Radiomics-based nomogram as predictive model for prognosis of hepatocellular carcinoma with portal vein tumor thrombosis receiving radiotherapy

Abstract

Background: This study aims to establish and validate a predictive model based on radiomics features, clinical features, and radiation therapy (RT) dosimetric parameters for overall survival (OS) in hepatocellular carcinoma (HCC) patients treated with RT for portal vein tumor thrombosis (PVTT).

Methods: We retrospectively reviewed 131 patients. Patients were randomly divided into the training (n = 105) and validation (n = 26) cohorts. The clinical target volume was contoured on pre-RT computed tomography images and 48 textural features were extracted. The least absolute shrinkage and selection operator regression was used to determine the radiomics score (rad-score). A nomogram based on rad-score, clinical features, and dosimetric parameters was developed using the results of multivariate regression analysis. The predictive nomogram was evaluated using Harrell's concordance index (C-index), area under the curve (AUC), and calibration curve.

Results: Two radiomics features were extracted to calculate the rad-score for the prediction of OS. The radiomics-based nomogram had better performance than the clinical nomogram for the prediction of OS, with a C-index of 0.73 (95% CI, 0.67-0.79) and an AUC of 0.71 (95% CI, 0.62-0.79). The predictive accuracy was assessed by a calibration curve.

Conclusion: The radiomics-based predictive model significantly improved OS prediction in HCC patients treated with RT for PVTT.

Keywords: hepatocellular carcinoma; portal vein tumor thrombosis; predictive model; radiation therapy; radiomics.

Figure 1

Study workflow. The region of interest (ROI) was segmented on all transverse contrast-enhanced computed tomography images by experienced radiation oncologists using the Eclipse system. After a three-dimensional reconstruction of the ROI, 48 textural features, including conventional features, histogram features, gray-level co-occurrence matrix, gray-level run-length matrix, neighborhood gray-level dependence matrix, and gray-level zone length matrix, were extracted. The extracted features were selected by least absolute shrinkage and selection operator regression. Based on the selected radiomics features, clinical features, and radiation therapy dosimetric parameters, a nomogram model was established to predict overall survival. The performance of the predictive model was evaluated with concordance index, area under the curve of the receiver operating characteristic curve, and calibration curve.

Figure 2

Radiation therapy plan for a patient. The prescribed dose to treat portal vein tumor thrombosis only is 50 Gy. The clinical target volume as the region of interest is contoured in red, and the volume is 280.7 cm³. The normal liver volume is 1,769.0 cm³, and the mean liver dose is 2,189.4 cGy.

Figure 3

Least absolute shrinkage and selection operator (LASSO) regression analysis for the selection of significant radiomics features from the 48 textural features. (A) Coefficient profile of the LASSO model. (B) Optimal tuning parameter (lambda) selection using 10-fold cross-validation with minimum criteria. Two significant radiomics features were extracted.

Figure 4

Survival curves of the high- and low-risk groups based on the radiomics score (rad-score) classification. The rad-scores in the high- and low-risk groups were more than −0.6 and less than −0.6, respectively. (A) Considering all patients, the median overall survival (OS) rates in the high- and low-risk groups were 7.4 months (95% CI, 6.5–10.7) and 12.4 months (95% CI, 10.0–16.8), respectively (p = 0.007). (B) Considering the training cohort, the median OS rates in the high- and low-risk groups were 7.5 months (95% CI, 6.5–11.2) and 11.8 months (95% CI, 9.6–16.8), respectively (p = 0.038). (C) Considering the validation cohort, the median OS rates in the high- and low-risk groups were 6.8 months (95% CI, 4.3–NA) and 12.6 months (95% CI, 10.7–NA), respectively (p = 0.033).

Figure 5

Nomograms for the prediction of overall survival. Nomograms with radiomics score, significant clinical features, and radiation therapy (RT) dosimetric parameters for the prediction of (A) median survival time and (B) 6-, 9-, and 12-month survival rates. Nomograms with selected clinical features and RT dosimetric parameters for the prediction of (C) median survival time and (D) 6-, 9-, and 12-month survival rates.

Figure 6

Receiver operating characteristic curves of different predictive nomograms for 9-month survival. (A) The area under the curve (AUC) was 0.71 (95% CI, 0.62–0.79) in the radiomics-based nomogram. (B) The AUC was 0.61 (95% CI, 0.51–0.71) in the clinical nomogram.

Figure 7

Calibration curves of (A) the radiomics-based nomogram and (B) clinical nomogram for the prediction of 9-month survival.