Prediction of Ki-67 expression in hepatocellular carcinoma with machine learning models based on intratumoral and peritumoral radiomic features

Figure 1 Flowchart for the inclusion and exclusion of hepatocellular carcinoma patients in this study. HCC: Hepatocellular carcinoma.

Figure 2 Computed tomography images before and after superresolution reconstruction. A: Image from a normal resolution; B: Image after superresolution reconstruction.

Figure 3 Examples of regions of interest delineated on computed tomography images with ITK-SNAP software. A: Original image; B: Intratumoral regions of interest (ROI); C: Peritumoral ROI; D: Fusion ROI (intratumoral tissue + peritumoral tissue).

Figure 4 The receiver operating characteristic curves were used to compare the intratumoral model, peritumoral model, and fusion model from the validation cohort. AUC: Area under the receiver operating characteristic curve; CI: Confidence interval.

Figure 5 Waterfall plot showing the model performance across the validation sets of the three models. A: Intratumoral model; B: Peritumoral model; C: Fusion model. Each bar in the plot represents the difference between the model’s predicted value and the established cut-off value. The bars extending above the Y = 0 line indicate model predictions of high Ki-67 expression, whereas the bars below the Y = 0 line signify predictions of low Ki-67 expression.

Figure 6 The receiver operating characteristic curve shows the performance of the different models in predicting Ki-67 expression. A: Receiver operating characteristic curve comparison of the training group for the three models; B: Receiver operating characteristic curve comparison of the validation group for the three models. AUC: Area under the receiver operating characteristic curve; CI: Confidence interval.

Figure 7 Clinical-radiomic nomogram. ALT: Alanine aminotransferase.

Figure 8 Decision curve analysis and calibration curves of different models. A: Calibration curves for three models (training group); B: Calibration curves for three models (validation group); C: Decision curve analysis for three models (training group); D: Decision curve analysis for three models (validation group).