Integrating Laboratory and Clinical Perspectives for AI-Based Prediction of Chronic Kidney Disease

Abstract

Chronic Kidney Disease (CKD) is a silent yet progressive condition, often diagnosed only at advanced stages when therapeutic options are limited and costly. Early prediction is therefore essential to reduce the incidence of end-stage renal disease (ESRD), improve patient outcomes, and optimize healthcare resources. Existing research on CKD prediction typically follows two separate paths: (i) laboratory-rich datasets, which provide precise biological indicators but suffer from small sample sizes and missing values, and (ii) large-scale clinical datasets, which offer population-level coverage but lack biological detail. This methodological divide constrains the development of comprehensive, clinically actionable prediction systems. In this study, we conduct a comparative analysis using both dataset types, applying a range of machine learning and deep learning models. Classical algorithms such as Logistic Regression, Random Forests, and XGBoost were evaluated alongside advanced architectures including Deep Neural Networks (DNNs) and Convolutional Neural Networks (CNNs). The results show that laboratory-based datasets yield high diagnostic precision, while clinical datasets support scalable prediction across diverse populations. Together, these findings highlight the complementary strengths of biological and clinical data, demonstrating that integrating both perspectives can enhance the robustness and practical relevance of AI-driven CKD prediction systems.

Ähnliche Arbeiten