A Novel Deep Learning Pipeline for Retinal Vessel Detection In Fluorescein Angiography

Abstract

While recent advances in deep learning have significantly advanced the state of the art for vessel detection in color fundus (CF) images, the success for detecting vessels in fluorescein angiography (FA) has been stymied due to the lack of labeled ground truth datasets. We propose a novel pipeline to detect retinal vessels in FA images using deep neural networks (DNNs) that reduces the effort required for generating labeled ground truth data by combining two key components: crossmodality transfer and human-in-the-loop learning. The cross-modality transfer exploits concurrently captured CF and fundus FA images. Binary vessels maps are first detected from CF images with a pre-trained neural network and then are geometrically registered with and transferred to FA images via robust parametric chamfer alignment to a preliminary FA vessel detection obtained with an unsupervised technique. Using the transferred vessels as initial ground truth labels for deep learning, the human-in-the-loop approach progressively improves the quality of the ground truth labeling by iterating between deep-learning and labeling. The approach significantly reduces manual labeling effort while increasing engagement. We highlight several important considerations for the proposed methodology and validate the performance on three datasets. Experimental results demonstrate that the proposed pipeline significantly reduces the annotation effort and the resulting deep learning methods outperform prior existing FA vessel

Automated vessel density detection in fluorescein angiography images correlates with vision in proliferative diabetic retinopathy

“…The highlighted rectangular regions in the leftmost image in Fig 2 contain many fine vessels that are difficult to identify in the original UWFFA images. Our automated detection algorithm, however, is not affected by changes in image contrast and is able to detect these vessels [21]. Careful contrast enhancement of these regions validates that the fine vessels are actually present.…”

“…Such an automated technique for UWFFA would provide the ability to study longitudinal changes in vessels and relationships between vessel density and clinical endpoints, such as visual acuity and CRT. Our group has worked to address this unmet need by developing an automated algorithm that can detect retinal vasculature on UWFFA using a deep learning approach [21].…”

“…The trained network achieved the area under the Precision-Recall curve of 0.930. Additional, technical details can be found in an independent publication [21], parts of which have been reported in preliminary form [19,26,27]. We apply the trained deep neural network to the (baseline) UWFFA images used in this study.…”

“…The output from the deep neural network is a vessel map where pixel intensity, ranging from 0 to 1, indicates the probability of a corresponding pixel being a vessel. The probabilistic vessel map is then converted into the binary representation by setting a threshold of 0.5, which is shown to be close to the optimal threshold [21].…”

Weakly-Supervised Vessel Detection in Ultra-Widefield Fundus Photography Via Iterative Multi-Modal Registration and Learning

“…The robustness against such outliers is particularly crucial in this application setting because, as noted earlier, some fine vessels appear only in Q a because the FA modality detects these much better than FP. We refer readers to [36] for detailed derivations of the parameter estimation with the EM approach. Here we only note that the key intuition can be understood from the fact that, in the EM approach, the arithmetic average in (1) is replaced by a weighted average where the weight for the squared error D j (q c k , q a j ) corresponding to the j th point in Q a corresponds to the estimated posterior probability that it is not an outlier (and has a corresponding point in Q c ).…”