Deteksi Penyakit Retinopati Diabetik Menggunakan Model Convolutional Neural Network

Diabetic retinopathy is one of the complication caused by diabetes. The symptoms happen on the blood vessels inside the retina and can be seen through retinal fundus scan images. Early detection for diabetic retinopathy is crucial, as further stage can cause a total blindness. Thus machine learning programs were built to help the medics as it can detect any signs of blood vessels’ abnormality from the shape of the blood vessel itself. CNN algorithm is already known for excellently performing image processing such as classfication and recognition. The unique layer structures and features make it trainable to perform it’s task precisely in a short amount of time. Hence a lot of CNN models were developed as a results of it’s changeable layer structures. One of the most eminent models is DenseNet121 where the structures allows every layers in it’s blocks to be connected with each other. The number ‘121’ annotates how many layer is used within. Now using DenseNet121 the research is aimed to perform diabetic retinopathy detection with a better results than the previous researches. Afterwards one basic 3-block CNN model were built to testify the competency of DenseNet121 on reaching the said goals. The proposed system achieved highest accuracy equals to 96.32% and Cohens Kappa score of 0.872 obtained by DenseNet121 model