Hough transform-based vascular network disorder features on baseline fluorescein angiography scans predict response to anti-vegf therapy in diabetic macular edema

An apparatus comprising:

• a processor;
• a memory configured to store a digitized image of a region of interest (ROI) demonstrating diabetic macular edema (DME), the digitized image having a plurality of pixels, a pixel having an intensity;
• an input/output (I/O) interface;
• a set of circuits;
• and
• an interface that connects the processor, the memory, the I/O interface, and the set of circuits, the set of circuits comprising:
• an image acquisition circuit configured to:
• access a two-dimensional (2D) fluorescein angiography (FA) image of an ROI demonstrating DME, where the ROI includes a retinal vascular network, the image including a a plurality of pixels, a pixel having an intensity;
• a skeletonization circuit configured to:
• define an image scene based on the 2D FA image, where the image scene includes the retinal vascular network;
• generate a segmented retinal vascular network by segmenting the retinal vascular network represented in the image scene;
• a Cartesian feature circuit configured to:
• generate a Cartesian 2D vessel network representation in the Cartesian domain based on the segmented retinal vascular network;
• compute a set of localized Hough transforms based on the Cartesian 2D vessel network representation;
• generate an aggregated set of peak orientations based on the set of localized Hough transforms;
• a vascular network organization via Hough Transforms (VaNgOGH) descriptor circuit configured to:
• generate a VaNgOGH descriptor based on the aggregated set of peak orientations;
• a classification circuit configured to:
• compute a probability that the ROI is a non-rebounder, where the classification circuit computes the probability based, at least in part, on the VaNgOGH descriptor using a linear discriminant analysis (LDA) machine learning approach;
• generate a classification of the ROI as a non-rebounder or rebounder based on the probability;
• and
• a display circuit configured to display the classification.

A non-transitory computer-readable storage device storing computer-executable instructions that, in response to execution, cause a processor to perform operations for distinguishing non-rebounders from rebounders to anti-vascular endothelial growth factor (VEGF) treatment, the operations comprising:

• accessing a pre-anti-VEGF treatment two-dimensional (2D) ultra-wide field FA (UWFA) image of region of interest (ROI) demonstrating diabetic macular edema (DME) or retinal vein occlusion (RVO), where the ROI includes retinal vascular network, the UWFA image including a a plurality of pixels, a pixel having an intensity;
• defining an image scene based on the UWFA image, where the image scene includes the retinal vascular network;
• segmenting the retinal vascular network represented in the image scene;
• generating a Cartesian 2D vessel network representation in the Cartesian domain based on the segmented retinal vascular network;
• computing a set of localized Hough transforms based on the Cartesian 2D vessel network representation;
• generating an aggregated set of peak orientations based on the set of localized Hough transforms;
• generating a vascular network organization via Hough transform (VaNgOGH) descriptor based on the aggregated set of peak orientations, where the VaNgOGH descriptor quantifies a disorder of the retinal vascular network;
• providing the VaNgOGH descriptor to a linear discriminant analysis (LDA) machine learning classifier configured to compute a probability that the ROI is a non-rebounder;
• receiving, from the machine learning classifier, a probability that the ROI is a non-rebounder, where the machine learning classifier computes the probability based, at least in part, on the VaNgOGH descriptor;
• generating a classification of the ROI as a non-rebounder or rebounder based on the probability;
• and
• displaying the classification.