Minireviews
Copyright ©The Author(s) 2021.
World J Gastroenterol. Oct 28, 2021; 27(40): 6825-6843
Published online Oct 28, 2021. doi: 10.3748/wjg.v27.i40.6825
Table 2 Applications of artificial intelligence classification methods using liver magnetic resonance imaging
Ref.
Task
Method
MR image
Accuracy
Sensitivity
Specificity
AUROC
Hectors et al[60], 2020Stage liver fibrosisVGG16 CNNGadoxetic acid-enhanced MRIF1-4: 0.69F1-4: 0.64F1-4: 0.90 F1-4: 0.77
F2-4: 0.85 F2-4: 0.82F2-4: 0.93F2-4: 0.91
F3-4: 0.85F3-4: 0.87F3-4: 0.83F3-4: 0.90
F4: 0.78F4: 0.73F4: 0.81F4: 0.85
Liu et al[55], 2021Classify cHCC-CC vs non-cHCC-CC and HCC vs non-HCCRadiomics + SVMGadoxetic acid-enhanced MRIcHCC-CC vs non-cHCC-CC: 0.77 cHCC-CC vs non-cHCC-CC: 0.65cHCC-CC vs non-cHCC-CC: 0.81 cHCC-CC vs non-cHCC-CC: 0.77
HCC vs non-HCC: - HCC vs non-HCC: 0.68 HCC vs non-HCC: 0.88HCC vs non-HCC: 0.79
Wu et al[48], 2020Classify tumours according to their LI-RADS gradeAlexNet CNNGadoxetic acid-enhanced MRI0.910.8350.95
Messaoudi et al[50], 2020Classify tumours into HCC or non-HCCPatch based CNNMultiphase 3D fast spoiled gradient echo T10.9???
Hamm et al[51], 2019Classify tumours into type and LI-RADS derived classesCNNMultiphase contrast-enhanced T1-weighted MRILesion class: 0.919Lesion class: 0.90Lesion class: 0.98LI-RADS (HCC): 0.922
LI-RADS: 0.943LI-RADS: 0.92LI-RADS: 0.97
Trivizakis et al[54], 2018Classify tumours into primary or metastatic3D CNN + SVMDiffusion weighted MRI0.830.930.670.8
He et al[65], 2019Correctly predict liver stiffness using clinical and radiomic dataRadiomics + SVMT2-weighted MRI0.8180.7220.870.84
Schawkat et al[61], 2020Stage liver fibrosis into low-stage (F0-2) and high-stage (F3-4)Radiomics + SVMT1-weighted MRI, T2-weighted MRIT1-w: 0.857 ??T1-w: 0.82
T2-w: 0.57
T2-w: 0.619
Lewis et al[56], 2019Distinguish HCC from other primary cancersRadiomics + Binary logistic regressionDiffusion weighted MRIObserver 1: 0.815Observer 1: 0.793 Observer 1: 0.889 Observer 1: 0.90
Observer 2: 0.80Observer 2: 0.862Observer 2: 0.778Observer 2: 0.89
Wu et al[57], 2019Classify tumours into HCC and HHRadiomics + logistic regressionT2-weighted MRI, Diffusion weighted MRI, T1-weighted GRE in phase and out of phase MRI?0.8220.7140.89
Oyama et al[58], 2019Classification of hepatic tumours into HCC, HH and MTRadiomics + logistic regression/XGBoostT1-weighted MRIHCC vs MT: 0.92HCC vs MT: 1.0 HCC vs MT: 0.84 HCC vs MT: 0.95
HCC vs HH: 0.9 HCC vs HH: 0.96 HCC vs HH: 0.84HCC vs HH: 0.95
MT vs HH: 0.73MT vs HH: 0.72MT vs HH: 0.74MT vs HH: 0.75
Wu et al[59], 2019Predict pre-operative HCC gradeCombined clinical data and Radiomics + logistic regressionT2/T1-weighted0.7610.850.650.8
Chen et al[69], 2019Predict pre-treatment immunscore in HCCCombined clinical data and radiomics + multi-vote decision treesGadoxetic acid-enhanced MRI0.8420.8460.8410.934
Park et al[63], 2019Stage liver fibrosisRadiomics + logistic regressionGadoxetic acid-enhanced MRIF2-4: 0.803 F2-4: 0.814F2-4: 0.784F2-4: 0.91
F3-4: 0.803F3-4: 0.789F3-4: 0.820 F3-4: 0.88
F4: 0.813F4: 0.921F4: 0.754F4: 0.87
Zhao et al[67], 2019Predict early reoccurrence of IMCCCombined clinical data and radiomics + logistic regressionT2-weighted MRI, gadoxetic acid-enhanced MRI0.8720.9380.8390.949
Reimer et al[68], 2018Predict therapy response to transarterial radioembolizationRadiomics + logistic regressionGadoxetic acid-enhanced MRI?Arterial phase: 0.83Arterial phase: 0.62Arterial phase: 0.73
Venous phase: 0.71Venous phase: 0.85Venous phase: 0.76
Zhen et al[53], 2020Classify liver tumours into benign , HCC, metastatic or other primary malignancyCNN with clinical inputT2, diffusion, Pre- contrast T1, late arterial, portal venous, and equilibrium phase0.919HCC: 0.957 HCC: 0.904 HCC: 0.951
Metastatic: 0.946Metastatic: 1.0Metastatic: 0.985
Other primary: 0.733Other primary: 0.964Other primary: 0.989
Yasaka et al[62], 2017Stage liver fibrosisCNNGadoxetic acid-enhanced MRIF4 vs F3-0: 0.75 F4 vs F3-0: 0.76F4 vs F3-0: 0.76 F4 vs F3-0: 0.84
F4-3 vs F2-0: 0.77 F4-3 vs F2-0: 0.78F4-3 vs F2-0: 0.74F4-3 vs F2-0: 0.84
F4-2 vs F1-0: 0.80F4-2 vs F1-0: 0.84F4-2 vs F1-0: 0.65F4-2 vs F1-0: 0.84
Kim et al[70], 2019Predict postoperative early and late recurrence of single HCCRadiomics + random forestsGadoxetic acid-enhanced MRIHarrel C-statistic: 0.716 in combined radiomic and clinicopathologic model, no significant difference to clinicopathologic model (0.696)
Kim et al[52], 2020Detect HCCCNNGadoxetic acid-enhanced MRI0.9370.940.990.97
Liu et al[71], 2020Identify clinically significant portal hypertensionCNN + logistic regression??0.9290.8460.94