<![CDATA[The vascular changes that occur in retinal are precursors of a diseases, such as heart disease, diabetic retinopathy, stroke and hypertension. Changes can be seen by analyzing retinal image, but it takes a long time. In this study we propose the automation of vascular segmentation processes in retinal image so that can assist in analysis process, which is an important step in retinal image analysis. The segmentation process is done by detecting the line using the Multi-Scale Line Operator algorithm and preprocessing image using K-Means algorithm. Line detection is performed on several different scales, then combines the results of each scale. Image preprocessing using the K-Means algorithm aims to ignore the optic disc area, which in that area will probably be detected as false positive. The performance of proposed algorithm was evaluated using the DRIVE and STARE dataset, the result showed that average accuracy of the DRIVE dataset reaches 0,940980219 with AUC 0,7462, and for STARE dataset reaches 0,949293361 with AUC 0,778. The results are obtained by using the number of K as much as 3 on the K-Means algorithm, which consists of background, foreground, and vessel.]]>
Copyrights © 2018
