US 12,254,674 B2
Method for recognizing arteries and veins on a fundus image using hyperspectral imaging technique
Hsiang-Chen Wang, Minsyong Township, Chiayi County (TW); Yu-Ming Tsao, Minsyong Township, Chiayi County (TW); Yong-Song Chen, Minsyong Township, Chiayi County (TW); Yu-Sin Liu, Minsyong Township, Chiayi County (TW); and Shih-Wun Liang, Minsyong Township, Chiayi County (TW)
Assigned to National Chung Cheng University, Minsyong Township (TW)
Filed by National Chung Cheng University, Minsyong Township, Chiayi County (TW)
Filed on Jul. 22, 2022, as Appl. No. 17/871,195.
Claims priority of application No. 111110879 (TW), filed on Mar. 23, 2022.
Prior Publication US 2023/0306720 A1, Sep. 28, 2023
Int. Cl. G06V 10/772 (2022.01); G06T 7/00 (2017.01); G06V 40/18 (2022.01)
CPC G06V 10/772 (2022.01) [G06T 7/0012 (2013.01); G06V 40/193 (2022.01); G06T 2207/30041 (2013.01); G06T 2207/30101 (2013.01)] 8 Claims
OG exemplary drawing
 
1. A method for recognizing arteries and veins on a fundus image, the method being implemented using a computer device executing a software application program and comprising steps of:
a) executing a pre-process operation on the fundus image, so as to obtain a pre-processed fundus image;
b) generating a fundus spectral reflection dataset associated with a plurality of pixels of the pre-processed fundus image, based on the pre-processed fundus image, and a spectral transformation matrix;
c) performing a principal components analysis (PCA) operation on the fundus reflection spectral dataset, so as to obtain a plurality of principle component scores associated with the pixels of the pre-processed fundus image, respectively; and
d) determining, for each of the pixels of the pre-processed fundus image that has been determined as a part of a blood vessel, whether the pixel belongs to a part of an artery or a part of a vein;
wherein step d) includes:
determining a threshold principle component score for recognizing the pixels as a part of an artery or a part of a vein;
comparing each of the plurality of principle component scores with the threshold principle component score; and
recognizing the pixels of the pre-processed fundus image as a part of an artery or a part of a vein using a relation of an associated principle component score with the threshold principle component score;
wherein the determining of the threshold principle component score includes
calculating an average of the plurality of principle component scores,
classifying the plurality of principle component scores into one of a first group and a second group, using the average of the plurality of principle component scores as an initial threshold;
calculating, for the principle component scores classified into the first group, a first average principle component score, and for the principle component scores classified into the second group, a second average principle component score,
calculating a current threshold based on the first average principle component score and the second average principle component score,
determining whether a difference between the current threshold and the initial threshold is less than a preset value,
when it is determined that the difference is less than the preset value, setting the current threshold as the threshold principle component score, and
when it is determined that the difference is not less than the preset value, setting the current threshold as the initial threshold, and repeating the classifying the plurality of principle component scores.