Ethical approval was obtained from the local Ethics Committee for this retrospective study, and informed consent was waived. This study was conducted in accordance with the Declaration of Helsinki.

From August 2015 to March 2021, 127 patients with TET underwent thorax CT examinations were included according to the following criteria: (a) TETs confirmed by pathology; (b) both non-contrast and contrast-enhanced CT images were available; (c) lesions larger than 1 cm in diameter based on the longest diameter; (d) the patient did not undergo biopsy or any treatment before CT examinations. In addition, a total of 5 cases were excluded according to the following exclusion criteria: (a) severe CT image artifact (n=3); (b) interval between CT and surgery was greater than two weeks (n=1); (c) exact histological subtype was unknown based on the analysis of puncture biopsy specimen (n=1). Thus, the final study group comprised 122 patients (67 male, 55 female; mean age, 52 years; age range, 18-78 years) ( Figure 1 ). Demographic and clinical data were retrieved from electronic medical records.

All thorax CT examinations were performed on 128-row CT scanner (Somatom Definition Flash, Siemens Healthcare, Forchheim, Germany), 256-row CT (Brilliance iCT, Philips Healthcare, Cleveland, USA), or 64-row CT (GE-Medical-Systems, Milwaukee, USA). Before the scan, the patient was instructed in breathing training. First, a non-enhanced thorax CT (210 mAs, 120 kV; slice thickness and interval, 5.0 mm) was performed. CT images were collected during a breath-holding period, and the scanning range was from the thoracic entrance to the costophrenic angle level in the cranio-caudal direction. Secondly, contrast medium was administered using a dual-head pump injector. A volume of 60-100 ml (1.2 ml/kg of body weight) iodixanol injection 320 (HengRui, JiangSu, China) was injected in the forearm vein at a flow rate of 3 ml/sec using a 20-G needle followed by a saline flush of 30 ml at the same rate. After intravenous injection of the contrast medium, the arterial and venous phase scan were performed, respectively. The scan parameters were as same as those of non-enhanced CT. The original data obtained from the scanning were reconstructed into axial, coronal, and sagittal images and then transmitted to the dedicated workstation.

CT features of the tumors were analyzed individually by two experienced radiologists (Y.-H.M. and W.-Q.Y., with 7 and 12 years of experience in chest imaging, respectively) who knew the patient had TETs but were blinded to the histological subtypes of the tumors. The disagreements between the two readers were resolved by negotiation with another senior radiologist (Y.-C.H., with 18 years of experience in chest imaging) until a consensus was reached.

Evaluation of tumor conventional CT features refers to the standard report terms for thymoma defined by the International Thymic Malignancy Interest Group (ITMIG) (14). The recorded CT features include tumor location (off-midline and midline), maximum diameter, contour (smooth and lobulated), homogeneity (almost homogenous and heterogeneous), enhancement degree, infiltration of adjacent pericardium or pleura, pericardiac or pleural effusion, and lymphadenopathy (short-axis diameter >10 mm). The maximum diameter of the tumor was measured at the level where the tumor appeared largest on the cross-sectional image. The enhancement degree was divided into mild to moderate [net enhanced value ≤ 40 Hounsfield Unit(HU)] and strong (net enhanced value > 40 HU) (16). Lung invasion was considered based on signs: a multilobular tumor convex to the lung with adjacent lung abnormalities, or deep lobulation at the tumor-lung interface (17, 18). The pericardiac/pleural invasion was considered when the space between the tumor and the pericardiac/pleura disappeared, with pericardiac/pleura thickening and/or cavity effusion (17).

The evaluated major mediastinal vessels included the aorta, branches of the aortic arch, pulmonary trunk, left and right pulmonary artery, SVC, innominate vein, and hilar pulmonary veins. According to the extent of tumor abutment with the adjacent vessel, the arterial invasion was divided into the following three grades: < 25%, 25-49%, and ≥ 50% ( Figure 2 ); the venous invasion was divided into two grades: < 50% and ≥ 50% ( Figure 3 ).

The final diagnosis was determined by surgical or puncture biopsy specimen and confirmed with histopathological examination. Based on the criteria of the WHO histological classification and Jeong simplification classification (5, 19), TET was divided into three subgroups: low risk thymoma (types A, AB, and B1), high risk thymoma (types B2 and B3) and thymic carcinoma. TETs were divided into stages I-IV according to the MK or TNM staging system (7, 9, 20).

The Pearson chi-square test or Fisher’s exact test was used to analyze categorical variables among defined groups. The Bonferroni method was used for pairwise comparison between groups. Multivariate logistic regression analysis was used to identify the risk factors associated with TET vascular invasion. Cohen’s kappa coefficient with a 95% confidence interval was adopted to assess the inter-reader agreement level in evaluating the degree of tumors abutment of adjacent vessels circumference; kappa values were interpreted as follows: poor, 0–0.2; fair, 0.21–0.4; moderate, 0.41–0.6; good, 0.61–0.8; and excellent, 0.81–1.0. Using the clinical and CT features as independent variables, and defined group (vessel invasion grade) as dependent variables, binary logistic regression analysis was conducted to establish a regression model. The receiver operating characteristic (ROC) curve analysis was performed for each statistically significant feature, and the predicted probability value of the regression model, and the area under the curve (AUC), sensitivity, and specificity were obtained. All statistical analyses above were performed with IBM SPSS 26 software (IBM Corp). Statistical significance was accepted as P < 0.05.

The demographic characteristics of the patients are summarized in Table 1 . A total of 122 patients (67 males, 55 females) with a mean age of 52 years were retrospectively entered into the study. Among these patients, 38 patients (31.1%) had been diagnosed with myasthenia gravis, and other clinical features included chest pain (17.2%, 21 of 122), respiratory symptoms (13.1%, 16 of 122), others (13.9%, 17 of 122), and no symptom in 30 patients (24.6%).

The pathological results were proved on the specimen, obtained by thoracotomy in 35 patients (28.7%), thoracoscopy in 74 patients (60.7%), and percutaneous biopsy in 13 patients (10.6%). After surgery and pathological analysis, 77 patients (63.1%) were found to be in TNM stage I or II, 45 (36.9%) in stage III or IV, and 71 patients (58.2%) in MK stage I or II, 51 (41.8%) in stage III or IV. Thirteen masses did not undergo surgery, and tumor stage was evaluated by puncture biopsy and imaging. Histological analysis revealed 106 (86.9%) thymomas (13 of type A, 33 of type AB, 10 of type B1, 32 of type B2, and 18 of type B3), 13 thymic squamous cell carcinoma, and 3 thymic neuroendocrine tumors ( Table 1 and Supplemental Table 1 ).

Comparisons of vascular abutment among groups with different types and stages of TETs are shown in Table 2 and Figure 4 . Overall, there were significant differences in the abutment rate of major mediastinal vessels among low-, high-risk thymomas, and thymic carcinomas, and between early and advanced stages of TETs (all P < 0.01).

For the aorta and its branches, 6 cases (12%) of high-risk thymoma and 5 cases (31.3%) of thymic carcinoma, or 11 cases of advanced TETs, had vascular abutment ≥ 50% of the vascular circumference. For the pulmonary artery and its branches, 3 cases (6%) of high-risk thymoma and 5 cases (31.3%) of thymic carcinoma, or 8 cases of advanced TETs, had vascular abutment ≥ 50% of the vascular circumference. As for SVC or innominate vein, the abutment rate differed among different types or stages (P < 0.001), with the highest abutment rate being thymic carcinoma (9/16, 56.3%), followed by advanced TETs, and high-risk thymoma (12/50, 24.0%). The abutment rate of the pulmonary veins was relatively low, with vascular abutment ≥ 50% of the vascular circumference in only 4 cases (8%) of high-risk thymoma and 3 cases (18.8%) of thymic carcinoma, or 7 cases of advanced TETs.

In addition, whether it is an artery or a vein, ≥ 50% of vascular abutment was mainly seen in stages III and IV, with significant statistical significance compared to stage I and II (P < 0.0167) based on TNM stages, except for pulmonary vein invasion among stage I-II and stages III (P > 0.0167); However, there was no significant statistical significance between the III and IV stages of vascular abutment with ≥ 50% (P > 0.0167).

The association between clinical and CT features and artery invasion grades of TETs is shown in Table 3 . Age, myasthenia gravis, and tumor location were not statistically different among the three groups (P > 0.05). It showed a higher rate of arterial invasion in male patients and tumors with a maximum diameter larger than 5.9 cm, lobulated contours, heterogeneous density, mild to moderate degree of enhancement, pericardial or pleural invasion, adjacent lung invasion, pericardial or pleural effusion, and mediastinal lymphadenopathy (P<0.05). The further comparison between every two groups showed that maximum diameter larger than 5.9 cm, heterogeneous density, and pericardial or pleural effusion were more common in grade III than in grades I and II.

Patients were divided into two groups according to the invasion degree of the SVC or innominate vein. Factors associated with greater than or equal to 50% venous invasion included male patients, tumors with a maximum diameter larger than 5.9 cm, lobulated contours, heterogeneous density, pericardial or pleural invasion, adjacent lung invasion, pericardial or pleural effusion, and mediastinal lymphadenopathy (P < 0.05) ( Table 4 ). Age, myasthenia gravis, tumor location, and degree of enhancement were not statistically different among the two groups (P > 0.05).

All indicators, including clinical and CT features, were included in the multivariable logistic regression analysis ( Table 5 ). For predicting 25% invasion of the aorta or pulmonary artery and its branches, the results showed that only the male patients (β = 1.549; odds ratio, 4.824; P = 0.007) and the pericardial or pleural invasion (β = 2.209; odds ratio, 9.110; P = 0.001) were independent predictors. As for 50% invasion of aorta or pulmonary artery and its branches, the midline location (β = 2.504; odds ratio, 12.234; P = 0.01) and mediastinal lymphadenopathy (β = 2.490; odds ratio, 12.06; P = 0.008) were independent predictors. Risk factors associated with ≥50% invasion of the SVC or innominate vein included midline location (β = 2.303; odds ratio, 10.0; P = 0.008), maximum tumor diameter larger than 5.9 cm (β = 4.038; odds ratio, 56.739; P = 0.001), and pericardial or pleural effusion (β = 1.460; odds ratio, 4.306; P = 0.033).

Based on the predictive probability values derived from the logistic regression model and univariate features for predicting 50% of great mediastinal vascular invasion in TET patients, ROC analysis was performed. The multivariate model achieved a higher diagnostic efficacy with an AUC of 0.944, 84.6% sensitivity, and 91.7% specificity for predicting 50% of artery invasion, and an AUC of 0.913, 81.8% sensitivity and 86.0% specificity for predicting 50% of venous invasion in TET patients ( Table 6 ).

According to the ROC analyses, some univariate features demonstrated different predictive efficacy for predicting 50% artery invasion in TET patients, including gender, tumor maximum diameter, contour, homogeneity, enhancement degree, pericardial or pleural invasion, lung invasion, pericardial or pleural effusion, and lymphadenopathy. Similarly, in addition to the enhancement degree, all of the above features were also helpful in predicting 50% venous invasion, with varying degrees of predictive efficacy ( Table 6 ).

Based on Cohen’s kappa coefficient, the consistency between the two readers was good in judging the invasion degree for the aorta or pulmonary artery and its branches, and SVC or innominate vein; Cohen’s kappa coefficient was 0.66 and 0.70, respectively.