Table 3 Characteristics of studies in artificial intelligence applications for appendicitis prognosis
Study, year | Objectives | Algorithms applied | Sample size (training: validating) | Age, mean ± SD (year) | Sex (male/female) |
|---|---|---|---|---|---|
Akbulut et al. [2], 2023 (Turkey) | Develop an ML model for the classification of AA and predicting perforated and non-perforated AA | CatBoost model | 1797 (80%/20%) (AA: 1465, NA: 332, AA non-perforated 1161, AA perforated: 304) | Male (median: 33; IQR: 23). female (median: 34; IQR: 26). AA (median: 33.1; IQR: 25), NA (median: 33; IQR: 24) | 993/804 |
Tuong-Anh Phan-Mai et al. [23], 2023 (Vietnam) | Develop and validate ML models for detecting CA | SVM, DT, KNN, LR, ANN, and GB | 1950 (483 CA, 1467 UCA) | Totally: 37:3 ± 15:9, CA: 40:6 ± 17:3, UCA: 36:2 ± 15:2 | Total: 678/826, CA: 233/250, UCA: 652/815 |
Li et al. [24], 2023 (China) | Develop a scoring system using clinical and imaging features to differentiate CA from UCA in pregnant individuals | LR, DT | 342 patients, 141 (41.23%) were diagnosed with CA, and 201 (58.77%) were diagnosed with UCA | Totally: 27.78 ± 3.36 CA: 27.87 ± 4.62 UCA: 27.72 ± 4.7 | All patients were female |
Lin et al. [25], 2023 (Taiwan) | Assess the ability of ANN models, to distinguish between UCA and CA | ANN (MLP) | 411 AA patients (288 (253 UCA, 35 CA): 123 (109 UCA, 14 CA) | Totally: 43.9 ± 17 Training set: 43.6 ± 16.5 Testing set: 44.8 ± 18.1 | Totally: 206/205 Training set: 144/144 Testing set: 62/61 |
Eickhof et al. [26], 2022 (Germany) | Create and validate an ML model for predicting postoperative outcomes of perforated appendicitis | RF classification is based on stratified under-sampling, i.e., an ensemble of DT | 163 (64 patients underwent laparoscopic surgery, 99 patients got an open procedure) | 38.1 ± 26.3 | 92/71 |
Xia et al. [27], 2022 (China) | Develop an accurate, rapid, noninvasive, and cost-effective diagnostic rule to differentiate between CA and UCA | OBLGOA-SVM, GOA-SVM, GS-SVM, RF, ELM, KELM, BPNN | 298 (150 UCA, 148 CA) | UCA: 42.23 ± 15.54, CA: 46.57 ± 19.73 | NR |
Kang et al. [28], 2021 (China) | Develop ML models to predict the pathological types of AA preoperatively | LR | 136 (acute SA = 8, acute PA = 104, acute GPA = 24). The sample size was divided 70/30 for training and testing SA/PA (112): training: 78, testing: 34. PA/GPA (128), training: 89, testing: 39 | SA: 39.12 ± 20.00, PA: 42.02 ± 17.31, GPA: 40.54 ± 15.31 | SA: 5/3, PA: 56/48, GPA: 12/11 |
Corinne Bunn et al. [29], 2021 (USA) | Apply different ML algorithms to predict the risk of postoperative sepsis after appendectomy, assess their effectiveness, and identify related risk factors | Multivariable LR, SVM, RFDT, and extreme gradient boosting (XGB) | 223,214 records for appendectomy (221,073 had no postoperative sepsis, 2141 had postoperative sepsis) | Postoperative sepsis: 48.09 ± 18.41 No postoperative sepsis: 39.8 ± 16.3 | Postoperative sepsis: 58.7%/41.3 No postoperative sepsis: 50.8%/49.2% |