Abstract:
Diabetes is a chronic disease characterised by disorders in the production or use of insulin in the body. Insulin is a hormone produced by pancreatic beta cells, essential in regulating blood glucose levels. One way to detect diabetes is to use iridology. Iridology is a field of study that can detect bodily abnormalities through the eye’s iris. Iridologists analyse the characteristics of the iris using a biomicroscope. This causes the results of analysis and diagnosis of disease to be subjective and take a long time. This research aims to build a program that is capable of detecting diabetes from iris images using k-Nearest Neighbor (k-NN) by comparing two texture feature extraction methods, namely Gray Level Co-Occurance Matrix (GLCM) and Gray Level Run Length Matrix (GLRLM). For GLCM, the features used are contrast, dissimilarity, homogeneity, energy, and correlation, while for GLRLM, the features used are short-run emphasis (SRE), long-run emphasis (LRE), gray-level non-uniformity (GLN), run percentage (RP), and run length non-uniformity (RLN). These features are then combined as input for the classification process of diabetic and normal patients using k-NN. The data used in this study were 140 eye images consisting of 70 eye images of diabetes patients and 70 normal eye images. The results showed that the accuracy of diabetes detection using GLCM feature extraction was better than GLRLM. The highest accuracy result with GLCM is 80.95%, while with GLRLM, it is 71.43%.
Authors: Farah Restuadji; Riries Rulaningtyas; Endah Purwanti; Soegianto Soelistiono; Fitriyatul Qulub
