Evaluation of potential image acquisition pitfalls during optical coherence tomography and their influence on retinal image segmentation.

The development of improved segmentation algorithms for more consistently accurate detection of retinal boundaries is a potentially useful solution to the limitations of existing optical coherence tomography (OCT) software. We modeled artifacts related to operator errors that may normally occur during OCT imaging and evaluated their influence on segmentation results using a novel segmentation algorithm. These artifacts included: defocusing, depolarization, decentration, and a combination of defocusing and depolarization. Mean relative reflectance and average thickness of the automatically extracted intraretinal layers was then measured. Our results show that defocusing and depolarization errors together have the greatest altering effect on all measurements and on segmentation accuracy. A marked decrease in mean relative reflectance and average thickness was observed due to depolarization artifact in all intraretinal layers, while defocus resulted in a less-marked decrease. Decentration resulted in a marked but not significant change in average thickness. Our study demonstrates that care must be taken for good-quality imaging when measurements of intraretinal layers using the novel algorithm are planned in future studies. An awareness of these pitfalls and their possible solutions is crucial for obtaining a better quantitative analysis of clinically relevant features of retinal pathology.

Comparison of PDE-based nonlinear diffusion approaches for image enhancement and denoising in optical coherence tomography.

A comparison between two nonlinear diffusion methods for denoising OCT images is performed. Specifically, we compare and contrast the performance of the traditional nonlinear Perona-Malik filter with a complex diffusion filter that has been recently introduced by Gilboa et al.. The complex diffusion approach based on the generalization of the nonlinear scale space to the complex domain by combining the diffusion and the free Schridinger equation is evaluated on synthetic images and also on representative OCT images at various noise levels. The performance improvement over the traditional nonlinear Perona-Malik filter is quantified in terms of noise suppression, image structural preservation and visual quality. An average signal-to-noise ratio (SNR) improvement of about 2.5 times and an average contrast to noise ratio (CNR) improvement of 49% was obtained while mean structure similarity (MSSIM) was practically not degraded after denoising. The nonlinear complex diffusion filtering can be applied with success to many OCT imaging applications. In summary, the numerical values of the image quality metrics along with the qualitative analysis results indicated the good feature preservation performance of the complex diffusion process, as desired for better diagnosis in medical imaging processing.

Delineating fluid-filled region boundaries in optical coherence tomography images of the retina.

We evaluate the ability of a deformable model to yield accurate shape descriptions of fluid-filled regions associated with age-related macular degeneration. Calculation of retinal thickness and volume by the current optical coherence tomography (OCT) system includes fluid-filled regions or lesions along with actual retinal tissue. In order to quantify these lesions independently from the retinal tissue, they must be outlined. A deformable model was applied to OCT images of retinas demonstrating cystoids and subretinal fluid spaces. Several implementation issues were addressed in order to choose appropriate parameters. The use of a nonlinear anisotropic diffusion filter to suppress speckle noise while at the same time preserving the edges of the original image was explored. Once the contours of the lesions were outlined, quantitative analysis of the surface area and volume of the lesions was performed. The deformable model could accurately outline fluid-filled regions within the retina. The detection method tested proved effective in capturing the complexity of fluid-filled regions in OCT images. Deformable models combined with nonlinear anisotropic diffusion filtering show promise in the detection of retinal features of interest for diagnosis in clinical OCT images. Thus, fluid-filled region detection may significantly aid in analysis of treatments and diagnosis.