An Extensive Detection of Hemorrhages from Diabetic Retinopathy Images


Author (s):

  1. C.A.SATHIYAMOORTHY, Faculty of Electronics Engineering, Satyabhama University, Chennai, India, satya.cas@gmail.com
  2. Dr. G. Kulanthaivel, NITTTR, Taramani, Chennai, India

Abstract:

One of the most serious impediments rampant among diabetic patients is diabetic retinopathy which affects blood vessels. The ophthalmologists can be able to provide an appropriate treatment to the patients using the most advanced technologies like lasertreatment to avoid blindness. In this study, the author proposes a method based on binary image processing combined fuzzy logic to detect hemorrhages from diabetic retinopathy images. In the starting stage, the hemorrhages are identified by binary operations, which eliminate the optic disc for fast and easy detection. Subsequently, hemorrhages are extracted by utilizing fuzzy logic methodologies, where it uses the RGB color space of the retinal images as the fuzzy sets and membership functions. The resultant image pixels received from the fuzzy operation are separated and grouped into red, green and blue channel based pixels of the hemorrhages. This fuzzy result is computed for hemorrhages according to the portion in the hemorrhages area detected from the diabetic retinopathy images. It is evaluated by comparing the detected hemorrhages with hand-drawn ground-truth images. The extensive experiment is carried out in MATLAB software then obtained sensitivity and specificity of detecting hemorrhages as 89.65 and 99.99% respectively.

No of downloads: 23