516 Artificial Neural Networks Predict the Need for Permanent CSF Diversion After Posterior Fossa Tumor Surgery

Abstract

INTRODUCTION: Resection of posterior fossa tumors (PFTs) can result in hydrocephalus that requires permanent CSF diversion. Clinicians need a durable method to predict occurrence of CSF diversion in this population. METHODS: We collected pre-operative and post-operative variables on 510 patients that underwent PFT surgery at our center in a retrospective fashion to train several statistical classifiers to predict the need for permanent CSF diversion as a binary class. A total of 62 classifiers relevant to our data structure were surveyed including regression models, decision trees, Bayesian models, and multilayer perceptron artificial neural networks (ANN). Models were trained using the (N=510) retrospective data using 10-fold cross validation to obtain accuracy metrics. Given the low incidence of our positive outcome (12%), we used the true positive rate along with the area under the receiver operating characteristic curve (AUC) to compare models. The best performing model was then prospectively validated on a set of 90 patients. RESULTS: 12% of patients in our dataset required permanent CSF diversion after PFT surgery. Of the trained models, 8 classifiers had an AUC greater than 0.5 on prospective testing. Artificial neural networks (ANN) demonstrated the highest AUC of 0.902 with a true positive rate of 83.3%. Despite comparable AUC, the remaining classifiers had a true positive rate below 35% (compared to ANN, P<0.0001). The negative predictive value of the ANN model was 98.8%. CONCLUSION: ANN-based models can reliably predict the need for ventriculoperitoneal shunt after PFT surgery.