RCC Subtype Classification Using Contrast-Enhanced CT Images and Convolutional Neural Networks

Renal cell carcinoma (RCC) is an aggressive and widespread form of kidney cancer that requires early and accurate diagnosis for effective treatment. Contrast-enhanced computed tomography (CECT) plays a crucial role in RCC detection, providing high-resolution imaging for detailed tumor analysis. This study introduces a deep learning-based approach using AlexNet for automated RCC classification, eliminating the need for manual image curation. A dataset of 580 abdominal CT images representing various RCC subtypes and non-tumor cases was preprocessed and categorized into five groups: clear cell RCC (ccRCC), papillary RCC (pRCC), chromophobe RCC (chRCC), healthy kidney, and non-kidney images. The model was trained using stochastic gradient descent with momentum (SGDM), achieving optimal classification accuracy with 13 epochs and a learning rate of 0.3. Performance evaluation using a confusion matrix demonstrated the model’s effectiveness in accurately distinguishing RCC subtypes. This study highlights the potential of deep learning in aiding radiologists with automated RCC diagnosis, ultimately enhancing early detection and clinical decision-making.