Automatic Detection of Proliferative Diabetic Retinopathy With Hybrid Feature Extraction Based on Scale Space Analysis and Tracking

Sabilla, Wilda Imama (Institut Teknologi Sepuluh Nopember, Surabaya)
Soelaiman, Rully (Institut Teknologi Sepuluh Nopember, Surabaya)
Fatichah, Chastine (Institut Teknologi Sepuluh Nopember, Surabaya)

IPTEK Journal of Proceedings Series
No 1 (2015): 1st International Seminar on Science and Technology (ISST) 2015

Feature extraction is a process to obtain the characteristics or features of an object where the value of the features will be used for analysis in the next process. In retinal image, extraction of blood vessels’ characteristics can be used for detection of proliferative diabetic retinopathy (PDR). Retinal blood vessels’ features can be obtained directly with segmented image and with additional spatial method. For PDR detection, we need the suitable method that can produce maximum feature representation. This paper proposed hybrid feature extraction using a scale space analysis method and tracking with Bayesian probability. The result of the retinal images classification from STARE database using soft threshold m-Mediods classifier shows the best accuracy of 98.1%.