Development and validation of CT-based fusion model for preoperative prediction of invasion and lymph node metastasis in adenocarcinoma of esophagogastric junction

Abstract

Purpose: In the context of precision medicine, radiomics has become a key technology in solving medical problems. For adenocarcinoma of esophagogastric junction (AEG), developing a preoperative CT-based prediction model for AEG invasion and lymph node metastasis is crucial.

Methods: We retrospectively collected 256 patients with AEG from two centres. The radiomics features were extracted from the preoperative diagnostic CT images, and the feature selection method and machine learning method were applied to reduce the feature size and establish the predictive imaging features. By comparing the three machine learning methods, the best radiomics nomogram was selected, and the average AUC was obtained by 20 repeats of fivefold cross-validation for comparison. The fusion model was constructed by logistic regression combined with clinical factors. On this basis, ROC curve, calibration curve and decision curve of the fusion model are constructed.

Results: The predictive efficacy of fusion model for tumour invasion depth was higher than that of radiomics nomogram, with an AUC of 0.764 vs. 0.706 in the test set, P < 0.001, internal validation set 0.752 vs. 0.697, P < 0.001, and external validation set 0.756 vs. 0.687, P < 0.001, respectively. The predictive efficacy of the lymph node metastasis fusion model was higher than that of the radiomics nomogram, with an AUC of 0.809 vs. 0.732 in the test set, P < 0.001, internal validation set 0.841 vs. 0.718, P < 0.001, and external validation set 0.801 vs. 0.680, P < 0.001, respectively.

Conclusion: We have developed a fusion model combining radiomics and clinical risk factors, which is crucial for the accurate preoperative diagnosis and treatment of AEG, advancing precision medicine. It may also spark discussions on the imaging feature differences between AEG and GC (Gastric cancer).

Keywords: Adenocarcinoma of esophagogastric junction (AEG); Fusion model; Lymph node metastasis; Prediction; Radiomics nomogram; Serous invasion.

Fig. 1

Flow chart of image radiomics analysis. The intraclass correlation coefficient ICC (inter-rater agreement) was calculated to evaluate the consistency of measurement results to select stable radiomics features

Fig. 2

Development and validation of fusion model for predicting T stage. Fusion model of T stage prediction based on radiomics features and clinical factors (A). Calibration curve for predicting the depth of tumour invasion in training folds (B) and test folds (C), internal validation cohort (D) and external validation cohort (E). The diagonal line represents the performance of ideal model, and the red line represents the consistency between the built model and the actual model

Fig. 3

Comparison of ROC curves analysis of the model for predicting T stage in four cohorts. ROC curves of radiomics nomogram, clinic model and fusion model in training folds (A), test folds (B), internal validation cohort (C) and external validation cohort (D)

Fig. 4

Development and validation of fusion model for predicting N stage. Fusion model of N stage prediction based on radiomics features and clinical factors (A). Calibration curve for predicting the lymph node metastasis in training folds (B), test folds (C), internal validation cohort (D) and external validation cohort (E). The diagonal line represents the performance of ideal model, and the red line represents the consistency between the built model and the actual model

Fig. 5

Comparison of ROC curves analysis of the model for predicting N stage in four cohorts. ROC curves of radiomics nomogram, clinic model and fusion model in training folds (A), test folds (B), internal validation cohort (C) and external validation cohort (D)

Fig. 6

Decision curve analysis (DCA) of radiomics nomogram, clinic model and fusion model. DCA for predicting T stage (A) and N stage (B). The black line represents the net benefit of none of the patients receiving general treatment interventions; the gray line represents the net benefit of patients receiving general treatment intervention; the blue line represents the net benefit of patients receiving the fusion model intervention; the red line represents the net benefit for patients receiving the radiomics nomogram intervention. The blue line is above the red line, indicating that the fusion model has higher clinical efficacy