Deep Learning-Based Survival Prediction in Pediatric Kidney Transplantation: A Comprehensive Model for Improved Post-Transplant Outcomes

Kidney transplantation is the optimal treatment for pediatric patients with end-stage renal disease. Accurate survival prediction post-transplantation is essential for improving patient outcomes and tailoring medical decisions. This paper presents a survival prediction model using the Deeplearning4J Multi-Layer Perceptron classifier, specifically designed for pediatric kidney transplant recipients with living donors. We utilized a real-world dataset including pediatric patient records collected from United Network for Organ Sharing consisting of clinical, demographic, and transplant-specific factors to train and evaluate the model. Top ranked attributes were selected through dimensionality reduction technique with ten-fold cross validation. Experimental results with performance evaluation show that the proposed classifier outperforms traditional machine learning models, with significant improvements in predictive accuracy of 92.8%, achieving an Area Under the Curve of 0.913. To prove the accuracy of model, the results were compared with existing approaches and machine learning techniques.