The appendix is in the right lower abdomen, where food debris and bacteria accumulate and produce inflammation1,2,3. Acute appendicitis is the most common surgical emergency worldwide, with a lifetime risk of 7–8%4. The social impact and healthcare burden are very high5, and it is also one of the common causes of surgical acute abdomen in the elderly. Acute appendicitis can be categorized into two types: UA, which is defined as cellulitis appendicitis without signs of necrosis or perforation6. And CA, this has focal or transmural necrosis that may eventually lead to perforation7. It is essential to differentiate between these two conditions because UA can be treated conservatively with antibiotics without surgery8,9 and may even resolve independently without antibiotic therapy10,11. Patients with CA require emergency appendectomy.Elderly people often suffer from a range of underlying diseases such as respiratory, circulatory, endocrine and metabolic disorders, and reduced immune function. Conservative treatment may also be attempted in elderly UA patients with mild clinical symptoms who have a strong desire to avoid surgery and accept the risk of recurrence. However, the probability of appendiceal perforation and death is significantly higher in elderly patients than in patients without perforation12. Moreover, delayed diagnosis of complicated appendicitis may lead to associated complications such as perforation and peritonitis, resulting in severe morbidity and mortality, especially in the elderly patient population with comorbidities13, as well as prolonged hospitalization, loss of employment, increased costs due to additional investigations, and psychosocial problems14.Therefore, essential to diagnose acute appendicitis correctly, and a “two-stage” approach is generally used in clinical practice. The first stage involves diagnosing “acute appendicitis.” In patients who do not have acute appendicitis, the cause of the abdominal pain needs to be identified and treated promptly. After the diagnosis of acute appendicitis is confirmed, the second stage involves differentiating between UA and CA and adopting different treatment protocols for the other conditions.Currently, historical information and laboratory findings are still considered the cornerstone of the diagnosis of acute appendicitis. Still, there is much intra-observer variability, and accuracy could be better. It has been shown that physicians fail to make a correct clinical diagnosis in all patients with acute abdominal pain based solely on history and routine laboratory tests15.In addition to clinical tests, complete blood count parameters (leukocytes, neutrophils, lymphocytes, platelets, platelet derivatives), which are part of routine blood biochemistry parameters, as well as markers such as total bilirubin (TBil), C-reactive protein (CRP), and procalcitonin are widely used as a next step in the diagnosis of acute appendicitis, as they vary according to the presence and severity of inflammation16. Individually, these markers of inflammation have a weak discriminatory ability, but when combined, they have a higher discriminative ability in diagnosing acute appendicitis versus non-appendicitis17. However, according to a prospective data study including 1024 patients with clinical suspicion of acute appendicitis, this combination was found not to rule in or rule out appendicitis18 adequately.Scoring systems such as AIRS and Alvarado, which consist of history, physical examination findings, blood biochemical parameters, and radiologic instruments and their combinations, can differentiate between simple and complex appendicitis19. None of these studies mentioned diagnostic accuracy measures, so sensitivity and specificity could not be calculated. Two other studies reported on the design of scoring systems, including clinical and biochemical features, neither reported diagnostic accuracy measures20,21. Imaging is essential in differentiating simple from complex appendicitis, with ultrasound and computed tomography (CT) imaging improving diagnostic sensitivity and specificity. Still, more sensitivity is needed for all parameters, while these tools have the disadvantages of being highly operator-dependent and radiation-exposed, respectively22.The above suggests that the diagnosis of appendicitis relies on clinical evaluation, laboratory tests, and imaging studies, including ultrasound and computed tomography (CT) scans. Still, these methods are fraught with limitations, such as diagnostic inaccuracies and time-consuming procedures, which can lead to severe complications such as appendiceal perforation and sepsis.To overcome these challenges, artificial intelligence (AI) has been widely used in clinical settings to assist doctors in medical diagnoses. Artificial intelligence refers to a machine’s ability to mimic human cognitive processes to perform tasks autonomously. Relevant literature has shown that AI techniques are advantageous in diagnosing acute appendicitis. Alramadhan et al. found that artificial neural networks (ANNs) can accurately predict the risk of intra-abdominal abscess (IAA) after appendectomy with an accuracy of 89.84%, sensitivity of 70%, and specificity of 93.61% on the test set23.Xia et al. constructed a diagnostic model using an improved grasshopper optimization algorithm based support vector mechanism to distinguish between complex and uncomplicated appendicitis. The optimal model yielded an average of 83.56% accuracy, 81.71% sensitivity, 85.33% specificity, and 0.6732 Matthews correlation coefficient24. Kim et al. developed a model using multivariate logistic regression and Bayesian information criterion to assess the model’s performance in the validation dataset through calibration plots and area under the curve (AUC), respectively. The model’s calibration and discrimination performance could identify patients with definite uncomplicated appendicitis who could benefit from non-surgical treatment with a low risk of failure25.Shahmoradi et al. compared the output of the optimized SVM artificial neural network with pathological findings. They showed that the network’s sensitivity, specificity, and accuracy for diagnosing acute appendicitis were 91.7%, 96.2%, and 95%, respectively26. Erkent et al. used a decision tree approach to determine the severity of acute appendicitis (AA) without imaging methods27.The “Validity of machine learning in detecting complicated appendicitis in a resource-limited setting: findings from Vietnam” article also builds and validates a machine learning model to facilitate the detection of complex appendicitis28. Phan–Mai et al. used several machine learning methods, including SVM, to classify patients with complicated appendicitis and patients with uncomplicated appendicitis. The experimental results show that the GB model has high validity. Both articles are about detecting and determining complex appendicitis from a machine learning perspective. The differences are as follows: (1) The datasets are different; (2) The number of machine learning models used is different; (3) In our paper, after determining GB as the optimal model, we used SHAP technology to visualize the weights of each parameter; (4) We developed the complex appendix diagnostic Shiny app to diagnose UA and CA to maximize generalizability and facilitate translation into real-world clinical practice.The clinical data for this study was obtained from the article “Development and Validation of a Clinical Prediction Model for Complicated Appendicitis in the Elderly”. The article used the SPSS 26. 0 and R 4.0.2 software to create a CA prediction model to help clinicians quickly determine the type of acute appendicitis. It was found that three parameters based on abdominal pain duration, peritonitis, and total bilirubin could help physicians quickly and effectively determine UA or CA. we re-examined the study based on the clinical data provided in this article using machine learning methods. We used nine machine methods to construct the most suitable clinical prediction model and further demonstrated the importance of each parameter using SHAP technique. Finally we develop the Shiny application for complicated appendicitis diagnosis to assist clinicians in quickly and effectively recognizing patients with CA and UA.Unlike traditional medical statistical methods, machine learning techniques predict new observations by learning based on existing data. However, a significant problem with many state-of-the-art machine learning models is the need for more transparency and interpretability. To be able to interpret the results of predictions and judgments of machine learning models, explainable artificial intelligence (XAI) techniques are applied in clinical research, among which the SHAP technique is one of the methods of XAI, which determines values showing the direction and magnitude of the contribution of a variable to the estimation of an ML model and provides a visualization of the variable’s contribution30.In this study, we predicted UA and CA by ML modeling using patients’ clinical and biochemical examination indexes and interpreted the model results using SHAP technique. The main findings and contributions of this paper are as follows:(1) ML models were created to predict patients with UA and CA accurately.(2) The GBM model performed well in differentiating patients.(3) To explain the model, we utilized SHAP technique to show the importance of different parameters. And we have developed CA diagnosis Shiny application which helps clinicians to diagnose UA and CA.
