Machine learning models from computed tomography to diagnose thymic epithelial tumors requiring combined resection

Abstract

Background: Minimally invasive approaches have been a standard choice of surgery for noninvasive thymic epithelial tumors (TETs), but we sometimes experience cases requiring combined resection of adjacent structures. We develop and validate machine learning models to predict combined resection based on preoperative contrast-enhanced computed tomography (CT).

Methods: This study included 212 patients with TETs (140 in the training cohort and 72 in the validation cohort) who underwent radical surgery. Radiomics features were extracted from contrast-enhanced CT and predicted with five feature selection methods and seven machine learning models in nested cross validation. The clinical utility of the models was analyzed by a decision curve analysis (DCA).

Results: Fifty-five patients in the training cohort and 28 in the validation cohort required combined resection. The classifiers random forest (RF), gradient boosting (GB), and eXtreme Gradient Boosting (XGB) indicated high predictive performance, with the XGB classifier based on features selected by GB performing the best, with an area under the curve (AUC) of 0.797. In the validation cohort, the classifier had an AUC of 0.817. The DCA showed the validity of the model with a threshold range of 15-72%. When restricted to combined pulmonary and pericardial resection, the respective AUCs were 0.736 and 0.674 for the training cohort and 0.806 and 0.924 for the validation cohort.

Conclusions: The machine learning model based on preoperative CT images was able to diagnose TETs requiring combined resection with high accuracy. The DCA demonstrated a wide range of model validity and may aid in surgical approach selection.

Keywords: Thymic epithelial tumor (TET); computed tomography (CT); machine learning; radiomics; surgical procedure.

Figure 1

The flow chart of patient selection. (A) For the training cohort, a total of 288 patients underwent thymectomy in CUH, and 140 met the criteria. (B,C) For the validation cohort, a total of 122 patients underwent thymectomy in CCC and KCH, and 72 met the criteria. CUH, Chiba University Hospital; CT, computed tomography; CCC, Chiba Cancer Center; KCH, Kimitsu Chuo Hospital.

Figure 2

The prediction results for TCR. (A) ROC curves show the results of each machine learning model with the best performing feature selection. (B) The results of predicting TCR in the validation cohort. (C) Combining the training and validation cohorts and showing how many models predicted as TCR when the cut-off of the machine learning model was set to 0.5. (D) Results of a DCA based on the predicted probability of TCR for all patients. LR, logistic regression; SVM, support vector machine; KNN, k-nearest neighbor; NB, Naïve Bayes; RF, random forest; GB, gradient boosting; XGB, eXtreme Gradient Boosting; TCR, tumors requiring combined resection of adjacent structure; ROC, receiver operating characteristic; DCA, decision curve analysis.

Figure 3

Representative cases. (A) Case 1 is an 84-year-old woman. The tumor was 7.5 cm in diameter with a smooth surface and uniform interior. Thymectomy was performed through a median sternotomy approach, but no adhesion or invasion was observed. The predicted probability of TCR was 6.2% for XGB. (B) Case 2 is a 72-year-old man. The tumor was 7.0 cm with irregular margins and internal calcification. A median sternotomy approach was initiated, but the tumor was firmly adhered to or had infiltrated the aorta, brachiocephalic vein, and pulmonary artery. Therefore, additional left fourth intercostal thoracotomy was performed. The brachiocephalic vein and transverse nerves had to be reconstructed. The predicted probability of TCR was 98.2% by XGB. LR, logistic regression; SVM, support vector machine; KNN, k-nearest neighbor; NB, Naïve Bayes; RF, random forest; GB, gradient boosting; XGB, eXtreme Gradient Boosting; TCR, tumors requiring combined resection of adjacent structure.

Figure 4

The prediction results for combined resection of lung and pericardium. (A) ROC curves show the results of the GB feature selector and XGB classifier for lung and pericardium. (B) The results of prediction in the validation cohort. (C) Results of a DCA based on the predicted probability of lung. (D) Results of a DCA based on the predicted probability of pericardium. XGB, eXtreme Gradient Boosting; ROC, receiver operating characteristic; GB, gradient boosting; DCA, decision curve analysis.