[Lecture Notes in Computer Science] Pattern Recognition and Computer Vision Volume 11256 (First Chinese Conference, PRCV 2018, Guangzhou, China, November 23-26, 2018, Proceedings, Part I) || Automated and Robust Geographic Atrophy Segmentation for Time Series SD-OCT Images

摘要： Geographic atrophy (GA), mainly characterized by atrophy of the retinal pigment epithelium (RPE), is an advanced form of age-related macular degeneration (AMD) which will lead to vision loss. Automated and robust GA segmentation in three-dimensional (3D) spectral-domain optical coherence tomography (SD-OCT) images is still an enormous challenge. This paper presents an automated and robust GA segmentation method based on object tracking strategy for time series SD-OCT volumetric images. Considering the sheer volume of data, it is unrealistic for experts to segment GA lesion region manually. However, in our proposed scenario, experts only need to manually calibrate GA lesion area for the ?rst moment of each patient, and then the GA of the following moments will be automatically detected. In order to fully embody the outstanding features of GA, a new sample construction method is proposed for more e?ectively extracting histogram of oriented gradient (HOG) features to generate random forest models. The experiments on SD-OCT cubes from 10 eyes in 7 patients with GA demonstrate that our results have a high correlation with the manual segmentations. The average of correlation coe?cients and overlap ratio for GA projection area are 0.9881 and 82.62%, respectively.