Klasifikasi Diabetic Retinopathy Menggunakan Metode Random Forest dengan Ekstraksi Fitur LBP dan GLCM

Abstract

Diabetic retinopathy is a disease that occurs due to damage to blood vessels in the retina. This disease is one of the complications of diabetes mellitus, which can lead to blindness. The detection of this disease is typically performed by experts through the observation of funduscopy examination results, which often takes a considerable amount of time and carries a high risk of errors. Researchers have developed computer vision techniques to detect diabetic retinopathy through fundus retina images. The LBP method is used to extract texture from the images, and the resulting LBP images are further processed using the GLCM method to extract 16 features for classification using the Random Forest method. To determine the best approach, various experiments are conducted involving the number of decision Trees in the random forest, data augmentation, and comparing the classification results using only LBP features, only GLCM features, and the combined LBP and GLCM features. The determination of the best model is based on the accuracy, precision, and recall values obtained. From the conducted experiments, the classification of images without data augmentation shows suboptimal results compared to the other approaches. Meanwhile, the highest classification results obtained using the combined LBP and GLCM feature extraction, LBP only, and GLCM only are 85.3%, 80.3%, and 90%, respectively, with different numbers of decision Trees. Thus, it can be concluded that increasing the number of decision Trees does not necessarily guarantee better classification results.