Automatic Retinal Minimum Distance Band (MDB)Computation from SD-OCT Images

In this paper, we have proposed a novel automatic method to determine the Minimum Distance Band (MDB) from the retinal SD-OCT image. MDB is the minimum distance between the retinal pigment epithelium (RPE) layer and the Optic Nerve Head (ONH) surface. It is an effective biomarker for early detection of glaucoma. Our proposed method is the first automatic method for computing the Minimum Distance Band (MDB). This method uses the approximate location of three benchmark reference (TBMR) layers of the retina that help to reduce the search space. This approach is highly accurate in detecting the boundary layers even in the presence of pathology. The terminal points of the RPE serve as the ONH region. The accuracies of ONH and MDB are tested against 13 manually graded optic disc centred SD-OCT volumes (11 B-scan per volume) of Glaucoma patients as a gold standard. Our method is very effective with mean and standard deviation of the error of the ONH width and MDB of 5.36±5.55 and 14.96±17.75 respectively. The precision, recall and F1 score for the ONH region are 94.90, 98.14 and 96.49 respectively.