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-12035971-B2
Publication Date
2024-07-16
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 for automatic determination of the arteriolar-to-venular diameter ratio (AVR) in blood vessels, including retinal blood vessels and others in vertebrates. This automatic determination is important because the AVR serves as a predictor for the risk of stroke, cerebral atrophy, cognitive decline, myocardial infarct, and other cardiovascular and brain diseases. The disclosed methods and systems perform this automatic measurement by detecting the optic disc, selecting a region of interest (ROI), segmenting vessels, classifying vessels as arteries or veins, measuring vessel widths, and calculating the AVR.
The problem addressed by the invention is that until now, AVR has been determined manually from retinal fundus images, which is time-consuming and requires expert evaluation. Clinicians are only capable of making gross estimates for substantially abnormal arteriovenous ratios, not precise numeric determinations. This manual process impedes efficient and accurate risk analysis for cardiovascular and neurological conditions. The invention solves this by automating the entire process, enabling rapid, objective, and reproducible measurement of AVR.
The methods utilize image preprocessing techniques such as field of view mirroring and background removal, vessel segmentation using supervised classifiers, skeletonization to extract vessel centerlines, and vessel width measurement employing both tobogganing methods to identify homogeneous regions ('splats') and graph-based boundary delineation with multiscale cost functions. Further, artery and vein classification is accomplished using local color and texture features normalized per image and a voting procedure based on vessel pairing around the optic disc ROI. The AVR calculation refines width measurements of matched artery-vein pairs iteratively, providing a robust quantitative metric.
By integrating these steps into an automated protocol, the invention offers an objective, accurate determination of AVR from retinal images and other tissue images with visible vasculature using two-wavelength imaging, enabling widespread clinical use and potential applications in medicine, neurology, ophthalmology, and other fields.
Claims Coverage
The patent contains three independent claims describing a method, a system, and a non-transitory computer-readable medium for automatic determination of the arteriovenous ratio (AVR) in tissue images. Six main inventive features are disclosed that cover image processing, classification, measurement, and calculation steps for AVR estimation.
Automatic identification of vessels in a region of interest with voting procedure
The method receives an image, determines an ROI composed of centerline pixels, assigns artery or vein labels to each pixel via a trained classifier, and applies a voting procedure repeated over multiple diameters to identify arteries and veins accurately in the ROI.
Vessel segmentation augmented by trained classification
Performing vessel segmentation on the image followed by classifying vessel pixels or segments using a trained classifier, possibly enhanced by vessel tree analysis or blood flow information to improve artery/vein identification.
Vessel width measurement using graph search and tobogganing
Determining vessel width measurements by using techniques such as graph search with multiscale cost functions derived from wavelet kernel lifting, multiscale pixel feature-based tobogganing methods to assign vesselness likelihoods to homogeneous splats, or profile fitting.
Region of interest defined by optic disc diameter
The ROI for measurement is defined as an annular region encapsulated between two concentric circles, where the diameters relate to the diameter of the optic disc detected in the image.
Applicability to diverse image types and tissue locations
The method and system can process various image types including color, multispectral, or Optical Coherence Tomography images, and images depicting tissues such as retina, iris, skin, brain surface, or other tissue with visible blood vessels.
AVR estimation correlating vessel width measurements in identified arteries and veins
Estimating the AVR by calculating ratios of vessel width measurements obtained from the identified arteries and veins in the ROI after classification and measurement steps.
The claims cover an integrated system and method for automated AVR determination based on vessel segmentation, classification, precise width measurement using advanced image analysis techniques, and robust identification of arteries and veins in a defined ROI relative to the optic disc, applicable across various image modalities and tissue types.
Stated Advantages
Provides a rapid, objective, and reproducible assessment of arteriolar-to-venular diameter ratio, improving clinical risk analysis for cardiovascular and brain diseases.
Eliminates the need for time-consuming manual AVR measurements by experts.
Enables detailed artery/vein classification and accurate vessel width measurement, even in small vessels and low contrast areas.
Applicable to various tissue types and imaging modalities, expanding potential clinical and research uses.
Supports batch processing of large image sets for population-level disease propensity assessment.
Documented Applications
Cardiovascular event risk analysis in patients undergoing fundus imaging, including diabetic retinopathy screening.
Neurological risk analysis for stroke, cerebral atrophy, and cognitive decline prediction using retinal imaging.
Assessment of Plus disease in retinopathy of prematurity through vascular tortuosity and AVR measurement.
Determination of properties of blood vessels in other body parts suitable for multi-wavelength imaging, such as iris, skin, and brain surface.
Large-scale automated analysis for disease propensity in groups of persons.
Interested in licensing this patent?