Journal of Theoretical and Applied Information Technology
15th February 2019. Vol.97. No 3
© 2005 – ongoing JATIT & LLS
MOKHLED S ALTAThe Exudates detection is an important task in diabetic retinopathy screening. Chain processing method (CPM) for exudates segmentation and detection has been designed for this purpose. CPM proposed new pre-processing approach, which perform not only enhancement task, but also detect and crop object region of interest in the image, which is the main contribution of this work. In post processing stage, a double segmentation and clustering based on chaining watershed and K-means clustering is proposed. The importance of the study lies in the use of clustering segmentation based on chain processing. The method has been validated on DIARETDB1 database; it is compared to 4 ground truths expert's of database evaluation protocols. Following the database evaluation protocol under statistical measurement factors to evaluate overall performance, it is found that the sensitivity specificity, predictive value, and accuracy of 96.12%, 99.76%, 92.64% and 99.69% respectively.
