Week 3 Report

This past week the group completed some Literature review of an article titled, "A novel method for blood vessel detection from retinal images." Our purpose for reviewing this article was to assist our team with one of the first steps in our algorithm, blood vessel detection. In the paper, the medical team sought out to find a method that would distinguish and contrast the large and thin blood vessels in retinal images. They planned on using adaptive thresholding to create a binary image that contrasts the large and thin vessels. In their method, they used feature extraction and image processing to determine the most accurate threshold to use for the images, which included analyzing each of the images channels and HSV values. Their results proved to be 93.2% accurate when tested using their online DRIVE image database of retinal images. While there is no true ground truth for blood vessel detection, we can test our algorithm on their original images and resulting images to see how accurate our algorithm is. Also, with this algorithm we aren't subject to as much subjective evaluation because we have a scientific result to compare directly to.Then, if necessary, we can improve our algorithm to produce an even higher accuracy.