Artificial intelligence in percutaneous coronary intervention: improved risk prediction of PCI-related complications using an artificial neural network

Objectives Complications after percutaneous coronary intervention (PCI) are common and costly. Risk models for predicting the likelihood of acute kidney injury (AKI), bleeding, stroke and death are limited by accuracy and inability to use non-linear relationships among predictors. Our objective was to develop and validate a set of artificial neural networks (ANN) models to predict five adverse outcomes after PCI—AKI, bleeding, stroke, death and any adverse outcome. Methods We conducted a study of 28 005 patients (training and test cohorts of 21 004 and 7001 patients, respectively) undergoing PCI at five hospitals in the Barnes-Jewish Hospital system. We used an ANN multi-layer perceptron (MLP) architecture based on a set of 278 preprocessed variables. Model accuracy was tested using area under the receiver operating-characteristic curve (AUC). Improved prediction by the MLP model was assessed using integrated discrimination improvement (IDI) and Brier score. Results The fully trained MLP model achieved convergence quickly (<10 epochs) and could accurately predict AKI (77.9%), bleeding (86.5%), death (90.3%) and any adverse outcome (80.6%) in the independent test set. Prediction of stroke was not satisfactory (69.9%). Compared with the currently used models for AKI, bleeding and death prediction, our models showed a significantly higher AUC, IDI and Brier score. Conclusions Using neural network-based models, we accurately predict major adverse events after PCI. Larger studies for replicability and longitudinal studies for evidence of impact are needed to establish these artificial intelligence methods in current PCI practice.