Development of a predictive model for one-year postoperative recovery in patients with lumbar disc herniation based on deep learning and machine learning

Abstract

AbstractBackgroundThe aim of this study is to develop a predictive model utilizing deep learning and machine learning techniques that will inform clinical decision-making by predicting the one-year postoperative recovery of patients with lumbar disc herniation.MethodsThe clinical data of 273 inpatients who underwent tubular microdiscectomy (TMD) between January 2018 and January 2021 were retrospectively analyzed as variables. The dataset was randomly divided into a training set (n = 191) and a test set (n = 82) using a ten-fold cross-validation technique. Various deep learning and machine learning algorithms including decision trees, random forests, extreme gradient boosting, support vector machines, parsimonious Bayes, K-nearest neighbors, L2-regularized logistic regression, unregularized logistic regression, and neural networks were employed to develop predictive models for the recovery of patients with lumbar disc herniation one year after surgery. The cure rate score of lumbar JOA score one year after TMD was used as an outcome indicator, and the area under the receiver operating characteristic curve (AUC) was selected as the main measure of learner superiority.ResultsThe correlation matrix heat map indicated that there was no need to use data reduction techniques prior to model development. The predictive model employing both machine learning and deep learning algorithms was constructed using 43 collected variables. Among the nine algorithms utilized, the L2-regularized logistic regression algorithm demonstrated the highest value of the area under the receiver operating characteristic curve (AUC).ConclusionsOur study findings demonstrate that the L2-regularized logistic regression algorithm provides superior predictive performance for the recovery of patients with lumbar disc herniation one year after surgery.