An optimized intelligent boosting model for diabetic retinopathy segmentation severity analysis using fundus images

Abstract

In today's scenario, many people suffer from Diabetic Retinopathy (DR), due to different lifestyles and cultures. Hence, the exact severity analysis system is the most required application to avoid vision loss. The Neural network with multiple decision functions already existed for this severity analysis case. However, those models do not give the proper outcome in exact segmentation, leading to improper severity analysis outcomes. So, the current study aims to design a novel Squirrel Search-based Extreme Boosting (SSbEB) for accurately segmenting and estimating the severity range. Initially, the DR database was filtered and entered into the classification layer, then the features were extracted, and the abnormal region was segmented. Here, incorporating the squirrel features inthe extreme boosting has afforded the finest feature analysis and segmentation outcome, which help predict the DR severity level with themaximum possible rate. The severity score of the segmented region was determined as normal, mild, severe, moderate, and proliferative. Hence, the designed model is implemented in the python platform, and the performance parameters, such as precision, specificity, accuracy, and recall, have been measured and compared with other models. Hence, the recorded exact severity analysis score is 94.4%, which is quite better than the past models. Thus, the implemented model is suitable for the DR severity analysis system and supported for real-time disease analysis applications.

Keywords

• severity classification
• affected region segmentation
• squirrel optimization
• extreme boosting
• feature analysis