Optimized Feature Selection for Kidney Ultrasound Image Classification Using Binary Coati Weighted Mean Vector Algorithm

The analysis of medical images presents many challenges, especially when making precise diagnoses. In pediatric Chronic Kidney Disease (CKD), early identification is critical because of its gradual progression to significant kidney failure. This study proposes a diagnostic framework for pediatric ultrasound image classification that incorporated machine learning and advanced feature selection methods. This approach is divided into four stages: Preprocessing, feature extraction, feature selection, and classification. Initially, pediatric kidney ultrasound images are enhanced using gaussian median filter. Radiomics features were then extracted, including Gray Level Co-Occurrence Matrix (GLCM), Gray Level Size Zone Matrix (GLSZM), Gray Level Run Length Matrix (GLRLM), Neighboring Gray Tone Difference Matrix (NGTDM), Gray Level Dependence Matrix (GLDM), and first-order statistics. To optimize this feature space, we introduce the Binary Coati Weighted Mean Vector (BinCoWmv) optimization algorithm, which uses a customized fitness function. Herein, the selected features were evaluated using different classifiers: Random Forest (RF), Decision Tree (DT), Support Vector Machine (SVM), Naïve Bayes (NB), K-Nearest Neighbor (KNN), and XG-Boost. Comparative evaluations with existing optimizers, such as the Coati Optimization Algorithm (COA), weighted average vector (INFO), Firefly Algorithm (FFA), and Harris Hawk Optimization (HHO), showed that BinCoWmv achieved a higher classification accuracy. Our framework improves diagnostic reliability and assists radiologist and nephrologist in the early detection of chronic kidney disease in children.