Extraction of Disease Area from Retinal Optical Coherence Tomography Images Using Three Dimensional Regional Statistics

We propose a new extraction method of the macular disease area in the human retinal layer from OCT images using three dimensional regional statistics. In previous researches, we extracted disease area by using the mean and standard deviation of the two dimensional disease part pointed out by a clinical doctor. However, the previous method cannot extract disease area for some disease OCT images precisely. In this paper, we propose a new extraction method of the disease area using three dimensional regional statistics. We use a set of 128 images (3D-OCT image) consisted of 2 dimensional OCT retinal image about one retina of a patient. The regional mean and regional standard deviation of gray level are calculated in the three dimensional region of interest (ROI, 125 (=5 × 5 × 5) pixels) in the abnormal area pointed by a clinical doctor. These values are compared with every ROI in the abnormal area to extract the disease area, and the proposal system measures the volume of the disease area. We apply the proposed method to OCT images of 5 patients with retinal diseases. As a result, we can measure the volume of the abnormal area with 80.7% average accuracy.