Integration of Physical Examination, Old and New Biomarkers, and Ultrasonography by Using Neural Networks for Pediatric Appendicitis.

Abstract

The objective of this study was to evaluate physical examinations, imaging, and laboratory analyses individually and combined using innovative statistical analysis methods for the accurate diagnosis of pediatric appendicitis. Patients admitted to hospital with symptoms of abdominal pain whose pediatric appendicitis scores greater than 3 were included in the study. Clinical, radiologic, and laboratory findings and as a new biomarker calprotectin (CPT) concentrations were evaluated individually and combined using artificial neural networks (ANNs), which revealed latent relationships for a definitive diagnosis. Three hundred twenty patients were evaluated (190 appendicitis [43 perforated] vs 130 no appendicitis). The mean ± SD age was 11.3 ± 3.6 years and 63% were male. Pediatric appendicitis scores, white blood cell (WBC) count, absolute neutrophil count (ANC), C-reactive protein (CRP) level, procalcitonin (PCT) and CPT concentrations were higher in the appendicitis group; however, only WBC and ANC were higher in first 24 hours of pain. White blood cells and CRP were diagnostic markers in patients whose appendix could not be visualized using ultrasonography (US). On classic receiver operating characteristic (ROC) analysis, the areas under the curve (AUCs) were not strong enough for differential diagnosis (WBC, 0.73; ANC, 0.72; CRP, 0.65; PCT and CPT, 0.61). However, when the physical examination, US, and laboratory findings were analyzed in a multivariate model and the ROC analysis obtained from the variables with ANN, an ROC curve could be obtained with 0.91 AUC, 89.8% sensitivity, and 81.2% specificity. C-reactive protein and PCT were diagnostic for perforated appendicitis with 0.83 and 0.75 AUC on ROC. Although none of the biomarkers were sufficient for an accurate diagnosis of appendicitis individually, a combination of physical examination and laboratory and US was a good diagnostic tool for pediatric appendicitis.