Glaucoma detection in mobile phone retinal images based on ADI-GVF segmentation with EM initialization

Abstract

The advanced development of a mobile phone and retinal lens technology has made fundus 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 slit-lamp machine. Existing image processing algorithms have a problem with handling the retinal images with narrow field of view and poor-quality taken by a mobile phone. This paper enhances the accuracy of our recently proposed scheme, Alternated Deflation-Inflation Gradient Vector Flow model (ADI-GVF), to improve the segmentation of the optic disk (OD) and the optic cup (OC) for glaucoma detection [1]. We integrated the exclusion method (EM), a precise algorithm for localizing the OD, with the ADI-GVF algorithm. This work has been experimentally proved that it can detect the boundaries of the OD and OC very precisely, resulting in a very accurate value of the cup-to-disk area ratio (CDAR) for precise glaucoma prescreening. The proposed method has been tested using a mobile phone dataset and two standard datasets (Drishti-GS and HFS). In mobile phone dataset, it obtains TPR up to 93.33%, and FOR as low as 6.66%. Satisfactory rates of TPR and FOR are also reported for those two standard datasets. In addition, the comparison on three datasets manifests that the proposed algorithm outperforms other state-of-the-art methods.