Advanced Deep Learning Model for Enhanced Kidney Disease Detection

Automated Chronic Kidney Disease (CKD) is a swiftly growing non-communicable condition that contributes to a substantial number of deaths and illnesses worldwide. Affecting over 10% of the global population, CKD results in millions of fatalities each year. In this study, an advanced deep learning model for kidney disease detection is proposed that combines convolutional and involution layers to improve classification accuracy. The model processes input images by first applying a 2D convolution with 32 filters of size 3x3, followed by ReLU activation to extract key features. Max-pooling layers are then used to reduce spatial dimensions while retaining crucial information. The convolutional layers are sequentially stacked with larger filter sizes, resulting in progressively richer feature representations. To enhance feature extraction, the model integrates custom involution layers, which extend traditional convolution operations and are particularly effective for processing large-scale image data. These involution layers are applied after the convolutional steps, utilizing 128 and 256 filters of size 3 × 3, followed by max-pooling to further refine the extracted features. A comparative analysis was carried out between the proposed custom Convolutional Neural Network (CNN) model and previous research works, utilizing metrics such as Recall, Precision, F1-Score, Accuracy, Loss, Area Under the Curve (AUC) and Receiver Operating Characteristic (ROC) curves. The results indicate that the proposed model achieves superior classification prediction performance, with an accuracy of 99% and reduced computational complexity. Additionally, it demonstrates a enhanced F1-Score of 100%, 98%, 98% and 100% for Cyst, Normal, Stone and Tumor respectively.