Automated determination of arteriovenous ratio in images of blood vessels
Inventors
Abramoff, Michael D. • Niemeijer, Meindert • Xu, Xiayu • Sonka, Milan • Reinhardt, Joseph M.
Assignees
University of Iowa Research Foundation UIRF • US Department of Veterans Affairs
Publication Number
US-11638522-B2
Publication Date
2023-05-02
Expiration Date
2032-01-20
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Abstract
The methods and systems provided can automatically determine an Arteriolar-to-Venular diameter Ratio, AVR, in blood vessels, such as retinal blood vessels and other blood vessels in vertebrates. The AVR is an important predictor of increases in the risk for stroke, cerebral atrophy, cognitive decline, and myocardial infarct.
Core Innovation
The invention provides methods and systems that automatically determine an Arteriolar-to-Venular diameter Ratio (AVR) in blood vessels, such as retinal blood vessels and other vertebrate blood vessels. The AVR serves as an important predictor of increased risk for stroke, cerebral atrophy, cognitive decline, and myocardial infarct.
The problem addressed is the manual determination of AVR from retinal color fundus images, which is time-consuming and requires an expert. Clinicians typically make only gross estimates and cannot determine numeric ratios. The invention solves this by introducing an automated system that detects the optic disc, defines an appropriate region of interest, classifies vessels as arteries or veins, measures vessel widths, and calculates the AVR, thereby facilitating risk analysis for cardiovascular events and related conditions.
The methods integrate preprocessing to remove image gradients and background variation, vessel segmentation producing a likelihood vessel map, skeletonization to extract vessel centerlines, and supervised classification to assign artery or vein labels to vessel segments. Vessel width is measured perpendicular to vessel angles using techniques such as tobogganing and graph-based approaches for boundary delineation. Iterative matching pairs arteries and veins to compute the AVR. The automatic method aligns with defined clinical protocols and provides assessment in a region centered on but excluding the optic disc.
Claims Coverage
The patent claims encompass inventive features related to automatic determination of arteriovenous ratio (AVR) from images through region of interest determination, vessel classification, vessel width measurement, and AVR estimation using advanced processing techniques and classifiers.
Automatic region of interest determination excluding optic disc
The method includes determining a region of interest (ROI) in an image centered around but excluding the optic disc, comprising multiple centerline pixels, and using concentric circles derived from the optic disc diameter to define the ROI.
Vessel pixel classification using normalized color features and supervised classifier
Extraction and normalization of color features from vessel centerline pixels, followed by labeling each pixel as artery or vein using a trained classifier applying derivatives, texture, and color properties.
Voting procedure across multiple ROI diameters for artery and vein identification
A voting algorithm is performed multiple times for ROIs of varying diameters to assign artery or vein labels consistently to vessel segments in the ROI based on classifier outputs.
Vessel width measurement using graph search and multiscale pixel feature tobogganing methods
Determining vessel widths for identified arteries and veins using methods including graph search with multiscale cost functions derived from wavelet kernel lifting (Gabor, Gaussian derivative, Difference of Gaussians) and multispectral pixel-based tobogganing segmentation and splats.
Iterative matching and calculation of AVR from vessel widths
AVR estimation is accomplished by matching and using widths of arteries and veins in the ROI iteratively to calculate central retinal artery and vein equivalents, forming the ratio.
Together, these inventive features provide a fully automated system and method for accurately identifying arteries and veins, measuring vessel widths, and calculating AVR from images excluding the optic disc area, enabling disease risk assessment.
Stated Advantages
Automation significantly reduces time and expert involvement compared to manual AVR measurement, making quantitative assessment feasible.
The methods allow accurate and repeatable artery and vein classification and vessel width measurement, providing numerical AVR values rather than gross estimates.
The system can be applied to various tissue images beyond the retina, including iris, skin, brain surface, and other tissues visible via two or more wavelength imaging.
Graph-based vessel boundary detection provides improved accuracy and robustness over prior methods, enabling analysis even on small, low contrast vessels.
The invention enables batch processing of large image quantities for group-level disease propensity determination.
Documented Applications
Risk analysis and prediction of cardiovascular events, stroke, cerebral atrophy, cognitive decline, and myocardial infarct using retinal and other tissue blood vessel imaging.
Application in medical fields such as Neurology, Primary Care, Ophthalmology for disease risk assessment.
Analysis of vascular properties in retina, iris, skin, eardrum, brain surface, and other suitable vertebrate tissues using multi-wavelength imaging.
Use in diabetic retinopathy screening and retinopathy of prematurity (Plus disease) diagnosis via vessel tortuosity and AVR assessment.
Batch processing of images for disease propensity estimation in groups of patients.
Interested in licensing this patent?