A recursive Bayesian approach to describe retinal vasculature geometry

Abstract

Demographic studies suggest that changes in the retinal vasculature geometry, especially in vessel width, are associated with the incidence or progression of eyerelated or systemic diseases. To date, the main information source for width estimation from fundus images has been the intensity profile between vessel edges. However, there are many factors affecting the intensity profile: pathologies, the central light reflex and local illumination levels, to name a few. In this study, we introduce three information sources for width estimation. These are the probability profiles of vessel interior, centreline and edge locations generated by a deep network. The probability profiles provide direct access to vessel geometry and are used in the likelihood calculation for a Bayesian method, particle filtering. We also introduce a geometric model which can handle non-ideal conditions of the probability profiles. Our experiments conducted on the REVIEW dataset yielded consistent estimates of vessel width, even in cases when one of the vessel edges is difficult to identify. Moreover, our results suggest that the method is better than human observers at locating edges of low contrast vessels.

One DAG to Rule Them All

“…Besides the already mentioned LUT technique, iterations based on decision trees have been proposed in [63], to further speedup the ZS algorithm. These solutions have been proposed some decades ago, but are still commonly used [25], [26], [27] and included in many image processing libraries, such as OpenCV. Contrary to CCL, each thinning proposal provides different outputs and the choice depends on the application needs.…”

“…E-mail: {name.surname}@unimore.it or to count objects [13]. Thinning, instead, is often used together with contour-tracing, morphological operators, and CCL, whenever a compact representation of the objects inside an image is required [20], as in fingerprint analysis [25], vasculature geometry detection [26], [27], and road mapping [15]. Therefore, also deep learning pipelines can benefit from efficient implementations of binary image processing algorithms.…”

Retinal Vessel Segmentation Combined With Generative Adversarial Networks and Dense U-Net

“…However, fixed segmentation rules often cannot match the diversity of vascular morphological distribution. The main idea of the supervised learning algorithms is to train the segmentation model using fundus images with segmentation annotations and allow the model to automatically extract the vascular features to achieve vessel segmentation, such as Bayes model-based algorithms [11], support vector machinebased algorithms [12], and deep learning-based algorithms [13][14][15][16]. However, supervised learning algorithms require huge data with manual label, which is hard to get.…”

Improving the Performance of Thinning Algorithms with Directed Rooted Acyclic Graphs

“…It can be defined as the successive removal of outermost layers of an object until only a skeleton of unit width remains [10]. Firstly introduced in the 1950 as a data compression strategy [11], the thinning procedure is nowadays used as a pre-or post-processing step in many different applications, ranging from medical imaging [33,34] to handwritten text recognition [7,19] and fingerprint analysis [22]. Therefore, having an efficient and effective algorithm is extremely important.…”

“…These solutions have been proposed some decades ago, but are still commonly used [22,33,34] and included in many image processing libraries, such as OpenCV [27].…”