Deep Neural Network-Based Survival Analysis for Skin Cancer Prediction in Heart Transplant Recipients

Heart-transplant recipients are at high risk of developing skin cancer, while Squamous Cell Carcinoma (SCC) and Basal Cell Carcinoma (BCC) are commonly detected. This paper utilized the database from the United Network for Organ Sharing (UNOS) to study the incidence rate of SCC and BCC among heart transplant recipients. Cox proportional hazards model and two deep neural network-based models were studied, and their performance were compared. In addition, Lasso regression, Chi-square test, and Wilcoxon signed-rank test were applied to identify key risk factors. The neural network-based survival models showed better accuracy compared to the standard Cox regression model, which indicates the advantage of deep learning approaches in survival analysis and risk prediction for post-transplant skin cancer. Clinical relevance- This study investigates the performance of deep learning (DL) models in clinical applications for predicting the risk of skin cancer in heart transplant recipients. The DL models outperform the standard models in assessing the incidence rate of skin cancer across different time spans.