Reconstruction of 3D Surface Maps from Anterior Segment Optical Coherence Tomography Images Using Graph Theory and Genetic Algorithms

Automatic segmentation of anterior segment optical coherence tomography images provides an important tool to aid management of ocular diseases. Previous studies have mainly focused on 2D segmentation of these images. A novel technique capable of producing 3D maps of the anterior segment is presented here. This method uses graph theory and dynamic programming with shape constraint to segment the anterior and posterior surfaces in individual 2D images. Genetic algorithms are then used to align 2D images to produce a full 3D representation of the anterior segment. In order to validate the results of the 2D segmentation comparison is made to manual segmentation over a set of 39 images. For the 3D reconstruction a data set of 17 eyes is used. These have each been imaged twice so a repeatability measurement can be made. Good agreement was found with manual segmentation for the 2D segmentation method achieving a Dice similarity coefficient of 0.96, which is comparable to the inter-observer agreement. Good repeatability of results was demonstrated with the 3D registration method. A mean difference of 1.77 pixels was found between the anterior surfaces found from repeated scans of the same eye