| Optical Coherence Tomography (OCT) is a noninvasive method to create a volume of data that is a measurement of the reflectance properties of tissue. The retina of the human eye has a layered structure that presents as bands in the OCT images. Evaluation of OCT images can be used to diagnose abnormal structure due to disease or damage. The principal task in evaluating the OCT images is segmentation. While the images can be evaluated by trained personnel, the quantity of OCT image data is increasing rapidly, therefore, the creation of methods for automatic segmentation is vital. Prior work in segmenting the bands in OCT images utilize a variety of means including edge and region detection, spline fitting, graph cuts and shape and texture analysis. The method presented here uses a combination of filtering and curve and parametric model based fitting to find the four posterior hyper-reflective bands in an OCT image of the retina. A strength of the parametric model based method is that additional information is added to the image that is not present in the image itself. Future work will expand to find all of the band features in the retina. Future work will also explore alternative filtering at different stages, alternate curve fitting methods, and alternate band models in 1, 2 and 3 dimensions. The goal is to create an automatic system that will not require human intervention to find all of the band structures in the retina and be tolerant of noise and variations in features.;Keywords: shape, image processing, segmentation, optical coherence tomography, retina. |
