Optic Nerve Head Segmentation Using Hough Transform and Active Contours

Handayani Tjandrasa, Ari Wijayanti, Nanik Suciati
2012 TELKOMNIKA (Telecommunication Computing Electronics and Control)  
Abstrak Optic nerve head merupakan bagian retina tempat sel ganglion axon keluar dari mata untuk membentuk optic nerve. Hilangnya fiber saraf akibat glaucoma menurunkan ukuran optic disk dan melebarkan ukuran cup. Karenanya evaluasi optic nerve head adalah penting untuk diagnosis dini glaucoma. Studi ini mengimplementasikan deteksi optic nerve head pada citra fundus retina berdasarkan Hough Transform dan Active Contour Model. Proses dimulai dengan perbaikan citra menggunakan filter homomorphic
more » ... ntuk koreksi iluminasi, kemudian dilanjutkan dengan penghapusan pembuluh darah untuk memfasilitasi proses segmentasi berikutnya. Hasil lingkaran transformasi Hough menjadi level set awal untuk active contour model. Hasil uji coba menunjukkan bahwa algoritma segmentasi mampu mendeteksi optic nerve head dengan akurasi rata-rata sebesar 75.56% dengan menggunakan 30 citra retina dari DRIVE database. Abstract Optic nerve head is part of the retina where ganglion cell axons exit the eye to form the optic nerve. Glaucomatous changes related to loss of the nerve fibers decrease the neuroretinal rim and expand the area and volume of the cup. Therefore optic nerve head evaluation is important for early diagnosis of glaucoma. This study implements the detection of the optic nerve head in retinal fundus images based on the Hough Transform and Active Contour Models. The process starts with the image enhancement using homomorphic filtering for illumination correction, then proceeds with the removal of blood vessels on the image to facilitate the subsequent segmentation process. The result of the Hough Transform fitting circle becomes the initial level set for the active contour model. The experimental results show that the implemented segmentation algorithms are capable of segmenting optic nerve head with the average accuracy of 75.56% using 30 retinal images from the DRIVE database.
doi:10.12928/telkomnika.v10i3.833 fatcat:amcf7dvxyfbkfbbtidameibgye