Deteksi Penyakit Retinopati Diabetik Dengan Metode 3D Gray-Levelco-Occurrence Matrix Dan Jaringan Saraf Tiruanpropagasi Balik
Total View This Week0
Total View This Week0
Institusion
Institut Teknologi Telkom Purwokerto
Author
Asep, Karyana
Subject
T Technology (General)
Datestamp
2021-09-28 04:30:41
Abstract :
Diabetic Retinopathy (DR) is a disease that attacks a person's eyes caused by Diabetes Mellitus (DM). DR disease analysis was carried out through eye fundus images by doctors. However, often poor image quality makes it difficult for doctors during the analysis process. Therefore, in this study, a system for detecting DR in eye fundus images is developed. This detection system is made by utilizing 3D Gray-Level Co-occurrence Matrix (3D GLCM) technology as a feature extraction method andArtificial Neural Network (ANN) Backpropagation as a classification method. In addition, hold-out validation is also used as a method of validating classification results and confusion matrix to see the level of accuracy of the system being created. The image used is 2949 pieces obtained from the MESSIDOR website with details of 2065 training images (70%) and test images as many as 884 pieces (30%). Tests were carried out on 13 3D GLCMoffset directions with 10, 50, and 100 hidden neuron variations. The results showed that the highest accuracy rate was 65.9% which occurred in the direction of the 1st offset ([0 1 0]) with 100 hidden neurons.The results of this accuracy are good enough but can still be improved. Keywords:Diabetic Retinopathy,Artificial Neural Network, 3D GLCM, confusion matrix, hold-out validation
Diabetic Retinopathy (DR) is a disease that attacks a person's eyes caused by Diabetes Mellitus (DM). DR disease analysis was carried out through eye fundus images by doctors. However, often poor image quality makes it difficult for doctors during the analysis process. Therefore, in this study, a system for detecting DR in eye fundus images is developed. This detection system is made by utilizing 3D Gray-Level Co-occurrence Matrix (3D GLCM) technology as a feature extraction method andArtificial Neural Network (ANN) Backpropagation as a classification method. In addition, hold-out validation is also used as a method of validating classification results and confusion matrix to see the level of accuracy of the system being created. The image used is 2949 pieces obtained from the MESSIDOR website with details of 2065 training images (70%) and test images as many as 884 pieces (30%). Tests were carried out on 13 3D GLCMoffset directions with 10, 50, and 100 hidden neuron variations. The results showed that the highest accuracy rate was 65.9% which occurred in the direction of the 1st offset ([0 1 0]) with 100 hidden neurons.The results of this accuracy are good enough but can still be improved. Keywords:Diabetic Retinopathy,Artificial Neural Network, 3D GLCM, confusion matrix, hold-out validation
Download
book
BibTex
Latex, Jabref
cloud_download
Original Resource
url resource Institution
