Glaucoma Detection in Mobile Phone Retinal Images Based on ADI-GVF Segmentation with EM initialization

Abstract

The advanced development of mobile phone and lens technology has made retinal imaging more convenient than ever before. In the digital health era, mobile phone fundus photography has evolved into a low-cost alternative to the standard ophthalmoscope. Existing image processing algorithms have a problem with handling the narrow field of view and poor quality of retinal images from a mobile phone. This paper enhances the accuracy of our previously proposed scheme, ADI-GVF snakes, to improve the segmentation of the optic disk (OD) and the optic cup (OC) for glaucoma pre-screening [1] from retinal images obtained from a mobile phone. This work integrated a better OD localization method, namely, the exclusion method (EM) with ADI-GVF segmentation for the OD and the OC. The improved algorithm can segment the regions of the OD and OC more accurately, resulting in a more precise value of the cup-to-disk area ratio (CDAR). The proposed method yields as high as 93.33% for true positive rate (TPR) and 93.87% for true negative rate (TNR) and as low as 6.12% and 6.66% for false omission rate (FOR), and false discovery rate (FDR). It also improves TPR, TNR, FOR, and FDR of the previous scheme [1] by 4.45%, 4.08%, 4.08%, and 4.44% respectively.

Automated optic disk segmentation for optic disk edema classification using factorized gradient vector flow

“…We compared the factorized gradient vector flow (FGVF) 8 , 9 used in our work against four other comparative methods: alternated deflation-inflation gradient vector flow (ADI-GVF) 37 , traditional gradient vector flow (GVF) 36 , region growing (RG) 29 , and super-pixel clustering (SPC) 30 . All methods except super-pixel clustering required initial points.…”

“…The oval fitting model is used to segment higher accurate boundary 229 images from DIARETDB0, DRSHTI-GS Overlap 66.59, Accuracy 96.30, F1-score 95.1 Abdullah et al 35 Using the fuzzy clustering mean method to localize the location. The active contour model is applied for OD segmentation 320 images from DRIVE, STARE, DIARETDB1, DRIONS-DB Sensitivity 87.26, Overlap 84.56, DICE 88.40, Accuracy 99.46 Kusumandari et al 36 Comparing Gradient Vector Flow (GVF) snake active contour model and ellipse fitting method in OD detection 64 images from the local database C/D ratio of area: 84.38 (GVF), 81.25 (Ellipse Fit) Khaing et al 37 Using an alternated deflation-inflation gradient vector flow (ADI-GVF) model for OD and optic cup segmentation in Glaucoma prescreening application. The ADI-GVF represents a balloon model that repeatedly deflates and inflates alternately until it converges at the edge of the targeted boundary 225 images from mobile phone database, Drishti-GS, HFS Recall 88.50, Precision 84.35, F-Measure 84.06 Gagan et al 38 Using basis splines-based active contour.…”

Calculation and Function of Composition Proportion in Photography Based on Digital Information Technology

“…In order to evaluate the effectiveness of our algorithm, we use the short name of mean reciprocal rank (MRR) to evaluate the algorithm of Gaussian kernel function; as a mechanism to evaluate the advantages and disadvantages of Mobile Information Systems ambiguous algorithms, MRR has certain universality in the world. It mainly measures whether the original image corresponding to the image to be matched in the database is blurred and arranged in front [18]. For a group of Q to be queried, the reciprocal formula of average sorting is as follows:…”

“…Let the invariant coefficient of the image after photographing be σ ′ , and formula (18) proves that the invariant coefficient is fixed before and after photographing.…”