Nur, Nahya (2018) Deteksi Region dengan Completed Local Binary Pattern dan Color Feature untuk Segmentasi Exudate menggunakan Metode Saliency pada Retina Fundus Images. Masters thesis, Institut Teknologi Sepuluh Nopember.
Preview |
Text
5116201035-Master-Thesis.pdf - Accepted Version Download (2MB) | Preview |
Abstract
Salah satu pertanda awal dari penyakit tersebut adalah munculnya luka berupa exudates yang terjadi karena terdapat lipid atau lemak bocor pada pembuluh darah abnormal dan bisa menyebabkan kebutaan bila berada di sekitar daerah macula. Deteksi dini kemunculan exudates diharapkan mampu untuk menurunkan resiko kebutaan terhadap penderita penyakit diabetic retinopathy.
Salah satu tantangan dalam proses pendeteksian exudates adalah ukuran dari objek tersebut yang cukup kecil dibandingkan keseluruhan image. Dalam penelitian ini mengusulkan pendeteksian region atau daerah exudates menggunakan CLBP dan color feature untuk segmentasi exudates dengan saliency method. Terdapat tiga tahapan utama dalam penelitian ini, yaitu penghapusan optic disk, pendeteksian lokasi exudates, dan segmentasi exudates. Penghapusan optic disk dilakukan dengan menggunakan algoritma midpoint circle. Pada tahapan pendeteksian lokasi exudates, image akan dibagi menjadi sub-sub image yang lebih kecil kemudian diklasifikasikan menjadi exudates patch dan exudate-free patch berdasarkan fitur yang diperoleh dengan CLBP dan color feature. Sub image yang diklasifikasikan sebagai exudates patch kemudian disegmentasi dengan menggunakan saliency method dan renyi entropi thresholding.
Evaluasi metode dilakukan pada dataset diaretDB1 dengan menghitung nilai akurasi, sensitivity, dan specificity. Metode yang diajukan dapat mendeteksi exudates secara lebih akurat dengan rata-rata nilai akurasi 99.63 %, sensitivity 83.23%, dan specificity 99.57%.
==============================================================================================Diabetic retinopathy is a complication of diabetes that attacks the eye organs. One of the early signs of the disease is the appearance of an exudate wound that occurs because there is lipid or leaked fatty in the abnormal blood vessels and can cause blindness if it appears around the macula. Early detection of exudates is expected to reduce the risk of blindness to diabetic retinopathy patients.
One of the problems in the detection process of exudates is that the size of the object is quite small compared to the overall image. This study proposes to detect the exudate region using CLBP and color feature for segmentation exudates with saliency method. There are three main stages in this research, such as optical disk removal, location detection of exudates, and exudates segmentation. Optical disc removal is done by using midpoint circle algorithm. At the detection stage of the exudates location, the image will be divided into blocks then classified it into exudate patch and exudate-free patch based on features that obtained using CLBP and color features. Sub images that are classified as exudate patch then segmented by using the saliency method and renyi entropy thresholding.
The method evaluation is performed on the diaretDB1 dataset by calculating the accuracy, sensitivity, and specificity values. The proposed method can detect exudates more accurately as shown with the average of accuracy, sensitivity, and specificity value of 99.63%, 83.23% , and 99.57% respectively.
| Item Type: | Thesis (Masters) |
|---|---|
| Uncontrolled Keywords: | diabetes retinopati, citra fundus, exudates, CLBP, saliency method, segmentasi. |
| Subjects: | T Technology > TA Engineering (General). Civil engineering (General) > TA1637 Image processing--Digital techniques T Technology > TA Engineering (General). Civil engineering (General) > TA174 Computer-aided design. |
| Divisions: | Faculty of Information and Communication Technology > Informatics > 55101-(S2) Master Thesis |
| Depositing User: | Nahya Nur |
| Date Deposited: | 21 Jul 2021 22:48 |
| Last Modified: | 21 Jul 2021 22:48 |
| URI: | http://repository.its.ac.id/id/eprint/57508 |
Actions (login required)
 |
View Item |