# US 12,223,646 B2 · Automated selection of ultra-widefield angiography images > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 12,223,646 B2 - Document ID: US12223646B2 - Status: Granted - Kind: B2 - Application number: 17/734,895 - Filed: 2022-05-02 - Published: 2022-11-03 - Granted: 2025-02-11 - Earliest claimed: 2021-04-30 - Assignees: The Cleveland Clinic Foundation - Applicants: The Cleveland Clinic Foundation - Inventors: Justis P. Ehlers; Duriye Damla Sevgi; Sunil K. Srivastava - Web page: https://patentgrep.com/patent/US12223646B2 ## CPC - G06T 7/0012: {Biomedical image inspection} - G06T 7/11: Region-based segmentation - G06T 2207/20081: Training; Learning - G06T 2207/20084: Artificial neural networks [ANN] - G06T 2207/30041: Eye; Retina; Ophthalmic - G06T 2207/30101: Blood vessel; Artery; Vein; Vascular - G06T 2207/30168: Image quality inspection - G06T: IMAGE DATA PROCESSING OR GENERATION, IN GENERAL ## Abstract Systems and methods are provided for provided for automated selection of UWFA images. A first set of images representing an early phase of a UWFA image acquisition and a second set of images representing a late phase of the UWFA image acquisition are received and segmented to provide a vascular structure map for each of the first set of images and the second set of images. An image quality metric is assigned to each of the first set of images and the second set of images from the vascular structure map associated with each image. An image of the first set of images having an extreme value for the image quality metric across the first set of images and an image of the second set of images having an extreme value for the image quality metric across the second set of images are selected. ## Claims ### Claim 1 A system for automated selection of ultra-widefield angiographic (UWFA) images, the system comprising: a processor; and a non-transitory computer readable medium storing executable instructions, the executable instructions comprising: an imager interface that receives a first set of images representing an early phase of a UWFA image acquisition and a second set of images representing a late phase of the UWFA image acquisition; a segmentation component that segments each of the first set of images and the second set of images to provide a vascular structure map for each of the first set of images and the second set of images; an image evaluation component that assigns an image quality metric to each of the first set of images and the second set of images from the vascular structure map associated with each image; and an image selector that selects an image of the first set of images having an extreme value for the image quality metric across the first set of images and an image of the second set of images having an extreme value for the image quality metric across the second set of images. ### Claim 2 (depends on 1) The system of claim 1, wherein the segmentation component comprises a a convolutional neural network trained on a set of UWFA images segmented by a human expert. ### Claim 3 (depends on 2) The system of claim 2, wherein the convolutional neural network uses a twelve-by-twelve convolutional kernel. ### Claim 4 (depends on 1) The system of claim 1, wherein the image selector assigns a first image quality metric to each of the first set of images and assigns a second image quality metric to each of the second set of images, wherein the first image quality metric is calculated differently from the second image quality metric such that a value of the first image quality metric for a given image is different from a value of the second image quality metric for the given image. ### Claim 5 (depends on 1) The system of claim 1, wherein the image selector assigns the image quality metric to each of the first set of images according to a total area of the vascular structure map for the image. ### Claim 6 (depends on 1) The system of claim 1, wherein the image selector comprises a predictive model, trained on a plurality of annotated vascular structure maps, that assigns the image quality metric to each of the first set of images based on the vascular structure maps associated with the first set of images. ### Claim 7 (depends on 6) The system of claim 6, wherein the predictive model assigns the image quality metric to each of the first set of images and the second set of images. ### Claim 8 (depends on 6) The system of claim 6, wherein the image selector comprises a predictive model, trained on a set of feature values extracted from a plurality of annotated vascular structure maps, that assigns the image quality metric to each of the first set of images on a corresponding set of features extracted from the vascular structure maps associated with the first set of images. ### Claim 9 (depends on 1) The system of claim 1, wherein the image selector assigns an image quality metric to each of the second set of images according to a similarity of the vascular structure map to a representative vascular structure map. ### Claim 10 (depends on 9) The system of claim 9, wherein the given image is an image of the second set of images, and the representative vascular structure map is the vascular structure map associated with the image of the first set of images having the extreme value. ### Claim 11 A method for automated selection of ultra-widefield angiographic (UWFA) images comprising: receiving a first set of images representing an early phase of a UWFA image acquisition and a second set of images representing a late phase of the UWFA image acquisition; segmenting each of the first set of images and the second set of images to provide a vascular structure map for each of the first set of images and the second set of images; assigning an image quality metric to each of the first set of images and the second set of images from the vascular structure map associated with each image; selecting an image of the first set of images having an extreme value for the image quality metric across the first set of images; and selecting an image of the second set of images having an extreme value for the image quality metric across the second set of images. ### Claim 12 (depends on 11) The method of claim 11, wherein segmenting each of the first set of images and the second set of images comprises providing each of the first set of images and the second set of images to a convolutional neural network trained on a set of UWFA images segmented by a human expert. ### Claim 13 (depends on 11) The method of claim 11, wherein assigning the image quality metric to each of the first set of images and the second set of images comprises: assigning a first image quality metric to each of the first set of images; and assigning a second image quality metric to each of the second set of images, wherein the first image quality metric is calculated differently from the second image quality metric such that a value of the first image quality metric for a given image is different from a value of the second image quality metric for the given image. ### Claim 14 (depends on 13) The method of claim 13, wherein assigning the image quality metric to each of the first set of images and the second set of images from the vascular structure map associated with each image comprises assigning the image quality metric for each image of the first set of images according to a total area of the vascular structure map for the image. ### Claim 15 (depends on 14) The method of claim 14, wherein selecting the image of the first set of images having the extreme value for the image quality metric across the first set of images comprises selectin the image of the first set of images having the maximum value of the total area of the vascular structure map. ### Claim 16 (depends on 14) The method of claim 14, wherein assigning an image quality metric to each of the first set of images and the second set of images from the vascular structure map associated with each image comprises assigning the image quality metric for each image of the second set of images according to a similarity of the vascular structure map to a representative vascular structure map. ### Claim 17 (depends on 16) The method of claim 16, wherein the representative vascular structure map is the vascular structure map associated with the image of the first set of images having the extreme value. ### Claim 18 A method for automated selection of ultra-widefield angiographic (UWFA) images comprising: receiving a first set of images representing an early phase of a UWFA image acquisition and a second set of images representing a late phase of the UWFA image acquisition; segmenting each of the first set of images and the second set of images to provide a vascular structure map for each of the first set of images and the second set of images; assigning a first image quality metric to each of the first set of images from the vascular structure map associated with each image; assigning a second image quality metric to each of the second set of images from the vascular structure map associated with each image, wherein the first image quality metric is calculated differently from the second image quality metric such that a value of the first image quality metric for a given image is different from a value of the second image quality metric for the given image; selecting an image of the first set of images having an extreme value for the first image quality metric across the first set of images; and selecting an image of the second set of images having an extreme value for the second image quality metric across the second set of images. ### Claim 19 (depends on 18) The method of claim 18, wherein assigning the first image quality metric to each of the first set of images comprises assigning the first image quality metric for each image of the first set of images according to a total area of the vascular structure map for the image. ### Claim 20 (depends on 18) The method of claim 18, wherein assigning a second image quality metric to each of the second set of images from the vascular structure map associated with each image comprises assigning the second image quality metric for each image of the second set of images according to a similarity of the vascular structure map to a representative vascular structure map.