A bio-inspired fall webworm optimization algorithm for feature selection and support vector machine optimization for retinal abnormalities detection

The early detection of retinal abnormalities such as diabetic retinopathy (DR) can be performed using the computerized analysis of retinal fundus images. The most significant complications associated with the DR detection are noise artifacts occurred as a result of unsuitable illumination, the overlapping of blood vascular structure and lesions as they have same intensities, and missing of data happened due to the analysis of large amount of data. Hence the improved technique capable of overcoming all these limitations must be presented in early detection of DR and other retinal abnormalities. Some of previous blood vessel segmentation methods provide better accuracy with normal retinal images and requires less computation time. In this work, an automated process using an optimized SVM classifier and a new feature extraction is presented. The proposed feature extraction process is sum of minimum (SOM) local difference pattern (LDP) (SOMLDP) which is developed from the computation of difference between pixels. This feature extraction produces precise feature information with reduced size. In addition, feature selection process is employed to select more important features, using a new optimization algorithm developed from the behavior of fall webworm (FWW) species. FWW optimization algorithm is also applied for the optimal tuning of support vector machine (SVM) classifier. The main objective of the paper is to present an automated detection of DR with more accuracy, less memory and reduced computation time. The performance of proposed technique is validated with publicly available standard dataset Messidor-2 by evaluating the metrics such as sensitivity, specificity and accuracy. The simulation results depict that sensitivity, specificity and accuracy of 0.8235, 0.9892 and 0.9879 is attained respectively. The FWW optimization algorithm is also validated by analyzing the computation time performance and comparing the performance of FWW algorithm with other optimization algorithms.