Sina Farsiu

Fast and robust multiframe super resolution

Super-resolution reconstruction produces one or a set of high-resolution images from a set of low-resolution images. In the last two decades, a variety of super-resolution methods have been proposed. These methods are usually very sensitive to their assumed model of data and noise, which limits their utility. This paper reviews some of these methods and addresses their shortcomings. We propose an alternate approach using L/sub 1/ norm minimization and robust regularization based on a bilateral prior to deal with different data and noise models. This computationally inexpensive method is robust to errors in motion and blur estimation and results in images with sharp edges. Simulation results confirm the effectiveness of our method and demonstrate its superiority to other super-resolution methods.

Kernel Regression for Image Processing and Reconstruction

In this paper, we make contact with the field of nonparametric statistics and present a development and generalization of tools and results for use in image processing and reconstruction. In particular, we adapt and expand kernel regression ideas for use in image denoising, upscaling, interpolation, fusion, and more. Furthermore, we establish key relationships with some popular existing methods and show how several of these algorithms, including the recently popularized bilateral filter, are special cases of the proposed framework. The resulting algorithms and analyses are amply illustrated with practical examples

Multiframe demosaicing and super-resolution of color images

In the last two decades, two related categories of problems have been studied independently in image restoration literature: super-resolution and demosaicing. A closer look at these problems reveals the relation between them, and, as conventional color digital cameras suffer from both low-spatial resolution and color-filtering, it is reasonable to address them in a unified context. In this paper, we propose a fast and robust hybrid method of super-resolution and demosaicing, based on a maximum a posteriori estimation technique by minimizing a multiterm cost function. The L/sub 1/ norm is used for measuring the difference between the projected estimate of the high-resolution image and each low-resolution image, removing outliers in the data and errors due to possibly inaccurate motion estimation. Bilateral regularization is used for spatially regularizing the luminance component, resulting in sharp edges and forcing interpolation along the edges and not across them. Simultaneously, Tikhonov regularization is used to smooth the chrominance components. Finally, an additional regularization term is used to force similar edge location and orientation in different color channels. We show that the minimization of the total cost function is relatively easy and fast. Experimental results on synthetic and real data sets confirm the effectiveness of our method.

Automatic segmentation of seven retinal layers in SDOCT images congruent with expert manual segmentation

Segmentation of anatomical and pathological structures in ophthalmic images is crucial for the diagnosis and study of ocular diseases. However, manual segmentation is often a time-consuming and subjective process. This paper presents an automatic approach for segmenting retinal layers in Spectral Domain Optical Coherence Tomography images using graph theory and dynamic programming. Results show that this method accurately segments eight retinal layer boundaries in normal adult eyes more closely to an expert grader as compared to a second expert grader.

Photoreceptor Layer Thinning over Drusen in Eyes with Age-Related Macular Degeneration Imaged In Vivo with Spectral-Domain Optical Coherence Tomography

PURPOSE Detect changes in the neurosensory retina using spectral-domain optical coherence tomography (SD OCT) imaging over drusen in age-related macular degeneration (AMD). Quantitative imaging biomarkers may aid in defining risk of disease progression. DESIGN Cross-sectional, case-control study evaluating SD OCT testing in AMD. PARTICIPANTS AND CONTROLS Seventeen eyes of 12 subjects with nonneovascular AMD and drusen and 17 eyes of 10 age-matched control subjects. METHODS Spectral-domain OCT imaging across the fovea in the study eye with multiple 10- to 12-mm scans of 1000 A scans each. MAIN OUTCOME MEASURES In summed SD OCT scans, the height of individual retinal layers either over drusen or at corresponding locations in the control eye and qualitative changes in retinal layers over drusen. Secondary measures included photoreceptor layer (PRL) area, inner retinal area, and retinal pigment epithelium (RPE)/drusen area. RESULTS The PRL was thinned over 97% of drusen, average PRL thickness was reduced by 27.5% over drusen compared with over a similar location in controls, and the finding of a difference was valid and significant (P=0.004). Photoreceptor outer segments were absent over at least 1 druse in 47% of eyes. Despite thinning of the PRL, inner retinal thickness remained unchanged. We observed 2 types of hyperreflective abnormalities in the neurosensory retina over drusen. Distinct hyperreflective speckled patterns occurred over drusen in 41% of AMD eyes and never in control eyes. A prominent hyperreflective haze was present in the photoreceptor nuclear layer over drusen in 67% of AMD eyes and more subtly in the photoreceptor nuclear layer in 18% of control eyes (no drusen). CONCLUSIONS With SD OCT as used in this study, we can easily detect and measure changes in PRL over drusen. Decreased PRL thickness over drusen suggests a degenerative process, with cell loss leading to decreased visual function. The hyperreflective foci overlying drusen are likely to represent progression of disease RPE cell migration into the retina and possible photoreceptor degeneration or glial scar formation. A longitudinal study using SD OCT to examine and measure the neurosensory retina over drusen will resolve the timeline of degenerative changes relative to druse formation.

Deblurring Using Regularized Locally Adaptive Kernel Regression

Kernel regression is an effective tool for a variety of image processing tasks such as denoising and interpolation . In this paper, we extend the use of kernel regression for deblurring applications. In some earlier examples in the literature, such nonparametric deblurring was suboptimally performed in two sequential steps, namely denoising followed by deblurring. In contrast, our optimal solution jointly denoises and deblurs images. The proposed algorithm takes advantage of an effective and novel image prior that generalizes some of the most popular regularization techniques in the literature. Experimental results demonstrate the effectiveness of our method.

Anti-amyloid therapy protects against retinal pigmented epithelium damage and vision loss in a model of age-related macular degeneration

Age-related macular degeneration (AMD) is a leading cause of visual dysfunction worldwide. Amyloid β (Aβ) peptides, Aβ1–40 (Aβ40) and Aβ1–42 (Aβ42), have been implicated previously in the AMD disease process. Consistent with a pathogenic role for Aβ, we show here that a mouse model of AMD that invokes multiple factors that are known to modify AMD risk (aged human apolipoprotein E 4 targeted replacement mice on a high-fat, cholesterol-enriched diet) presents with Aβ-containing deposits basal to the retinal pigmented epithelium (RPE), histopathologic changes in the RPE, and a deficit in scotopic electroretinographic response, which is reflective of impaired visual function. Strikingly, these electroretinographic deficits are abrogated in a dose-dependent manner by systemic administration of an antibody targeting the C termini of Aβ40 and Aβ42. Concomitant reduction in the levels of Aβ and activated complement components in sub-RPE deposits and structural preservation of the RPE are associated with anti-Aβ40/42 antibody immunotherapy and visual protection. These observations are consistent with the reduction in amyloid plaques and improvement of cognitive function in mouse models of Alzheimers disease treated with anti-Aβ antibodies. They also implicate Aβ in the pathogenesis of AMD and identify Aβ as a viable therapeutic target for its treatment.

Imaging the Infant Retina with a Hand-held Spectral-Domain Optical Coherence Tomography Device

PURPOSE To evaluate and treat infant retina through the use of a hand-held spectral-domain optical coherence tomography (SD OCT) device in selected cases of Shaken Baby syndrome (SBS). DESIGN Observational case series. METHODS A novel SD OCT system was optimized and evaluated for infant imaging. An adult eye was imaged with both a standard clinical SD OCT system and the hand-held system. Four eyes of two infants with a history of SBS were imaged with the hand-held system. One infant was imaged again during follow-up examinations. Robust image processing algorithms were developed to create high-quality images. Images were assessed for usefulness in demonstrating pertinent morphologic features. RESULTS The novel SD OCT unit proved effective for data acquisition and comparable with conventional chin-rest SD OCT. Rapid data acquisition limited motion artifact within the B-scan, although there was slight motion between B scans. The SD OCT images provided previously unseen details with regard to the morphologic features of retinal lesions in these infant eyes. This information influenced prognosis and management. CONCLUSIONS As with adults, the hand-held customized SD OCT proved to be an invaluable tool in the differentiation of disease processes or injury in these eyes under study. SD OCT imaging systems may be considered a useful adjunct to RetCam fundus photography for assessment and clinical management in cases of SBS.

Validated Automatic Segmentation of AMD Pathology Including Drusen and Geographic Atrophy in SD-OCT Images

PURPOSE To automatically segment retinal spectral domain optical coherence tomography (SD-OCT) images of eyes with age-related macular degeneration (AMD) and various levels of image quality to advance the study of retinal pigment epithelium (RPE)+drusen complex (RPEDC) volume changes indicative of AMD progression. METHODS A general segmentation framework based on graph theory and dynamic programming was used to segment three retinal boundaries in SD-OCT images of eyes with drusen and geographic atrophy (GA). A validation study for eyes with nonneovascular AMD was conducted, forming subgroups based on scan quality and presence of GA. To test for accuracy, the layer thickness results from two certified graders were compared against automatic segmentation results for 220 B-scans across 20 patients. For reproducibility, automatic layer volumes were compared that were generated from 0° versus 90° scans in five volumes with drusen. RESULTS The mean differences in the measured thicknesses of the total retina and RPEDC layers were 4.2 ± 2.8 and 3.2 ± 2.6 μm for automatic versus manual segmentation. When the 0° and 90° datasets were compared, the mean differences in the calculated total retina and RPEDC volumes were 0.28% ± 0.28% and 1.60% ± 1.57%, respectively. The average segmentation time per image was 1.7 seconds automatically versus 3.5 minutes manually. CONCLUSIONS The automatic algorithm accurately and reproducibly segmented three retinal boundaries in images containing drusen and GA. This automatic approach can reduce time and labor costs and yield objective measurements that potentially reveal quantitative RPE changes in longitudinal clinical AMD studies. (ClinicalTrials.gov number, NCT00734487.).

Integration of a spectral domain optical coherence tomography system into a surgical microscope for intraoperative imaging.

PURPOSE To demonstrate an operating microscope-mounted spectral domain optical coherence tomography (MMOCT) system for human retinal and model surgery imaging. METHODS A prototype MMOCT system was developed to interface directly with an ophthalmic surgical microscope, to allow SDOCT imaging during surgical viewing. Nonoperative MMOCT imaging was performed in an Institutional Review Board-approved protocol in four healthy volunteers. The effect of surgical instrument materials on MMOCT imaging was evaluated while performing retinal surface, intraretinal, and subretinal maneuvers in cadaveric porcine eyes. The instruments included forceps, metallic and polyamide subretinal needles, and soft silicone-tipped instruments, with and without diamond dusting. RESULTS High-resolution images of the human retina were successfully obtained with the MMOCT system. The optical properties of surgical instruments affected the visualization of the instrument and the underlying retina. Metallic instruments (e.g., forceps and needles) showed high reflectivity with total shadowing below the instrument. Polyamide material had a moderate reflectivity with subtotal shadowing. Silicone instrumentation showed moderate reflectivity with minimal shadowing. Summed voxel projection MMOCT images provided clear visualization of the instruments, whereas the B-scans from the volume revealed details of the interactions between the tissues and the instrumentation (e.g., subretinal space cannulation, retinal elevation, or retinal holes). CONCLUSIONS High-quality retinal imaging is feasible with an MMOCT system. Intraoperative imaging with model eyes provides high-resolution depth information including visualization of the instrument and intraoperative tissue manipulation. This study demonstrates a key component of an interactive platform that could provide enhanced information for the vitreoretinal surgeon.