A novel hybrid approach for multi stage kidney cancer diagnosis using RCC ProbNet.
Authors
Affiliations (6)
Affiliations (6)
- Institute of Information Technology, Khwaja Fareed University of Engineering and Information Technology, RahimYar Khan, 64200, Pakistan.
- Institute of Information Technology, Khwaja Fareed University of Engineering and Information Technology, RahimYar Khan, 64200, Pakistan. [email protected].
- Department of Computer Science, COMSATS University Islamabad, Sahiwal Campus, Sahiwal, 57000, Pakistan.
- Centre for Smart Systems and Automation, CoE for Robotics and Sensing Technologies, Faculty of Artificial Intelligence and Engineering, Multimedia University, Persiaran Multimedia, 63100, Cyberjaya, Selangor, Malaysia.
- Department of Computer Engineering, Gachon University, Seongnam-si, 13120, South Korea.
- Centre for Smart Systems and Automation, CoE for Robotics and Sensing Technologies, Faculty of Artificial Intelligence and Engineering, Multimedia University, Persiaran Multimedia, 63100, Cyberjaya, Selangor, Malaysia. [email protected].
Abstract
Kidney renal cell carcinoma (RCC) is a highly invasive malignancy, where early-stage diagnosis can significantly reduce treatment complexity and mortality risk. This study aims to enhance diagnostic precision through a novel model, RCC-ProbNet, capable of identifying and classifying RCC across various stages. We introduce RCC-ProbNet, a hybrid deep learning model that incorporates an initial stage where features are extracted from medical imaging data, and a subsequent construction of a probabilistic feature model. This hybridized method could allow a more granular feature representation through incorporating a variety of features and enhance the model's ability to differentiate different RCC stages. The model is integrated with a Logistic Regression (LR) classifier for the final stage prediction. We validate performance using k-fold cross-validation. Overall, we obtain a notable diagnostic accuracy of 99.93% by RCC-ProbNet + LR, surpassing the current state-of-the-art techniques. Next, the model shows strong stability in different cross-validation folds. The comparative results also demonstrate that our model outperforms traditional and deep learning methods for RCC classification tasks consistently. The outstanding performance and the invulnerability of RCC-ProbNet + LR make it a powerful classifier for early RCC screening and staging in clinical practice. Its high performance indicates it has the potential to assist in timely and personalised treatments. Its future iterations may broaden the use toward real-time clinical decision support systems and consequently optimise patients' outcomes regarding renal oncology.