Enhancing vessel visibility in fundus images to aid the diagnosis of retinopathy of prematurity

Abstract

Retinopathy of prematurity is a disease that can affect premature or in similar conditions babies. For diagnosing of retinopathy of prematurity, the infant is examined as soon as possible. Due to the nature of the examination, the images obtained are poor in quality. This article presents an automated method for processing fundus images to improve the visibility of the vascular network. The method includes several processing tasks whose parameters are predicted using an artificial neural network. A set of 88 clinical images were used in this work. The performance of our proposal is efficient, and the average processing time was 42 ms. The method was assessed using both the contrast improvement index and expert opinions. The contrast improvement index average was 2; this means the processed image successfully improved its contrast. Three pediatric ophthalmologists validated the proposed method and agreed that the visual enhancement can help observe clearly the retinal vessels.

Emerging Trends and Research Foci in Artificial Intelligence for Retinal Diseases: Bibliometric and Visualization Study (Preprint)

“…Zhou et al [44] proposed the learning of discriminative CNN features and enhanced thin vessels in color fundus images to further improve the segmentation performance. This algorithm improves the contrast of the retinal vessels and was verified by three pediatric ophthalmologists [45]. Goel et al [46] showed that using a development learning model to transfer learning can improve the accuracy of correct classification of different aneurysms in the retina area caused by DR. AI technology can improve diagnosis accuracy and can also save time for both doctors and patients by increasing the contrast between image structures, such as segmenting distinct blood arteries or calculating normal and pathological structures.…”

Emerging Trends and Research Foci in Artificial Intelligence for Retinal Diseases: Bibliometric and Visualization Study

“…Zhou et al [ 44 ] proposed the learning of discriminative CNN features and enhanced thin vessels in color fundus images to further improve the segmentation performance. This algorithm improves the contrast of the retinal vessels and was verified by three pediatric ophthalmologists [ 45 ]. Goel et al [ 46 ] showed that using a development learning model to transfer learning can improve the accuracy of correct classification of different aneurysms in the retina area caused by DR.…”

Features extraction using encoded local binary pattern for detection and grading diabetic retinopathy

“…The candidate regions of exudates were detected using the k-means clustering technique, which provides a good result when the retinal OD is fully visible and fails if only a portion of the OD is visible. Some researchers confirm the need for pre-processing to improve the image quality [13,14]. Pazmino et al [13] presented a method for processing fundus images to improve the visibility of the vascular network.…”

“…Some researchers confirm the need for pre-processing to improve the image quality [13,14]. Pazmino et al [13] presented a method for processing fundus images to improve the visibility of the vascular network. Luangruangrong [14] used contrast limited adaptive histogram equalization (CLAHE) for image enhancement, then optic disk and blood vessel detection followed by classifying exudates using hierarchical fuzzy-c-mean clustering.…”