Artificial Intelligence–Based Prediction of Recurrence and Personalized Treatment Planning in Pediatric Pilonidal Sinus Disease

Abstract

Purpose

This study aimed to assess the feasibility of artificial intelligence (AI)–based models for predicting recurrence risk and supporting individualized treatment strategies in pediatric pilonidal sinus disease (PSD).

Methods

Clinical data from 242 pediatric PSD patients were retrospectively analyzed. Two machine learning (ML) models were developed: (1) a binary classifier for recurrence prediction and (2) a multiclass classifier for treatment modality selection. Model performance was evaluated using 5-fold cross-validation based on accuracy, area under the receiver operating characteristic curve (ROC-AUC), F1 score, and confusion matrix. Model interpretability was assessed using SHapley Additive exPlanations (SHAP) analysis to identify the most influential predictive features.

Results

Among 242 patients, 59.1% underwent surgical excision (unroofing with secondary healing), 33.1% received crystallized phenol, and 7.8% were treated with silver nitrate. The overall recurrence rate was 5.7%, with no significant differences among treatment modalities. Longer healing duration, larger lesion size, higher body mass index (BMI), and an increased number of sinus tracts were associated with higher recurrence risk. SHAP analysis revealed that healing time, BMI, age, and the number of sinus tracts were the most influential predictors in both recurrence and treatment selection models.

Conclusion

AI-based predictive models can effectively estimate recurrence risk and assist in tailoring individualized management strategies for pediatric PSD. Further multicenter, prospective studies are warranted to validate these findings and support clinical integration of AI-assisted decision systems.