VascX Models: Model Ensembles for Retinal Vascular Analysis from Color Fundus Images

TL;DR

VascX represents an important improvement in segmentation quality that translates into better vascular features to support more robust analyses of the retinal vasculature, relevant for the detection, prevention, and monitoring of disease.

Abstract

We introduce VascX models, a comprehensive set of model ensembles for analyzing retinal vasculature from color fundus images (CFIs). Annotated CFIs were aggregated from public datasets . Additional CFIs, mainly from the population-based Rotterdam Study were annotated by graders for arteries and veins at pixel level, resulting in a dataset diverse in patient demographics and imaging conditions. VascX models demonstrated superior segmentation performance across datasets, image quality levels, and anatomic regions when compared to existing, publicly available models, likely due to the increased size and variety of our training set. Important improvements were observed in artery-vein and disc segmentation performance, particularly in segmentations of these structures on CFIs of intermediate quality, common in large cohorts and clinical datasets. Importantly, these improvements translated into significantly more accurate vascular features when we compared features extracted from VascX segmentation masks with features extracted from segmentation masks generated by previous models. With VascX models we provide a robust, ready-to-use set of model ensembles and inference code aimed at simplifying the implementation and enhancing the quality of automated retinal vasculature analyses. The precise vessel parameters generated by the model can serve as starting points for the identification of disease patterns in and outside of the eye.

Figures (7)

• Figure 1: Statistics of the Rotterdam development sets used to train and evaluate VascX models. NA indicates the information was not available.
• Figure 2: Screen capture of the software used for all the annotations on Rotterdam CFIs.
• Figure 3: Sample of pre-processed images from the Rotterdam sets, showing: original images (first row), the results of bounds detection defined by the intersection of a circle and optional lines (second row) and the cropped and contrast-enhanced image (third row).
• Figure 4: Example model outputs from the benchmarked systems showing A/V segmentation and disc and fovea outputs were applicable. We observed more consistency in A/V and disc segmentation from VascX. The CFIs shown correspond to the images with the median Dice score for A/V segmentation with VascX; for each dataset.
• Figure 5: Box plots displaying the distribution of model performance (Dice score) over image quality and anatomic region (AR) bins/slices of the Rotterdam dataset. For vessel segmentation, the performance of the models is stable across quality and anatomic region bins. For artery-vein segmentation, the biggest improvement is in usable images. For disc segmentation, the model achieves remarkably stable performance across image quality and AR; while the images pose a challenge for Automorph.