Leveraging Machine Learning for Pediatric Appendicitis Diagnosis: A Retrospective Study Integrating Clinical,...

Appendicitis is the most common surgical emergency in pediatric patients, requiring timely diagnosis to prevent complications. This study introduces an innovative approach by integrating clinical, laboratory, and imaging features with advanced machine-learning techniques to enhance diagnostic accuracy in pediatric appendicitis.

A retrospective analysis was conducted on 782 pediatric patients from the Regensburg Pediatric Appendicitis Data set. Clinical scores, laboratory markers, and imaging findings were analyzed. Statistical comparisons were performed using independent t-tests and χ ² tests, with significance set at p < 0.05. Predictive models, including logistic regression and machine learning classifiers, were developed and evaluated using accuracy, precision, recall, and F1-score.

Significant differences were observed in clinical scores (e.g., Alvarado Score and Pediatric Appendicitis Score) and laboratory markers (e.g., WBC count and neutrophil percentage) between appendicitis (AA) and non-appendicitis (Non-AA) groups (p < 0.001). Imaging features, including appendix diameter, also demonstrated diagnostic value. Among predictive models, the Random Forest classifier achieved the highest accuracy (94.5%), with strong precision (93.8%) and recall (95.2%) for appendicitis diagnosis.

This study represents a novel application of machine learning models, particularly Random Forest, to enhance diagnostic accuracy for pediatric appendicitis. The integration of clinical, laboratory, and imaging features offers a comprehensive and precise diagnostic framework. Further validation in diverse populations is recommended.