Different optical methods for retinal imaging provide a significant improvement for image analysis and help with data interpretation. The use of tunable light sources, which have been optimized for contrast enhancement of various retinal features or lesions in retinal images, could simplify the eye fundus examination through enhanced image quality. In this study, we have developed and described a research prototype which consists of a spectrally tunable light source based on a digital micromirror device which is further coupled to a fundus camera. The overall aim of this construction was to generate illuminations optimized for enhanced retinal image feature visibility. The optimized illumination conditions were compared to traditional red-free imaging and the measurements were executed for an artificial eye followed by in vivo measurements of the eyes of three volunteers. In all cases, the retinal image contrast was observed to improve compared to the traditional red-free imaging. Depending on the observed retinal feature, the perceptual improvements in the contrast varied from a few percent to nearly 70 percent.
Abstract. Spectral retinal images have signi cant potential for improving the early detection and visualization of subtle changes due to eye diseases and many systemic diseases. High resolution in both the spatial and the spectral domain can be achieved by capturing a set of narrowband channel images from which the spectral images are composed. With imaging techniques where the eye movement between the acquisition of the images is unavoidable, image registration is required. In this paper, the applicability of the state-of-the-art image registration methods for the composition of spectral retinal images is studied. The registration methods are quantitatively compared using synthetic channel image data of an eye phantom and semisynthetic set of retinal channel images subjected to known transformations. The experiments show that Generalized dual-bootstrap iterative closest point method outperforms the other evaluated methods in registration accuracy and the number of successful registrations.
Abstract. Spectral retinal images have signicant potential for improving the early detection and visualization of subtle changes due to eye diseases and many systemic diseases. High resolution in both the spatial and the spectral domain can be achieved by capturing a set of narrowband channel images from which the spectral images are composed. With imaging techniques where the eye movement between the acquisition of the images is unavoidable, image registration is required. In this paper, the applicability of the state-of-the-art image registration methods for the composition of spectral retinal images is studied. The registration methods are quantitatively compared using synthetic channel image data of an eye phantom and semisynthetic set of retinal channel images subjected to known transformations. The experiments show that Generalized dual-bootstrap iterative closest point method outperforms the other evaluated methods in registration accuracy and the number of successful registrations.
Abstract. Most feature-based lesion detection and computer-aided diagnosis methods for medical images require representative data of each region of interest (ROI) for parameter selection. Furthermore, the spatial accuracy of the segmentation of the ROIs from the background can significantly affect certain image features extracted from the ROIs. However, requiring spatially accurate manual segmentations of the ROIs to be used as the ground truth is infeasible for large image sets due to the amount of manual work involved. To relax the requirement of spatial accuracy and to enable spatial refinement of coarse manual segmentations to have more representative feature data, a method based on color information and maximally stable extremal regions of lesion likelihoods is presented. The proposed method is quantitatively compared to several segmentation approaches by using a challenging set of retinal images with spatially accurate ground truth of exudates. The experiments show that the proposed method produces good results measured as Dice coefficients between the refined segmentation and ground truth.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.