Automated Drusen Detection Technique Based On Multiple Optical Coherence Tomography Images

In recent years,the disease detection technique based on optical coherence tomography(OCT)has been a hot spot in medical image processing field.It uses computer technique to recognize disease region in OCT image.OCT image detection technique is applied to many kinds of eye disease detection widely,such as diabetic retinopathy,choroidal neovascularization,ellipsoid zone lost and so on.This paper investigates OCT image detection technique for drusen.This technique is the premise of early age-related macular degeneration detection,and results of detection contribute to comparison,quantification and monitoring of drusen in processing of its development and treatment.This paper mainly investigates automated detection technique based on multiple OCT images.The main work and contribution is following:Firstly,OCT volumetric data was preprocessed,including segmenting Bruch's membrane,flattening retina and moving up Bruch's membrane to obtain multiple C-scan.The hybrid contrast ratio was calculated to identify normal eye and drusen eye.Secondly,multiple OCT images extracted from OCT volumetric data were 3D mean filtered and normalized to improve contrast of every C-scan.Then,the processed multiple C-scans were segmented by top-hat transform and Otsu to obtain multiple segmentation results.Finally,multiple segmentation results were integrated by continuity of drusen.The vessel shadow in outer C-scan was removed by opening operation to obtain final result.This paper introduces a new automated drusen detection technique with minimum need for layer segmentation.The method is based on detection of drusen areas in C-scans at certain distances above the Bruch's membrane,circumventing the difficult task of pathologic retinal pigment epithelium segmentation.The method was tested by thirteen drusen patients' OCT images.Compared with results produced by two ophthalmologists,the means of accuracy rates are 90.54% and 85.22%,respectively.Based on the drusen detection method of OCT image proposed by this paper,ophthalmologist can observe position of disease region conveniently.The method provides technical support for investigation of relationship between drusen area and capillary perfusion and shows its potential value for research of drusen.