Segmentasi Pembuluh Darah Retina Pada Citra Fundus Menggunakan Gradient Based Adaptive Thresholding Dan Region Growing

Deni Sutaji
Chastine Fatichah
Dini Adni Navastara

Abstract


 

Segmentasi pembuluh darah pada citra fundus retina menjadi hal yang substansial dalam dunia kedokteran, karena dapat digunakan untuk mendeteksi penyakit, seperti: diabetic retinopathy, hypertension, dan cardiovascular. Dokter membutuhkan waktu sekitar dua jam untuk mendeteksi pembuluh darah retina, sehingga diperlukan metode yang dapat membantu screening agar lebih cepat.

Penelitian sebelumnya mampu melakukan segmentasi pembuluh darah yang sensitif terhadap variasi ukuran lebar pembuluh darah namun masih terjadi over-segmentasi pada area patologi. Oleh karena itu, penelitian ini bertujuan untuk mengembangkan metode segmentasi pembuluh darah pada citra fundus retina yang dapat mengurangi over-segmentasi pada area patologi menggunakan Gradient Based Adaptive Thresholding dan Region Growing.

Metode yang diusulkan terdiri dari 3 tahap, yaitu segmentasi pembuluh darah utama, deteksi area patologi dan segmentasi pembuluh darah tipis. Tahap segmentasi pembuluh darah utama menggunakan high-pass filtering dan tophat reconstruction pada kanal hijau citra yang sudah diperbaiki kontrasnya sehingga lebih jelas perbedaan antara pembuluh darah dan background. Tahap deteksi area patologi menggunakan metode Gradient Based Adaptive Thresholding. Tahap segmentasi pembuluh darah tipis menggunakan Region Growing berdasarkan informasi label pembuluh darah utama dan label area patologi. Hasil segmentasi pembuluh darah utama dan pembuluh darah tipis kemudian digabungkan sehingga menjadi keluaran sistem berupa citra biner pembuluh darah. Berdasarkan hasil uji coba, metode ini mampu melakukan segmentasi pembuluh darah retina dengan baik pada citra fundus DRIVE, yaitu dengan akurasi rata-rata 95.25% dan nilai Area Under Curve (AUC) pada kurva Relative Operating Characteristic (ROC) sebesar 74.28%.                          

Kata Kunci: citra fundus retina, gradient based adaptive thresholding, patologi, pembuluh darah retina, region growing, segmentasi.

 

 

Segmentation of blood vessels in the retina fundus image becomes substantial in the medical, because it can be used to detect diseases, such as diabetic retinopathy, hypertension, and cardiovascular. Doctor takes about two hours to detect the blood vessels of the retina, so screening methods are needed to make it faster.

The previous methods are able to segment the blood vessels that are sensitive to variations in the size of the width of blood vessels, but there is over-segmentation in the area of pathology. Therefore, this study aims to develop a segmentation method of blood vessels in retinal fundus images which can reduce over-segmentation in the area of pathology using Gradient Based Adaptive Thresholding and Region Growing.

The proposed method consists of three stages, namely the segmentation of the main blood vessels, detection area of pathology and segmentation thin blood vessels. Main blood vessels segmentation using high-pass filtering and tophat reconstruction on the green channel which adjusted of contras image that results the clearly between object and background. Detection area of pathology using Gradient Based Adaptive thresholding method. Thin blood vessels segmentation using Region Growing based on the information main blood vessel segmentation and detection of pathology area. Output of the main blood vessel segmentation and thin blood vessels are then combined to reconstruct an image of the blood vessels as output system.This method is able to segment the blood vessels in retinal fundus images DRIVE with an accuracy of 95.25% and the value of Area Under Curve (AUC) in the relative operating characteristic curve (ROC) of 74.28%.

Keywords: Blood vessel, fundus retina image, gradient based adaptive thresholding, pathology, region growing, segmentation.


References


Fraz, M. M., Remagnino, P., Hoppe, A., Uyyanonvara, B., Rudnicka, A., Owen, C., & Barman, S. (2012). Blood vessel segmentation methodologies in retinal images – A survey. Computer Methods and Programs in Biomedicine, 108(1), 407–433.

Hakim, L. (2014). Segmentasi Pembuluh Darah Retina Pada Citra Fundus Menggunakan Gaussian Pyramid dan Phase Congruency. Explore IT.

Ilyas, S. (2004). Sari Ilmu Penyakit Mata Edisi II. Jakarta: Fakultas Kedokteran Universitas Indonesia.

ISI. (2004). DRIVE: Digital Retinal Images for Vessel Extraction. Retrieved from Image Sciences Institute: http://www.isi.uu.nl/Research/Databases/DRIVE/

Roychowdhury, S., Koozekanani, D. D., & Parhi, K. K. (2014). Blood Vessel Segmentation of Fundus Images by Major Vessel Extraction and Subimage Classification. IEEE Journal of Biomedical and Health Informatics, 19(3), 1118-1128.

Sabilla, W. I., Soelaiman, R., & Fatichah, C. (2015). Automatic Detection of Proliferative Diabetic Retinopathy with Hybrid Feature Extraction Based on Scale Space Analysis and Tracking. The 1st International Seminar on Science and Technology (hal. 95-96). Surabaya: Postgraduate Program Institut Teknologi Sepuluh Nopember.

Welfer, D., Scharcanski, J., & Marinho, D. R. (2010). A Coarse-to-Fine Strategy for Automatically Detecting Exudates in Color Eye Fundus Images. Computerized Medical Imaging and Graphics, 34(3), 228-235.

Yazid, H., & Arof, H. (2013). Gradient based adaptive thresholding. Journal of Visual Communication and Image Representation, 24(7), 926–936.

Zhao, Y. Q., Wang, X. H., Wang, X. F., & Shih, F. Y. (2014). Retinal vessels segmentation based on level set and region growing. Pattern Recognition, 47(7), 2437–2446.




DOI: https://doi.org/10.26594/register.v2i2.553

Article metrics

Abstract views : 456 | views : 340

Refbacks

  • There are currently no refbacks.



Indexed in:

                                   


 

Creative Commons License
Register: Jurnal Ilmiah Teknologi Sistem Informasi is licensed under a Creative Commons Attribution 4.0 International License.