The Development of a Tool for the Detection of Cotton Wool Spots, Haemorrhage, and Exudates Using Multi-resolution Analysis

Develop a graphical user interface tool for the diagnosis of diabetic retinopathy complications such cotton wool spots, haemorrhages, and exudates (EX). By employing digital image processing techniques and multi-resolution analysis techniques (Wavelet decomposition). In order to do a research on this technique, we have utilised local databases created by Dr. Manoj Saswade in addition to web databases such as DiarectDB0, DiarectDB, STARE, and DRIVE. In all, we have 1180 high resolution images of the fundus. The optic disc is removed from each and every fundus picture using preprocessing, which is the initial step in this approach. The following stage is to implement numerous algorithms for DIP, such as the intensity transformation, complement function, histogram equalisation, and threshold. Other algorithms to build include the histogram equalisation and threshold. Wavelet decomposition will be utilised within the symlet wavelet. Then the removal of the lesions and apply statistical methods. After that, analyse the performance of the proposed method by computing the sensitivity and specificity using a receiver operating characteristic curve. Using several statistical methods, we were able to obtain an overall result of 94%.