Fully automated detection, grading and 3D modeling of maculopathy from OCT volumes

Maculopathy is one of the prime causes of blindness and it is effectively diagnosed through optical coherence tomography (OCT) images. The two major forms of maculopathy are macular edema and central serous retinopathy and many clinicians have recommended OCT imaging over other eye testing techniques to examine maculopathy. However, to the best of our knowledge, there is a scarcity of literature related to the automated severity analysis and 3D modeling of maculopathy affected human retina. Therefore, this paper presents a support vector machines (SVM) based fully automated classification model to diagnose two prime forms of maculopathy. The proposed method not only diagnose the maculopathy but also grades it as per the clinical standards. The proposed system first computes coherent tensors from the OCT volume and then it forms a 7D feature vector in which three features are extracted from retinal thickness profile and four features are extracted from the retinal fluids. SVM was trained on thirty OCT volumes (ten ME, ten CSR and ten normal). Total ninety OCT volumes (thirty normal, thirty CSR and thirty ME) of seventy-three patients are used for validation. The proposed system achieved the accuracy, true positive rate and true negative rate of 97.78%, 96.77%, 100% respectively.