Advanced Machine Learning Models for Accurate Kidney Cancer Classification Using CT Images

Abstract

Kidney cancer, particularly renal cell carcinoma (RCC), poses significant challenges in early and accurate diagnosis due to the complexity of tumor characteristics in computerized tomography (CT) images. Traditional diagnostic approaches often struggle with variability in data and lack the precision required for effective clinical decision-making. This study aims to develop and evaluate machine learning (ML) models for the accurate classification of kidney cancer using CT images, focusing on improving diagnostic precision and addressing potential challenges of overfitting and dataset heterogeneity. Two ML models, Support Vector Machines (SVM) and Multi-Layer Perceptrons (MLP), were employed for classification. Key attribute extraction techniques, including grayscale-level co-occurrence matrix (GLCM) and Gabor filters, were utilized to capture texture and structural features of CT images. Data normalization and preprocessing ensured consistency and enhanced model reliability. The SVM model achieved an accuracy of 93%, while the MLP model demonstrated superior performance with a 99.64% accuracy rate. These results highlight the MLP model's ability to capture complex patterns in the data. However, the exceptional accuracy of the MLP model raises concerns about potential overfitting, warranting further evaluation on more diverse datasets. This study underscores the potential of ML techniques, particularly MLP, in enhancing the accuracy of kidney cancer diagnosis. Integrating such advanced ML models into clinical workflows could significantly improve patient outcomes.