Use of Deep Neural Networks and Bayesian Analysis to Identify Risk Factors for Poor Outcome After Pediatric Cardiac Surgery

Pediatric cardiac surgery with cardiopulmonary bypass is associated with significant morbidity and mortality. Also score systems for risk factors, such as Risk Adjustment for Congenital Heart surgery (RACHS 1) score or the ARISTOTLE score, have been developed, outcome prediction remains difficult. New mathematical methods using deep neural networks associated with Bayesian statistical methods have been developed to give a better understanding of the complex interaction between different risk factors, to identify risk factors and group them in related families. This method has been successfully used to predict mortality in dialysis patient as well as to better describe complex psychiatric syndromes.

The primary hypothesis of this study is that the use of these tools will give a better understanding on the factors affecting outcome after pediatric cardiac surgery.

A network analysis using Gaussian Graphical Models, Mixed Graphical models and Bayesian networks will be used to identify single or groups of risk factors for morbidity and mortality after pediatric cardiac surgery under cardiopulmonary bypass.