9 Jurnal Fisika 9 (1) (2019) 9-20 Jurnal Fisika https://journal.unnes.ac.id/nju/index.php/jf/index Deteksi Awal Retinopati Hipertensi Menggunakan Jaringan Syaraf Tiruan pada Citra Fundus Mata Violetta Vincentia, Nurhasanah  , Iklas Sanubary Jurusan Fisika, Universitas Tanjungpura, Pontianak, Indonesia Info Artikel ________________ Sejarah Artikel: Diterima: 20 Maret 2019 Disetujui: 10 Juli 2019 Dipublikasikan: 11 Juli 2019 ________________ Keywords: Citra Fundus Mata, GLCM, Jaringan Syaraf Tiruan, Retinopati Hipertensi ________________ ABSTRAK ________________________________________________________________________ Identifikasi fundus mata abnormal (retinopati hipertensi) dari citra fundus mata manusia telah dilakukan menggunakan Jaringan Syaraf Tiruan (JST). Data yang digunakan berupa citra fundus mata normal dan fundus mata abnormal. Pengolahan awal citra dilakukan dengan menyeragamkan ukuran citra fundus menjadi 256 x 256 piksel. Citra fundus yang semula Red Green Blue (RGB) diubah menjadi citra grayscale. Citra diolah menggunakan perataan kontras, filter, penghapusan background, segmentasi dan masking untuk mendapat citra pembuluh darah.Citra diekstraksi dengan menghitung ciri statistik menggunakan grey level co-occurence matrix (GLCM) 4 arah yaitu 0 o , 45 o ,90 o , dan 135 o pada jarak spasial 1. Ciri statistik yang dihitung yaitu energi, kontras, korelasi, dan homogenitas sebagai input pada JST. Data dari ekstraksi ciri diidentifikasi menggunakan jaringan syaraf tiruan propagasi balik dengan arsitektur jaringan [17 7 1] dan fungsi pelatihan traingdm. Hasil dari pelatihan jaringan menunjukkan Mean Square Error (MSE) sebesar 0,00025 sementara pengujian jaringan menunjukkan nilai MSE sebesar 0,0464 dan akurasi 80%. Metode JST dapat digunakan untuk deteksi awal retinopati hipertensi. ABSTRACT _________________________________________________________________ An identification of hypertensive retinopathy had been carried out from the fundus image of the human eye using artificial neural networks (ANN). Data that were used in this research, the form of normal eye fundus and abnormal eye fundus (hypertensive retinopathy) image. The preprocessing of the image was done by standardizing the size of the fundus image to 256 x 256 pixel. Then the fundus image that was originally RGB was converted into a grayscale image. After that, the image was processed by using contrast smoothing, filtering, background removal, segmentation and masking to get an image with only blood vessels. After the preprocessing, the image was extracted by calculating statistical characteristics used 4-ways grey level co-occurrence matrix (GLCM) (0 o , 45 o , 90 o , and 135 o ) at spatial distance 1. The characteristics of statistics that were calculated energy, contrast, correlation, and homogeneity as input on ANN. Data from feature extraction were identified by using backpropagation neural networks with architecture [17 7 1] and traingdm as training functions. The results of network training showed a Mean Square Error (MSE) of 0,00025 while network testing showed an MSE value of 0,0464 and an accuracy of 80%. From the results obtained, it could be concluded that this method could be used for early detection of hypertensive retinopathy. © 2019 Universitas Negeri Semarang  Alamat korespondensi: Jurusan Fisika, Universitas Tanjungpura, Pontianak, Indonesia E-mail: nurhasanah@physics.untan.ac.id p-ISSN 2088-1509 e-ISSN 2684-978X