Automatic Detection of Exudates in Retinal Image Using Statistical Techniques.

Khare, Vijay

Diabetic retinopathy is an eye disease seen widely in diabetes patient. It is one of the leading causes of vision loss or blindness characterised by exudates as one of its symptoms. In this paper an objective is to develop algorithm for exudates detection in poorly contrasted or low-quality images. To identify the exudates, initially pre-processing operation is performed on retinal images to retrieve the colour information using histogram specification operation. A new method is proposed to obtain the textural feature of each pixel named as gray level pixel count matrix (GLPCM). The GLPCM method is compared with existing statistical technique i.e. gray level co-occurrence matrix (GLCM) and gray level run length matrix (GLRLM). The classification operation is performed using BPNN classifier. The result of proposed feature extraction technique has been validated based on the ground truth details provided in the dataset and achieved specificity and sensitivity 98.9%, 90.6% on DIARETDB0, DIARETDB1 and DRIVE dataset. In this study also compared different segmentation technique on similar dataset.

RETINAL imaging; AUTOMATIC detection in radar; DIABETIC retinopathy

Ingénierie des Systèmes d'Information, 2023, Vol 28, Issue 1, p255

1633-1311

Academic Journal

10.18280/isi.280130