Translator Disclaimer
4 April 2019 Multiscale Hessian filtering for enhancement of OCT angiography images
Author Affiliations +
Proceedings Volume 10858, Ophthalmic Technologies XXIX; 108581K (2019)
Event: SPIE BiOS, 2019, San Francisco, California, United States
The comprehensive architectural analysis of the retinal vasculature would greatly aid with the diagnosis and management of many ocular diseases. Optical coherence tomography angiography (OCTA) is a powerful micrometerlevel resolution, high sensitivity, and potentially large field of view retinal imaging modality that allows assessment of 2D and 3D microvascular networks. However, the quality of retinal OCTA images are often degraded by the noise and poor vascular contrast. Digital image filters are widely used in medical diagnostic to selectively enhance specific local intensity profiles or structures such as vasculatures. Most successful feature enhancement filters employ Hessian matrix and eigen values-based approach. In this paper, we demonstrate the feasibility of multiscale Hessian filters for the enhancement of the OCTA images of a mouse retina. We show that the enhancement filter based on the ratio of Hessian eigenvalues proposed by Jerman et al. (Jerman filter) performs better than the most commonly used Frangi’s method. This improved performance included close-to-uniform response in all vascular structures and enhancement of visibility of vascular structures with non-circular cross-sections. To evaluate and compare performance, different multi-scale Hessian filtering was performed on OCTA images of different inner retina vascular beds of a mouse eye.
© (2019) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Ratheesh K. Meleppat, Eric B. Miller, Suman K. Manna, Pengfei Zhang, Edward N. Pugh Jr., and Robert J. Zawadzki "Multiscale Hessian filtering for enhancement of OCT angiography images", Proc. SPIE 10858, Ophthalmic Technologies XXIX, 108581K (4 April 2019);

Back to Top