A NOVEL RETINAL SEGMENTATION BASED ON ADAPTIVE FUZZY THRESHOLDING
Abstract
When assessing health concern, eye examination can be as successful as physical examination. Retina testing may be first warning to identify a different of unknown health problems, includees diabetes & pre-diabetes. Ophthalmologists depend more on the binary segmented image of the medical diagnosis and prognosis system of the retina fundus, where the efficient of the segmented lines, optical disks and irregular lesions greatly affects the reliability of the diagnosis. This paper suggests automated retinal fundus entity segmentation system based on the same core segmentation algorithm, consisting of three segmentation subsystems. With respect to wide variations in characteristics and properties, each subsystem recovers Retinal vessels, optic disks & lesions exudate without surface examination and synthesis. Such anatomical structures can be described in a single session with high efficient, even in the pathological retinal image, for the sake of streamlined diagnosis and full medical understanding. The proposed system blends adaptive fuzzy threshold with computational morphology using a robust hybrid segmentation algorithm. Four reference datasets are used to check the proposed system: DRIVE & STARE (vessels), DRISHTI-GS (optical disks), DIARETDB1 (exuding lesions). Competitive segmentation calibre is achieved, outperforms a number of current systems and shows the ability to manage certain anatomical heterogeneous models.
Keywords
Diabetes , Retinal vessels, optic disks, drishti-gs (optical disks)