Comparing the Accuracy of Neural Network Models and Conventional Tests in Diagnosis of Suspected Acute Appendicitis

Background and Diagnosis of acute appendicitis can be difficult due to similarity of symptoms to many abdominal diseases. Delayed diagnosis could expose the patient to serious conditions. In this study we compared the Artificial Neural Network (ANN) models and conventional laboratory tests in diagnosis of appendicitis. The study population included 100 patients with suspected appendicitis. White Blood Cells (WBC), Procalcitonin (PCT), C-reactive protein (CRP) and PMN were measured as conventional diagnostic tests and ANN was applied as a combinational test. Definite diagnosis of appendicitis was made based on pathology results. For each test, Receiver Operating Characteristic (ROC) curve and sensitivity and specificity tables were used. The mean age of patients was 28.01±12.68 years and 71 (71%) were male. The sensitivity of ANN model was 97.59 and the sensitivities of CRP and WBC were 92.77% and 85.54%, respectively. The highest accuracy in diagnosis of acute appendicitis was achieved by ANN (88%). This study showed that combinational test using ANN could be more beneficial in diagnosis of acute appendicitis.