Stage-wise Classification of Chronic Kidney Disease (Stages 1–5) using a Deep Artificial Neural Network on Routine Clinical Blood Parameters

Abstract

Chronic kidney disease (CKD) is often diagnosed late because early stages usually have no symptoms. Identifying the stage of CKD is important for guiding treatment and monitoring progression. In this study, we developed a deep neural network to classify patients into all five CKD stages using routine clinical and laboratory data. The model uses patient demographics, lab results, and comorbidity information to make predictions without invasive tests. The model is trained and evaluated on a single-center dataset comprising 500 anonymized patient records. On the test set, it achieved 98% overall accuracy, with high precision and recall for Stages 1 to 4. The dataset exhibits notable class imbalance, with advanced CKD stages being underrepresented, and Stage 5 performance is limited due to the very small number of available samples. The results show that early and intermediate stages can be identified reliably using standard data. This approach could help diagnostic laboratories support clinicians in decision-making and monitoring. Future work will focus on validation using larger, multi-center datasets and on improving model interpretability.