Renal Cell Carcinoma Classification: Deep Learning with MLflow, DVC, and AWS CI/CD Deployment

One prevalent type of kidney cancer that can be challenging to diagnose and treat is renal cell carcinoma (RCC). This article offers a novel method for applying deep learning techniques to the classification of RCC data. In particular, we examine the performance of three distinct models: ResNet50, VGG16, and a special 2D convolutional neural network (CNN). Our method makes use of cutting-edge machine learning frameworks like MLflow and DVC to streamline model building, experimentation, and version control. We describe an effective deployment pipeline that facilitates seamless integration with healthcare systems using AWS CI/CD infrastructure. The dataset, which includes 8,143 normal images and $\mathbf{1 0, 8 7 8}$ tumor images, is sourced from the Cancer Imaging Archive. CNN-4 was the classification model that obtained the highest accuracy, and CNN-6, ResNet50, and VGG16 were the detection models that achieved the highest accuracy, with respective results of $99 \%, 98 \%$, and 98%. These results demonstrate how our approach can be utilized to improve patient outcomes and early detection of RCC while simultaneously improving the accuracy of RCC diagnoses.