The Foveal Avascular Zone (FAZ) is of clinical importance since the vascular arrangement around the fovea changes with disease and refractive state of the eye. Therefore, it is important to segment and quantify the FAZ accurately. Studies done to date have achieved reasonable segmentation but there is a need for considerable improvement. In order to test and validate newly developed automated segmentation algorithms, we have created a public dataset of these retinal fundus images. The 304 images in the dataset are classified into: diabetic (107), myopic (109) and normal (88) eyes. The images were classified by a clinical expert and include clinical grading of diabetic retinopathy and myopia. The images are of dimensions 420 x 420 pixels (6mm x 6mm of retina). Both clear and manually segmented by a clinical expert (ground truth) images are available (608 total images). In these images, the FAZ is the green region marked in manually segmented image. The images can be used to test newly developed techniques and the manual segmentation images can be used as a ground truth for making performance comparisons and validation. It should also be noted there are only a few studies using supervised learning to segment the FAZ and this dataset will potentially be useful for machine learning training and validation. The image database, The Foveal Avascular Zone Image Database (FAZID) dataset can be accessed from the ICPSR website at the University of Michigan (https://doi.org/10.3886/E117543V2).
The Foveal Avascular Zone (FAZ) is of clinical importance since the vascular arrangement around the fovea changes with disease and refractive state of the eye. Therefore, it is important to segment and quantify the FAZs accurately. Here we provide a new methodology for this measurement. Eighty normal fundus images of dimensions 420x420 pixels corresponding to 6mm x 6mm were used in this study. Each fundus image was manually segmented by a clinical expert (ground truth), the new methodology and an existing technique provided by the image acquisition device (Cirrus 5000 Carl Zeiss Meditec Inc., Dublin, CA). The images were first processed by a Difference of Gaussian (DoG) filter iteratively 25 times after being complemented. This is followed by a Prewitt edge detection and repeated image dilation at angles of 0,45 and 90 degrees. Image closure was then applied followed by noise and small object removal which resulted in the segmented boundary. For deeper insight into shape change, besides the diameter of the FAZ other parameters - eccentricity, perimeter, major axis, minor axis, incircle, circumcircle, Fmin, Fmax, tortuosity, vessel diameter index and vessel avascular density - were calculated. The mean diameter by manual segmentation was 673.04 ± 86.92 μm compared to 688.42 ± 72.18 μm by our technique. The corresponding value generated by the instrument was 623.60 ± 121.50 μm. This technique shows considerable improvement in accuracy (the mean value as well as the standard deviation) when compared to system segmentation and the ground truth. These aspects will be discussed in the paper.
To quantitatively describe and evaluate a new image processing technique for estimating the Foveal Avascular Zone (FAZ) in subjects with Diabetic Retinopathy and myopes. From a total of 328 images obtained from Diabetic Retinopathy (113), myopes (120) and normal (93), the FAZ dimensions were quantified using a new image processing algorithm. These parameters were also determined manually and by the OCT manufacturer’s inbuilt algorithm. In the new technique, the images were first pre-processed by using a DOG filter iteratively before being complemented followed by a Prewitt edge detection and repeated image dilation at angles of 00, 450 and 900. Image closure was then applied followed by noise and small object removal which resulted in the segmented boundary. For deeper insight into shape change, in addition to the diameter of the FAZ other parameters such as the area, diameter, major axis, minor axis, orientation, perimeter vessel avascular density (VAD), Vessel diameter Index (VDI), etc. were obtained. The circularity index was calculated using the FAZ area and perimeter parameters. The mean FAZ diameter (mm) by the new automated technique, manual-segmentation (ground truth), and inbuilt instrument algorithm were 0.67 ± 0.87, 0.67 ± 0.72 and 0.61 ± 0.14. The mean of FAZ area (mm2) was 0.36 ± 0.10, 0.33 ± 0.09 and 0.43 ± 0.14 in normal, myopia and diabetic subjects respectively. The new technique shows considerable improvement in accuracy (mean ± SD) when compared to the inbuilt system segmentation and the ground truth (manual marking by an expert clinician). The study results show that the FAZ area in Diabetic Retinopathy is significantly different (p=0.003) when compared to myopic eyes (p=0.016) and normals.