Radiomics to predict PNI in ESCC

Abstract

Objective: This study aimed to investigate whether contrast-enhanced computed tomography (CECT) based radiomics analysis could noninvasively predict the perineural invasion (PNI) in esophageal squamous cell carcinoma (ESCC).

Methods: 398 patients with ESCC who underwent resection between February 2016 and March 2020 were retrospectively enrolled in this study. Patients were randomly divided into training and testing cohorts in a 7:3 ratio. Radiomics analysis was performed on the arterial phase images of CECT scans. From these images, 1595 radiomics features were initially extracted. The intraclass correlation coefficient (ICC), wilcoxon rank-sum test, spearman correlation analysis, and boruta algorithm were used for feature selection. Logistic regression (LR), random forest (RF), and support vector machine (SVM) models were established to preidict the PNI status. The performance of these radiomics models was assessed by the area under the receiver operating characteristic curve (AUC). Decision curve analysis (DCA) was conducted to evaluate their clinical utility.

Results: Six radiomics features were retained to build the radiomics models. Among these models, the random forest (RF) model demonstrated superior performance. In the training cohort, the AUC value of the RF model was 0.773, compared to 0.627 for the logistic regression (LR) model and 0.712 for the support vector machine (SVM) model. Similarly, in the testing cohort, the RF model achieved an AUC value of 0.767, outperforming the LR model at 0.638 and the SVM model at 0.683. Decision curve analysis (DCA) suggested that the RF radiomics model exhibited the highest clinical utility.

Conclusions: CECT-based radiomics analysis, particularly utilizing the RF, can noninvasively predict the PNI in ESCC preoperatively. This novel approach could enhance patient management by providing personalized information, thereby facilitating the development of individualized treatment strategies for ESCC patients.

Keywords: Computed tomography; Contrast-enhanced; Esophageal squamous cell carcinoma; Perineural invasion; Radiomics.

Fig. 1

The flowchart illustrating the patient selection and exclusion in our study

Fig. 2

Representation segmentations of a PNI-positive ESCC in a 70-year-old male (a-d) and a PNI-negative ESCC in a 60-year-old female (f-i). Axial (a, f), Sagittal (b, g), and Coronal (c, h) images show tumor delineation using 3Dslicer software. (d, i) 3D regions of interest (ROIs) after refinement. Corresponding histopathological slices of the PNI-positive (e) and PNI- negative (j) tumors, H&E, ×20

Fig. 3

The general workflow of radiomics analysis

Fig. 4

The retained features and corresponding importance degree. The length of the blue bars corresponds to the importance degree of each radiomics feature, indicating their significance in predicting the PNI status

Fig. 5

Radiomics heat map illustrates the correlation coefficients among the retained 6 radiomics features in both the training cohort (a) and testing (b) cohort. The color scale on the right indicates the correlation coefficient (-1 to 1, ranging from blue to red)

Fig. 6

ROC curves of radiomics models for the training cohort (a) and testing cohort (b). AUCs obtained from the RF-radiomics model were higher than LR-radiomics model and SVM-radiomics model in the training cohort and testing cohort

Fig. 7

Calibration curves for the radiomics models of the training cohort (a) and testing cohort (b). The Y-axis represents the actual probability, while the X-axis indicates the predicted probability

Fig. 8

DCA for the training cohort (a) and testing cohort (b). The black horizontal line represents no patients as PNI-positive, and the grey line represents all patients as PNI-positive. The colored lines for each model respectively illustrate the net benefit for each patient based on different radiomics models. The closer the decision curves are to the black and gray curves, the lower the clinical decision net benefit of the model. The threshold probability is where the expected benefit of treatment is equal to the expected benefit of avoiding treatment. The RF-radiomics model had a higher net benefit compared to the other two radiomics models within a wide range of risk thresholds