Abstract
Abstract: :
Purpose:To develop a new algorithm for the determination of retinal thickness in optical coherence tomograms by detecting the interfaces of the internal limiting membrane and the retinal pigment epithelium. While the OCT records high-resolution A-scans of the retina, the software used to measure retinal thickness is prone to artifacts and errors. Methods:We propose a new method to measure the retinal thickness in OCT scans. The method was inspired by the Markov method proposed recently [Koozekanani D, et al. IEEE Trans Med Imaging 2001;20:900-16]. After weighting each A-scan with some a priori, we can choose the relevant edges according to following rules. First, because of the extremely low reflectivity of vitreous, the first non-noise edge must be ILM interface. Second, the edge with the biggest weight value among the M-H edges is the RPE interface. Compare to the Markove method, our method is less dependent on neighbor A-scans and simply extend the boundaries by choosing the longest continuous edges. Results:We have evaluated 226 OCT retinal images with the Markov method and our new algorithm to make comparison (see Table). The results show the new method of detecting retinal thickness is superior to the other two methods, since OCT software is prone to both errors and artifacts and Markov method is only robust to healthy retina. Conclusion:Optical Coherence Tomography is a method of generating high-resolution cross-sectional images of the retina and vitreous. The main use of OCT images is to guide treatment decisions and improve differential diagnosis. The software allows generating secondary images consisting of interpolated retinal thickness maps. These thickness maps depend on accurate determination of retinal thickness in each underlying A-scan. We have shown that the current software is prone to artifacts and are presenting an improved algorithm for measuring retinal thickness. This algorithm should proof useful for clinical applications of retinal thickness mapping with the OCT.
Keywords: 430 imaging/image analysis: clinical • 429 image processing • 308 age-related macular degeneration