CT Scan Image Classification Using Random Forest Method Based on Texture Features to Differentiate Between Normal Kidney and Cyst

Abstract

Kidney cysts are common benign conditions that require accurate diagnosis through medical imaging. Automated classification systems can assist radiologists in improving diagnostic accuracy and efficiency. This study aims to develop an automated classification system to differentiate normal kidneys from kidney cysts in CT images using the Random Forest algorithm and comprehensive texture feature analysis. We utilized 200 CT scan images from a public dataset, comprising 100 normal kidney images and 100 kidney cyst images. Texture feature extraction was performed using three complementary methods: Histogram (first-order statistics), Gray Level Co-occurrence Matrix (GLCM), and Gray Level Run Length Matrix (GLRLM). Classification was implemented using the Random Forest algorithm in the WEKA environment, with training-testing splits ranging from 5% to 50%. The Random Forest classifier achieved 100% accuracy, precision, and recall across all tested data splits. The combination of the three texture feature extraction methods proved highly effective in capturing distinctive patterns between normal and cystic kidney tissues. The proposed texture-based Random Forest approach demonstrates robust performance for automated kidney cyst classification, showing potential for clinical decision support systems; because every held-out split produced a perfect score, this result is examined in detail in the Results and Discussion section before it is taken as evidence of a deployable model.