A machine learning approach to predicting postoperative recurrence in pediatric chronic rhinosinusitis: identification of key metabolic biomarkers

Abstract

Background

Pediatric chronic rhinosinusitis (CRS) is a common chronic inflammatory disease with a high recurrence rate after surgery. This study aimed to construct and validate a machine learning-based predictive model to predict the risk of postoperative recurrence of pediatric CRS and to identify potential biomarkers.

Methods

One hundred and fifteen pediatric patients who underwent functional endoscopic sinus surgery were included. The dataset was divided into training and testing sets (7:3 ratio). Demographic characteristics and laboratory data of were collected and used as features in the predictive models. Eight machine learning algorithms, including Random forest (RF), were applied to construct predictive models. Feature selection was performed, and hyperparameters were optimized using a grid search with 10-fold cross-validation. Model performance was assessed using the area under the receiver operating characteristic curve (AUC) and F1 score.

Results

The Random Forest model performed best in predicting the postoperative recurrence of CRS in children, with AUC of 0.830. Feature selection analyses showed that metabolic markers, such as DBIL, Glu, and TBIL, had an important role in predicting CRS recurrence. In the test set, the AUC of the RF model reached 0.864 and an F1 score of 0.9, showing good stability and generalization ability.

Conclusion

In this study, we successfully constructed a model to predict the postoperative recurrence of pediatric CRS. The predictive model indicated that key metabolites were significantly associated with disease outcomes, and individualized management of postoperative pediatric CRS.