185
Views
3
CrossRef citations to date
0
Altmetric
ORIGINAL RESEARCH

Radiomics Based on Dynamic Contrast-Enhanced Magnetic Resonance Imaging in Preoperative Differentiation of Combined Hepatocellular-Cholangiocarcinoma from Hepatocellular Carcinoma: A Multi-Center Study

, , , ORCID Icon, & ORCID Icon
Pages 795-806 | Received 02 Feb 2023, Accepted 02 May 2023, Published online: 02 Jun 2023

Figures & data

Figure 1 The recruitment process of the cases for this study.

Abbreviations: cHCC-CC, combined hepatocellular carcinoma-cholangiocarcinoma; HCC, hepatocellular carcinoma; TACE, transarterial chemoembolization.
Figure 1 The recruitment process of the cases for this study.

Table 1 Sequences and Parameters of Dynamic-Enhanced MRI

Figure 2 The general flow of radiomics analysis in this study.

Figure 2 The general flow of radiomics analysis in this study.

Table 2 Comparison of Patient Characteristics of cHCC-CC and HCC in the Train and Test Set

Table 3 Performance of Each Model Based on Logistic Regression

Figure 3 Comparison of ROC curves in the test set of each model based on logistic regression.

Abbreviations: ROC, receiver operating characteristic; AUC, area under the curve.
Figure 3 Comparison of ROC curves in the test set of each model based on logistic regression.

Figure 4 Confusion matrix of the clinic-radiomics model in the test set based on logistic regression.

Abbreviations: cHCC-CC, combined hepatocellular carcinoma-cholangiocarcinoma; HCC, hepatocellular carcinoma.
Figure 4 Confusion matrix of the clinic-radiomics model in the test set based on logistic regression.

Figure 5 The global SHAP explanatory diagram shows the distribution of features’ influence on the clinic-radiomics model’s output. Red has a positive effect on the prediction and blue has a negative impact.

Abbreviations: SHAP, SHapley Additive exPlanations; CEA carcinoembryonic antigen; CA199, cancer antigens 19–9; HBeAg, hepatitis b e antigen, HBsAg, hepatitis b surface antigen; AP, arterial phase.
Figure 5 The global SHAP explanatory diagram shows the distribution of features’ influence on the clinic-radiomics model’s output. Red has a positive effect on the prediction and blue has a negative impact.

Figure 6 Distribution of importance for each feature in the fusion model based on AP. The value to the right of each red bar is the contribution coefficient of the feature to the model, which is the average value of the absolute value of the SHAP value for each feature.

Abbreviations: AP, arterial phase; SHAP, SHapley Additive exPlanations; CEA carcinoembryonic antigen; CA199, cancer antigens 19–9; HBeAg, hepatitis b e antigen, HBsAg, hepatitis b surface antigen.
Figure 6 Distribution of importance for each feature in the fusion model based on AP. The value to the right of each red bar is the contribution coefficient of the feature to the model, which is the average value of the absolute value of the SHAP value for each feature.

Figure 7 A single-sample SHAP interpretation plot that the model for HCC diagnosed.

Abbreviations: SHAP, SHapley Additive exPlanations; CA199, cancer antigens 19–9; AP, arterial phase.
Figure 7 A single-sample SHAP interpretation plot that the model for HCC diagnosed.