فهرست نشریات علمی دسترسی آزاد ایران

Background and Objectives: Renal diseases constitute a substantial global health burden, highlighting the critical need for efficient and timely treatments to optimize patient outcomes. Early diagnosis of chronic kidney disease (CKD) may be challenging due to its latent nature, which increases the risk of renal damage. Early detection is crucial to halt or reduce the progression of CKD. This research provides an Intelligent Medical Decision Support System (IMDSS) for regulating the course of kidney injury, based on the Deep Attention-Based Spectral Convolutional Neural Network (DA-SCNN) method.   Methods: The study uses the UCI CKD dataset. Min-max normalization is used to normalize the characteristics of the input, resulting in consistent scaling and accelerating the convergence of the Deep Learning (DL) model. In feature selection, Recursive Feature Elimination (RFE) is employed to identify the most advantageous traits for the best possible model performance. Spectral convolutional layers and attention approaches are incorporated in the DA-SCNN architecture to identify important patterns in the data.   Results: The research evaluates the proposed approach by comparing it with existing protocols for treating renal sickness using performance metrics including sensitivity, specificity, accuracy, and precision. The results demonstrate that the DA-SCNN model achieves significantly higher performance with 97% accuracy, 89% precision, 94% sensitivity, and 96% specificity, outperforming conventional methods such as Deep Belief Network (77% accuracy), Long Short-Term Memory (89% accuracy), and Gated Recurrent Unit (85% accuracy).   Conclusion: This study contributes to the advancement of IMDSS in renal healthcare and provides a way for improved patient outcomes and customized treatment regimens.