Characteristics of included studies
Author | Year | Country | Study design | Study period | Study setting | Disease condition | Surgical procedure | Sample size | Predictive model | Internal validation | External validation | Discrimination | Calibration | Predictor |
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
Liang et al. [27] | 2019 | China | Retrospective | 2011–2017 | Single-center | Rectal cancer | Total mesorectal excision | 108 | Radiomics + LR | Cross-validation | No | AUC = 0.740 | No | 22 radiomic features |
Radiomics + SVM | AUC = 0.770 | |||||||||||||
Yan et al. [28] | 2019 | China | Retrospective | 2004–2015 | Single-center | Colon carcinoma | Not specified | 32,819 | Cox regression; Nomogram | Bootstrapping | No | AUC = 0.825 | Yes | Age; CEA; tumor size; tumor grade; N staging |
Li et al. [29] | 2019 | China | Retrospective | 2015–2018 | Single-center | Colon cancer | Radical colectomy | 48 | Clinical + SVM | Cross-validation | No | AUC = 0.690 | No | Age; sex; tumor location; tumor histology; tumor size |
Radiomics + SVM | AUC = 0.850 | 6 radiomic features | ||||||||||||
Hybrid + SVM | AUC = 0.870 | 6 clinical and radiomic features | ||||||||||||
Taghavi et al. [15] | 2021 | Netherlands | Retrospective | 2006–2016 | Multicenter | Colorectal cancer | Not specified | 91 | Clinical + RF | Bootstrapping; Cross-validation | No | AUC = 0.860 | No | Age; sex (male/female); primary tumor site; tumor stage; nodal stage; CEA; chemotherapy |
Radiomics + RF | AUC = 0.710 | 101 radiomic features | ||||||||||||
Combined + RF | AUC = 0.860 | 104 clinical and radiomic features | ||||||||||||
Lee et al. [30] | 2020 | South Korea | Retrospective | 2008–2013 | Single-center | Colorectal cancer | Colectomy | 2,019 | CNN + LR | Cross-validation | No | AUC = 0.747 | No | Age; sex (male/female); T stage; N stage; CT image features |
CNN + RF | AUC = 0.697 | |||||||||||||
Xiao et al. [31] | 2022 | China | Retrospective | 2016–2017 | Single-center | Colorectal cancer | Radical colorectal resection | 611 | CNN | Cross-validation | No | AUC = 0.758 | Yes | Digital pathological images |
Cox regression; Nomogram | Bootstrapping | AUC = 0.848 | Digital pathological images; N stage; T stage; VE/LI/PI | |||||||||||
Hao et al. [32] | 2022 | China | Retrospective | 2016–2019 | Single-center | Colorectal cancer | Not specified | 293 | LR; Nomogram | Bootstrapping | No | AUC = 0.784 | Yes | Age; CEA; VI; T stage; N stage; family CRCLM history; KRAS |
AUC, area under the receiver operating characteristic curve; CEA, carcinoembryonic antigen; CNN, convolutional neural network; CRCLM, colorectal cancer liver metastasis; KRAS, Kirsten rat sarcoma viral oncogene homologue; LI, lymphatic invasion; LR, logistic regression; PI, perineurial invasion; RF, random forest; SVM, support vector machine; VE, vascular emboli; VI, vascular invasion.