p. 182-8.
2. Xu, X., et al., Differentiating Veins From Arteries on Optical Coherence Tomography Angiography by Identifying Deep Capillary Plexus Vortices. Am J Ophthalmol, 2019. 207: p. 363-372.

3. Taha, A.A. and A. Hanbury, Metrics for evaluating 3D medical image segmentation: analysis, selection, and tool. BMC Med Imaging, 2015. 15: p. 29.
4. Lotmar, W., A. Freiburghaus, and D. Bracher, Measurement of vessel tortuosity on fundus photographs. Albrecht Von Graefes Arch Klin Exp Ophthalmol, 1979. 211(1): p. 49-57.

Permission has been given by authors to upload to Crick Figshare. Copyright remains with the original holders.