CT-based radiomics features analysis for predicting the risk of anterior mediastinal lesions

Abstract

Background: To retrospectively validate CT-based radiomics features for predicting the risk of anterior mediastinal lesions.

Methods: A retrospective study was performed through February 2013 to March 2018 on 298 patients who had pathologically confirmed anterior mediastinal lesions. The patients all underwent CT scans before their treatment, including 130 unenhanced computed tomography (UECT) and 168 contrast-enhanced CT (CECT) scans. The lesion areas were delineated, and a total of 1,029 radiomics features were extracted. The least absolute shrinkage and selection operator (Lasso) algorithm method was used to select the radiomics features significantly associated with discrimination of high-risk from low-risk lesions in the anterior mediastinum. Then, 8-fold and 3-fold cross-validation logistic regression (LR) models were taken as the feature selection classifiers to build the radiomics models for UECT and CECT scan respectively. The predictive performance of the radiomics features was evaluated based on the receiver operating characteristics (ROC) curve.

Results: Each of the two radiomics classifiers included the optimal 12 radiomic features. In terms of the area under ROC curve, using the radiomics model in discriminating high-risk lesions from the low-risks, CECT images accounted for 74.1% with a sensitivity of 66.67% and specificity of 64.81%. Meanwhile, UECT images were 84.2% with a sensitivity of 71.43% and specificity of 74.07%.

Conclusions: The association of the two proposed CT-based radiomics features with the discrimination of high and low-risk lesions in anterior mediastinum was confirmed, and the radiomics features of the UECT scan were proven to have better prediction performance than the CECT's in risk grading.

Keywords: Radiomics; anterior mediastinum; contrast-enhanced computed tomography (CECT); histologic grade; unenhanced computed tomography (UECT).

Figure 1

Examples of anterior mediastinal lesion segmentation. (A) Venous-phase CECT images. A 56-year-old woman with thymic cyst; (B) UECT images. A 67-year-old woman with WHO type AB thymoma. UECT, unenhanced computed tomography; CECT, contrast-enhanced computed tomography.

Figure 2

Radiomics analysis workflow. First, the clinical UECT (A) and CECT (B) images of high- and low-risk anterior mediastinal lesions were collected. Second, image segmentation was used to delineate the anterior mediastinal lesions. Next, the image features were extracted by the automated high-throughput feature analysis algorithm. Finally, the statistical analysis was applied, and the sequential forward search was used for feature selection for the classification of anterior mediastinal lesions. UECT, unenhanced computed tomography; CECT, contrast-enhanced computed tomography.

Figure 3

Radiomics features selection of UECT using the LASSO regression. (A) Mean square error path diagram. The abscissa is log (alpha), and the dashed lines of different colors indicate that different features correspond to different alphas with different mean square errors; (B) LASSO path map, features corresponding to different alpha features. UECT, unenhanced computed tomography.

Figure 4

Radiomics features selection of CECT using the LASSO regression. (A) Mean square error path diagram. The abscissa is log (alpha), and the dashed lines of different colors indicate that different features correspond to different alphas with different mean square errors; (B) LASSO path map, features corresponding to different alpha features. CECT, contrast-enhanced computed tomography.

Figure 5

Receiver operating characteristic curve on UECT. (A) The 12-feature model was trained in training set with 8-fold cross-validation LR classifier in UECT. The AUC was 0.842 (95% CI: 0.763–0.922; sensitivity 71.43%; specificity 74.07%); (B) the 12-feature model of The UECT validation set with the AUC of 0.774 (95% CI: 0.632–0.916; sensitivity 74.36%; specificity 72.73%). UECT, unenhanced computed tomography.

Figure 6

Receiver operating characteristic curve on CECT. (A) The 12-feature model was trained in training set with 3-fold cross-validation LR classifier in UECT. The AUC was 0.741 (95% CI: 0.628–0.814; sensitivity 66.67%; specificity 64.81%); (B) the 12-feature model of The UECT validation set with the AUC of 0.734 (95% CI: 0.597–0.872; sensitivity 62.96%; specificity 66.67%). UECT, unenhanced computed tomography; CECT, contrast-enhanced computed tomography.