Table 7 Data-driven precision diagnosis in digestive diseases based on clinical data
From: Data-driven decision-making for precision diagnosis of digestive diseases
First author, year | Disease | n | Data source and specific task | ML method | Diagnostic performance | Refs. |
|---|---|---|---|---|---|---|
Li, 2021 | CRC | 1164 | Laboratory test data from electronic medical records; classification based on four laboratory indicators | LR/RF/KNN/SVM/Naïve Bayes | AUC: 0.849 (95%CI, 0.840–0.860) | [163] |
Hu, 2021 | GIST | 124 | Clinical examination data of pre-operation; classification based on CT and EUS features | XGBoost | AUC: 0.770 (95%CI, 0.570–0.900) | [164] |
Shung, 2020 | UGIB | 2357 | Clinical and laboratory indicators; classification based on variables of demography, comorbidity, clinical feature and laboratory | XGBoost | AUC: 0.90 (95%CI, 0.87 − 0.93) | [165] |
Wang, 2021 | Esophageal motility function | 229 | Esophageal HRM data sets; predicting esophageal motility function over HRM features | Conv3D/BiConvLSTM | Accuracy: 91.32% | [166] |
Zhu, 2020 | GC | 709 | Demographic and laboratory indicators from electronic medical records; classification based on a panel of independent predictors | GBDT | Accuracy: 83.0% | [167] |
Phan-Mai, 2023 | Complicated Appendicitis | 1950 | Medical record data; classification based on indicators of demography, blood test, and ultrasound of the appendix | SVM/DT/LR/KNN/ANN/GB | AUC: 0.64–0.89 | [168] |
Nemlander, 2023 | CRC | 2681 | PHC data; classification based on diseases diagnosed in PHC consultations and consultation number | SGB/LR | AUC: 0.830 (95%CI, 0.790–0.870) | [169] |
Popa, 2022 | EMD | 157 | Esophageal HRM images; classification based on the images | CNN | Accuracy: 93.0% | [170] |
Fan, 2022 | GC | 574 | Medical record data; classification based on age, sex and classical serum tumor markers | LR/RF | Accuracy: 86.8% | [171] |
Kou, 2022 | EMD | 1741 | Esophageal HRM data set; classification based on raw multi-swallow data | CNN/ANN/XGBoost/Bayes | Accuracy: 88.0–93.0% | [172] |
Ho, 2023 | EC | 819 | Questionnaire data from the SPIT and RISQ data sets; classification based on 17 features selected from questionnaire response | LDA/GLMNET/SVM/RF/KNN/CART/GLM | AUC: 0.710–0.920 | [173] |