Optimizing Chronic Kidney Disease Detection Through Multi-Modal Data Fusion Using CNN-MLP Architectures

Chronic Kidney Disease (CKD) is a progressive condition that can lead to severe complications if not identified in its early stages. Conventional diagnostic methods rely on laboratory tests, medical imaging, and expert interpretation. However, these approaches often present challenges such as subjective analysis, dependency on specialized professionals, and inefficiencies in integrating multiple data sources. The absence of a unified diagnostic framework that leverages both imaging and biochemical markers reduces the accuracy and reliability of CKD detection. This study presents a Multi-Modal Learning framework that integrates medical imaging with structured clinical data to improve CKD classification. The proposed model incorporates Convolutional Neural Networks (CNN) for extracting spatial features from kidney scans and a Multi-Layer Perceptron (MLP) for processing numerical biomarker data. By combining these complementary data sources, the system enhances diagnostic precision and minimizes uncertainties in disease identification. Advanced preprocessing techniques, including normalization, augmentation, and correlation-based feature selection, refine the dataset for optimal model performance. The integration of CNN and MLP outputs through a fusion layer ensures a comprehensive diagnostic representation. Model evaluation is conducted using standard performance metrics to validate effectiveness. The proposed methodology effectively addresses limitations related to missing values, noisy medical images, and redundant clinical biomarkers, leading to improved prediction outcomes. The fusion of imaging and biochemical data establishes a robust and efficient framework for automated CKD detection, reducing misclassification rates and enhancing diagnostic reliability.