DefinePK

Abstract

Kidney disease occurs when the kidneys become weakened and lose their ability to cleanse the blood. Most individuals show no symptoms in the early stages of kidney disease. As the condition progresses, toxins can accumulate in the bloodstream, causing complications such as anemia, hypertension, diabetes, osteopenia, and nerve damage. While these issues often develop gradually and without noticeable symptoms, they can eventually lead to sudden renal failure. Early identification of kidney disease allows for the most effective treatment. Predicting kidney function and disease using kidney ultrasound imaging is widely considered in clinical practice due to its safety, simplicity, and affordability. Several works on kidney disease prediction have already been done, but accuracy improvement is still needed. To solve this issue, the research proposes optimized Feature Selection (FS) and an Artificial Intelligence (AI) model for effective kidney disease prediction from ultrasound images. The own dataset with normal and diseased kidney images is created and processed. The processed image features are extracted using the Gray-Level Co-Occurrence Matrix (GLCM) technique. The most significant features are retrieved using Particle Swarm Optimization (PSO), a bio-inspired algorithm. The features of GLCM and PSO are given to the Support Vector Machine (SVM) and Convolutional Neural Network (CNN) models for the classification of diseased and normal images. The SVM model with GLCM and PSO features and CNN with GLCM and PSO features are evaluated using accuracy, precision, recall, and F1 score. Both models show better accuracy improvement by using PSO features. The experimental findings show that the PSO-CNN model gives the maximum accuracy of 98.57% when compared with other models