LAUSR

Computer-aided diagnosis of hypertensive retinopathy (CAD-HR) is performed by analyzing the retinal image. The analysis is carried out in several stages, one of which is image segmentation. The segmentation carried out so far generally uses a region-based and threshold-based approach. There is not yet a clustering-based approach, and there has been no previous analysis of why clustering-based is not yet widely used. This study aims to conduct clustering-based Segmentation analysis, specifically k-means clustering in CAD-HR. The research method used is divided into four stages, namely preprocessing, segmentation, feature extraction using fractal dimensions, statistical analysis for classification, and classification. Testing is done using the DRIVE and STARE datasets. The results of statistical tests showed that the number of clusters 3 was able to provide a significant difference between the fractal positive and negative dimensions of hypertensive retinopathy. The model of CAD-RH using the k-means algorithm for segmentation method is able to provide 80% sensitivity performance. The k-mean algorithm can be used as an alternative to segmenting retinal blood vessels.