Meta-Heuristic Optimized Dual-Attention Deep Network with Depth-wise Separability for High-Precision Renal Tumor Diagnosis

Abstract

The human kidney’s primary function is to maintain homeostasis by filtering met-abolic byproducts and toxins from the circulatory system; however, the uncon-trolled proliferation of atypical cells can lead to tumor formation, presenting diag-nostic challenges due to the limitations specifically overfitting, high complexity, and low accuracy of current kidney tumor classification methods. This paper in-troduces a novel deep learning architecture “Meta-Heuristic Optimized Dual-Attention Deep Network with Depth-wise Separability for High-Precision Renal Tumor Diagnosis” designed to analyze CT scan data to accurately localize, quanti-fy (size), and map the spatial distribution of renal tumors, thereby significantly improving therapeutic planning. The process begins with crucial pre-processing using the Guided Triple Gaussian Functioning Filter   to ensure effective noise reduction and optimal image clarity. Subsequently, a Pyramid Convolutional Co-ordinate Attention-based Residual Autoencoder   is utilized for semantic feature extraction, employing its pyramid layers and coordinate attention to focus selec-tively on tumor regions across multiple scales. These refined features are then fed into the Squeeze-and-Excitation based Dual-Attention Module Multi-scale Con-volutional Depthwise Separable Network  , a specialized classifier that integrates dual-attention mechanisms to simultaneously optimize the detection task by em-phasizing channel-specific and spatial feature importance. Finally, to ensure ro-bust generalization, the model's hyperparameters are optimally tuned using a me-taheuristic approach, the Hybrid Spiral Chimp Parrot Optimization Algorithm, re-sulting in a promising solution with demonstrated    an accuracy of  98.81 % , a precision of  97.73 % , a recall of  98.73 % , a Kappa score of  98.36 % , and an F1-Score of  98.81\% , also demonstrating a stable  0.98  K-Fold accuracy for enhanced oncological diagnosis and treatment.