Pairs Of Retinal Images Research Articles

Retinal image registration and comparison for clinical decision support

For eye diseases, such as glaucoma and age-related macular degeneration (ARMD), involved in long-term degeneration procedure, longitudinal comparison of retinal images is a common step for reliable diagnosis of these kinds of diseases.To provide a retinal image registration approach for longitudinal retinal image alignment and comparison.Two image registration solutions were proposed for facing different image qualities of retinal images to make the registration methods more robust and feasible in a clinical application system.Thirty pairs of longitudinal retinal images were used for the registration test. The experiments showed both solutions provided good performance for the accurate image registrations with efficiency.We proposed a set of retinal image registration solutions for longitudinal retinal image observation and comparison targeting a clinical application environment.

Retinal image restoration by means of blind deconvolution

Retinal imaging plays a key role in the diagnosis and management of ophthalmologic disorders, such as diabetic retinopathy, glaucoma, and age-related macular degeneration. Because of the acquisition process, retinal images often suffer from blurring and uneven illumination. This problem may seriously affect disease diagnosis and progression assessment. Here we present a method for color retinal image restoration by means of multichannel blind deconvolution. The method is applied to a pair of retinal images acquired within a lapse of time, ranging from several minutes to months. It consists of a series of preprocessing steps to adjust the images so they comply with the considered degradation model, followed by the estimation of the point-spread function and, ultimately, image deconvolution. The preprocessing is mainly composed of image registration, uneven illumination compensation, and segmentation of areas with structural changes. In addition, we have developed a procedure for the detection and visualization of structural changes. This enables the identification of subtle developments in the retina not caused by variation in illumination or blur. The method was tested on synthetic and real images. Encouraging experimental results show that the method is capable of significant restoration of degraded retinal images.

Robust Multiscale Stereo Matching from Fundus Images with Radiometric Differences

A robust multiscale stereo matching algorithm is proposed to find reliable correspondences between low contrast and weakly textured retinal image pairs with radiometric differences. Existing algorithms designed to deal with piecewise planar surfaces with distinct features and Lambertian reflectance do not apply in applications such as 3D reconstruction of medical images including stereo retinal images. In this paper, robust pixel feature vectors are formulated to extract discriminative features in the presence of noise in scale space, through which the response of low-frequency mechanisms alter and interact with the response of high-frequency mechanisms. The deep structures of the scene are represented with the evolution of disparity estimates in scale space, which distributes the matching ambiguity along the scale dimension to obtain globally coherent reconstructions. The performance is verified both qualitatively by face validity and quantitatively on our collection of stereo fundus image sets with ground truth, which have been made publicly available as an extension of standard test images for performance evaluation.

Evaluation of Vascular Disease Progression in Retinopathy of Prematurity Using Static and Dynamic Retinal Images

To measure accuracy and speed for detection of vascular progression in retinopathy of prematurity (ROP) from serial images. Two strategies are compared: static side-by-side presentation and dynamic flickering of superimposed image pairs. Prospective comparative study. Fifteen de-identified, wide-angle retinal image pairs were taken from infants who eventually developed plus disease. Image pairs representing vascular disease progression were taken ≥1 week apart, and control images without progression were taken on the same day. Dynamic flickering pairs were created by digital image registration. Ten experts independently reviewed each image pair on a secure website using both strategies, and were asked to identify progression or state that images were identical. Accuracy and speed were measured, using examination date and ophthalmoscopic findings as a reference standard. Using static images, experts were accurate in a mean (%) ± standard deviation (SD) of 11.4 of 15 (76%) ± 1.7 image pairs. Using dynamic flickering images, experts were accurate in a mean (%) ± SD of 11.3 of 15 (75%) ± 1.7 image pairs. There was no significant difference in accuracy between these strategies (P = .420). Diagnostic speed was faster using dynamic flickering (24.7 ± 8.3 seconds) vs static side-by-side images (40.3 ± 18.3 seconds) (P = .002). Experts reported higher confidence when interpreting dynamic flickering images (P = .001). Retinal imaging provides objective documentation of vascular appearance, with potentially improved ability to recognize ROP progression compared to standard ophthalmoscopy. Speed of identifying vascular progression was faster by review of dynamic flickering image pairs than by static side-by-side images, although there was no difference in accuracy.

Automated determination of cup-to-disc ratio for classification of glaucomatous and normal eyes on stereo retinal fundus images

Early diagnosis of glaucoma, which is the second leading cause of blindness in the world, can halt or slow the progression of the disease. We propose an automated method for analyzing the optic disc and measuring the cup-to-disc ratio (CDR) on stereo retinal fundus images to improve ophthalmologists' diagnostic efficiency and potentially reduce the variation on the CDR measurement. The method was developed using 80 retinal fundus image pairs, including 25 glaucomatous, and 55 nonglaucomatous eyes, obtained at our institution. A disc region was segmented using the active contour method with the brightness and edge information. The segmentation of a cup region was performed using a depth map of the optic disc, which was reconstructed on the basis of the stereo disparity. The CDRs were measured and compared with those determined using the manual segmentation results by an expert ophthalmologist. The method was applied to a new database which consisted of 98 stereo image pairs including 60 and 30 pairs with and without signs of glaucoma, respectively. Using the CDRs, an area under the receiver operating characteristic curve of 0.90 was obtained for classification of the glaucomatous and nonglaucomatous eyes. The result indicates potential usefulness of the automated determination of CDRs for the diagnosis of glaucoma.

Automatic point correspondence using an artificial immune system optimization technique for medical image registration

In this paper, an automatic method for determining pairs of corresponding points between medical images is proposed. The method is based on the implementation of an artificial immune system (AIS). AIS is a relatively novel, population based category of algorithms, inspired by theoretical immunologic models. When used as function optimizers, AIS have the attractive property of locating the global optimum of a function as well as a large number of strong local optimum points. In this work, AIS has been applied both for the extraction of an optimal set of candidate points on the reference image and the definition of their corresponding ones on the second image. The performance of the proposed AIS algorithm is evaluated against the widely used Iterative Closest Point (ICP) algorithm in terms of the accuracy of the obtained correspondences and in terms of the accuracy of the point-based registration by the two correspondence algorithms and the Mutual Information criterion, as an intensity-based registration method. Qualitative and quantitative results involving 92 X-ray dental and 10 retinal image pairs subject to known and unknown transformations are presented. The results indicate a superior performance of the proposed AIS algorithm with respect to the ICP algorithm and the Mutual Information, in terms of both correct correspondence and registration accuracy.

A Partial Intensity Invariant Feature Descriptor for Multimodal Retinal Image Registration

Detection of vascular bifurcations is a challenging task in multimodal retinal image registration. Existing algorithms based on bifurcations usually fail in correctly aligning poor quality retinal image pairs. To solve this problem, we propose a novel highly distinctive local feature descriptor named partial intensity invariant feature descriptor (PIIFD) and describe a robust automatic retinal image registration framework named Harris-PIIFD. PIIFD is invariant to image rotation, partially invariant to image intensity, affine transformation, and viewpoint/perspective change. Our Harris-PIIFD framework consists of four steps. First, corner points are used as control point candidates instead of bifurcations since corner points are sufficient and uniformly distributed across the image domain. Second, PIIFDs are extracted for all corner points, and a bilateral matching technique is applied to identify corresponding PIIFDs matches between image pairs. Third, incorrect matches are removed and inaccurate matches are refined. Finally, an adaptive transformation is used to register the image pairs. PIIFD is so distinctive that it can be correctly identified even in nonvascular areas. When tested on 168 pairs of multimodal retinal images, the Harris-PIIFD far outperforms existing algorithms in terms of robustness, accuracy, and computational efficiency.

Disparity-energy signals in perceived stereoscopic depth

Stereopsis, the ability to sense the world in three dimensions (3D) from pairs of retinal images, functions when both images have corresponding elements. When observers view stereograms lacking a global match, they do not perceive 3D structure, whereas several cortical areas encode stereoscopic depth in the disparity energy. Whether these neural representations are exploited or ignored in perceptual decisions remains elusive. By combining contrast-reversal and delay between stereo images, we found that disparity-energy signals mediate the reversal of stereoscopic depth judgments. A crisp, adjacent plane of reference was crucial for the signal to be used in the judgments. Disparity discrimination relies on the disparity-energy signal when the stimulus has no global binocular match and is accompanied by a fixed surface of reference.

The logic of how-questions

Philosophers and scientists are concerned with the why and the how of things. Questions like the following are so much grist for the philosopher’s and scientist’s mill: How can we be free and yet live in a deterministic universe?, How do neural processes give rise to conscious experience?, Why does conscious experience accompany certain physiological events at all?, How is a three-dimensional perception of depth generated by a pair of two-dimensional retinal images?. Since Belnap and Steel’s pioneering work on the logic of questions, Van Fraassen has managed to apply their approach in constructing an account of the logic of why-questions. Comparatively little, by contrast, has been written on the logic of how-questions despite the apparent centrality of questions such as How is it possible for us to be both free and determined? to philosophical enterprise.1 In what follows I develop a logic for how-questions of various sorts including how-questions of cognitive resolution, how-questions of manner, how-questions of method, of means, and of mechanism.

Image registration and subtraction for the visualization of change in diabetic retinopathy screening

Diabetic retinopathy (DR) is a leading cause of blindness in people of working age in developed countries. DR is characterized by lesions of the retinal microvasculature. This paper describes a technique for visualizing the changes which occur in retinal pathology over a 1 year interval. Pairs of retinal images of the same eye were acuired 12 months apart within a DR screening programme. The images were normalized, registered and subtracted to generate a contrast-enhanced difference image displaying changes in retinal appearance. The normalization factor and registration error were quantified for 104 pairs of retinal images.

Adaptive optics enhanced simultaneous en-face optical coherence tomography and scanning laser ophthalmoscopy

A novel combination of adaptive optics with a simultaneous en-face OCT/SLO system for high resolution imaging of the in-vivo human retina is presented. Pairs of retinal images are shown and performance of the system is evaluated with and without dynamic wavefront correction. The adaptive optics closed loop system operates at a frame rate of 9 Hz and incorporates a Shack-Hartmann wavefront sensor based on a highly sensitive Andor camera and a 37 actuator OKO membrane deformable mirror to correct for ocular aberrations. The system produces C-scan pairs of images at a frame rate of 2 Hz. The correction of aberrations produced by the adaptive optics closed-loop system increased the signal-to-noise ratio in images obtained from volunteer eyes by up to 6 dB in the OCT channel and up to 9 dB in the SLO channel. A slight improvement in the lateral resolution was also obtained, from 6.5 mum before to 5 mum after closing the adaptive optics loop.

Improved contrast of subretinal structures using polarization analysis.

To improve the ability to detect and quantify the early retinal changes associated with aging, age-related maculopathy, and age-related macular degeneration. A computational approach was implemented for analyzing images using a readily available polarimeter that is used for glaucoma diagnosis. This device, the GDx Nerve Fiber Analyzer (Laser Diagnostic Technologies, Inc., San Diego, CA), takes a series of images as a function of the polarization angle of the illuminating light. For each of 20 input polarizations, pairs of retinal images are digitized. One image is made of the light returning from the eye that is polarized parallel to the input light, and the other image is made of the light that is rotated by 90 degrees from the input polarization. Using the raw data from these 40 images, and a simplified model of the polarization properties of the eye, we calculated the amount of light that returns in a parallel polarized state, and the amount of light that is depolarized by multiple scattering. Measurements were made in seven subjects with small drusen. The depolarized light image produced a 3.4 times higher contrast of drusen and subretinal changes than the parallel polarized light images. Polarization-sensitive imaging combined with a simple computational approach allows the measurement of the retinal distribution of multiply scattered light. With this technique, retinal imaging of age-related changes in retinal and subretinal tissue can be improved.

A feature-based technique for joint, linear estimation of high-order image-to-mosaic transformations: mosaicing the curved human retina

An algorithm for constructing image mosaics from multiple, uncalibrated, weak-perspective views of the human retina is presented and analyzed. It builds on an algorithm for registering pairs of retinal images using a noninvertible, 12-parameter, quadratic image transformation model and hierarchical, robust estimation. The major innovation presented is a linear, feature-based, noniterative method for jointly estimating consistent transformations of all images onto the mosaic image. Constraints for this estimation are derived from pairwise registration both directly with the anchor image and indirectly between pairs of nonanchor images. An incremental, graph-based technique constructs the set of registered image pairs used in the solution. The estimation technique allows images that do not overlap the anchor frame to be successfully mosaiced, a valuable capability for mosaicing images of the retinal periphery. Experimental analysis on data sets from 16 eyes shows the average overall median transformation error in final mosaic to be 0.76 pixels. The technique is simpler, more accurate, and offers broader coverage than previously published methods.

Contributions of the cornea and the lens to the aberrations of the human eye

The relative contributions of optical aberrations of the cornea and the crystalline lens to the final image quality of the human eye were studied. The aberrations of the entire eye were obtained from pairs of double-pass retinal images, and the aberrations of the cornea were obtained from videokeratographic data. Third-order spherical aberration and coma were significantly larger for the cornea than for the complete eye, indicating a significant role of the lens in compensating for corneal aberrations. In a second experiment retinal images were recorded in an eye before and after we neutralized the aberrations of the cornea by having the subjects wear swimming goggles filled with saline water, providing a direct estimate of the optical performance of the crystalline lens.

Estimates of the ocular wave aberration from pairs of double-pass retinal images.

We apply a computational technique to retrieve the wave aberration of the eye from the point-spread function obtained from pairs of double-pass retinal images. The method consists of an adapted pyramidal version of a nonlinear least-squares fitting procedure to a wave aberration expressed as an expansion in Zernike polynomials. Although the procedure provides accurate estimates of the wave aberration, it presents several drawbacks that are discussed in detail. In particular, since a great deal of computational time is necessary to retrieve a single wave aberration, this technique is not useful for real-time applications. We present results of wave aberrations in five normal subjects in the fovea for a 4-mm-pupil diameter. In every case there is a clear presence of comalike aberrations, while the third-order spherical aberration is usually smaller than previous estimates. The root-mean-square error in the retrieved wave aberration, when defocus and astigmatism were corrected, ranges from 0.24 to 0.5 wavelength. The particular values of the aberration coefficients present a large intersubject variability.

Stereo capture: local rematching driven by binocularly attended 3-D configuration rather than retinal images.

Previous explanations for stereo capture were mainly based on the low-level perceptual processing of binocular stereopsis which usually shows that one pair of retinal images corresponds to only one 3-D perceptual configuration. Stereo capture, however, may encounter multiple perceptual configurations due to the matching ambiguity of wallpaper elements that may not be solved merely by bottom-up processing of the retinal stimuli. The present study suggests that binocular attention plays an important role in stereo capture by way of selecting and enhancing a perceptual configuration that is often ambiguous without attention involved. Stereo capture results from wallpaper's local rematching driven by binocularly attended 3-D configuration rather than retinal images.

The role of the vertical dimension in stereoscopic vision.

It is shown how the full three-dimensional structure of a scene may in principle be computed from a correlated pair of retinal images, when all that is known about the orientations of the two eyes is that the planes of their horizontal meridians accurately coincide. The vertical dimension plays a crucial role in the theory; visible points which lie on the horizontal meridian supply no information about the angle of convergence or the direction of gaze. (i) If the scene contains three or more nonmeridional points, not all lying in a vertical plane, then their positions in space are fully determined by the horizontal and vertical coordinates of their images on the two retinas. (ii) If just two nonmeridional points are visible, or more than two, lying in a vertical plane, then their retinal images admit, in general, just two distinct three-dimensional interpretations. One of these is usually unrealistic; but a choice between them may be perceptually difficult if the vertical plane containing the points is nearly perpendicular to the interocular axis. These results suggest that vertical disparities may play an important role in the binocular perception of absolute depth. Elsewhere (Mayhew, this issue) this suggestion is found to provide a quantitative explanation of Ogle's 'induced effect'.

A computational model of binocular depth perception.

We point out that the horizontal disparities between a pair of retinal images are inadequate for computing the three-dimensional structure of a scene unless supplemented by independent information about the distance and direction of the fixation point. We suggest that this supplementary information is derived not from non-visual sources, but from the vertical disparities of a few non-meridional image points. This hypothesis is shown to account quantitatively for Ogle's induced effect--the marked distortion of a scene by a vertically magnifying lens placed in front of one eye.