A diagnostic model based on clinical and enhanced computed tomography features to identify gastric schwannoma and gastric stromal tumor of different sizes: a multi-institution retrospective study.

BACKGROUND

The preoperative diagnosis of gastric stromal tumor (GST) and gastric schwannoma (GS) is of great significance for the selection of surgical methods. However, their similar clinical and imaging findings make it challenging to distinguish between the two tumors, and thus misdiagnosis occurs frequently before surgery. This study aimed to investigate the value of clinical data and enhanced computed tomography (CT) features in the differential diagnosis of GST and GS by a multi-institution retrospective analysis, for large (≥5 cm) and small (<5 cm) tumors.

METHODS

This study involved 493 patients with GST and 102 patients with GS. The enhanced CT imaging data and clinical characteristics of the participants were analyzed, including the complete clinical and imaging data, and postoperative pathological diagnosis. The cut-off value of 5 cm of the tumor was used to divide the patients into two groups (large group: 164 GST and 13 GS; small group: 329 GST and 89 GS). The categorical variables were statistically analyzed using the chi-squared test, and the continuous variables were statistically analyzed using independent samples t-test or Mann-Whitney U test. The statistically significant variables for the two groups were analyzed using multivariate logistic regression and receiver operating characteristic curve analysis.

RESULTS

A total of 177 patients were divided into a large GST (GST-L) group (n=164) and a large GS (GS-L) group (n=13), and 418 patients were divided into a small GST (GST-S) group (n=329) and a small GS (GS-S) group (n=89). Multivariate logistic regression analysis showed that the lymph node and necrosis were independent risk factors in large groups. In the small groups, multivariate logistic regression analysis showed that sex, location, growth pattern, lymph node, and necrosis were independent risk factors. The area under the curve was 0.983 in the large groups [accuracy (ACC) =95.5, sensitivity (SEN) =0.923, specificity (SPE) =0.957], and 0.931 in the small groups (ACC =77.8, SEN =0.816, SPE =0.922).

CONCLUSIONS

The two models have high diagnostic efficiency, and share significantly different features of lymph nodes and necrosis, whether in large or small groups, which may contribute to providing a precise reference for the planning of surgical methods.