Blood vessel segmentation in en-face OCTA images: a frequency based method

TL;DR

This paper presents a frequency-based method for segmenting retinal vessels in en-face OCTA images using a Gabor filter bank. It constructs three frequency-enhanced representations $I_1,I_2,I_3$ from 18 filters ($3$ frequency scalings and $6$ orientations) and performs vessel segmentation via thresholding and a Potts-model FAZ identification, yielding a piecewise-constant map for the FAZ. Quantitative evaluation on 10 eyes shows VD differences of $1.58\pm 1.08\%$ from the device and robust FAZ descriptor agreement with expert GT (Dice ≈ 0.89), while qualitative results demonstrate accurate small-vessel detection and boundary fidelity. The authors also introduce adaptive local vessel density maps to enable region-specific analysis of retinal blood flow and outline plans for broader validation across devices and datasets.

Abstract

Optical coherence tomography angiography (OCTA) is a novel noninvasive imaging modality for visualization of retinal blood flow in the human retina. Using specific OCTA imaging biomarkers for the identification of pathologies, automated image segmentations of the blood vessels can improve subsequent analysis and diagnosis. We present a novel segmentation method for vessel density identification based on frequency representations of the image, in particular, using so-called Gabor filter banks. The algorithm is evaluated qualitatively and quantitatively on an OCTA image in-house data set from $10$ eyes acquired by a Cirrus HD-OCT device. Qualitatively, the segmentation outcomes received very good visual evaluation feedback by experts. Quantitatively, we compared the resulting vessel density values with automated in-built values provided by the device. The results underline the visual evaluation. For the evaluation of the FAZ identification substep, manual annotations of $2$ expert graders were used, showing that our results coincide well in visual and quantitative manners. Lastly, we suggest the computation of adaptive local vessel density maps that allow straightforward analysis of retinal blood flow in a local manner.

Figures (9)

• Figure 1: New image representations are obtained by the convolution of the original image $I$ with Gabor filters. The images $I_1, I_2, I_3$ correspond to the $3$ chosen frequency ranges defined by $\{\omega_i \}_{i=1}^3$.
• Figure 2: The combined substeps described in Sections \ref{['vessels']} and \ref{['fazid']} yield the final image.
• Figure 3: Image representations for the FAZ identification
• Figure 4: Originals (top) and segmentations (bottom) from the proposed method
• Figure 5: FAZ identified by an expert grader(a) from the device(b) and our algorithm(c).