RESUMO
Objective:To evaluate the value of CT for differentiating gastric leiomyoma (GLM) from gastric schwannoma (GS).Methods:The clinical and imaging data of 42 patients with GLM (GLM group) and 41 patients with GS (GS group) were analyzed retrospectively. The general information and CT features were compared between two groups. The independent factors for differentiating GLM from GS were obtained by multivariate Logistic regression analysis. The receiver operating characteristic curve (ROC) was used to evaluate the diagnostic efficiency of the model.Results:The proportion of female and age in GLM group were significantly lower than those in GS group: 59.52% (25/42) vs. 85.37% (35/41), (51.83 ± 10.52) years old vs. (58.80 ± 10.63) years old, and there were statistical differences ( P<0.01). The upper part of the stomach rate, irregular shape rate, intraluminal growth rate, ratio of long diameter to short diameter and mild to moderate enhancement rate in GLM group were significantly higher than those in GS group: 71.43% (30/42) vs. 14.63% (6/41), 52.38% (22/42) vs. 21.95% (9/41), 92.86% (39/42) vs. 19.51% (8/41), 1.90 ± 0.55 vs. 1.34 ± 0.28 and 92.86% (39/42) vs. 51.22% (21/41), the cystic degeneration rate, ulcer rate, incidence of tumor-associated lymph node, CT values of venous phase and delayed phase in GLM group were significantly lower than those in GS group: 2.38% (1/42) vs. 26.83% (11/41), 7.14% (3/42) vs. 24.39% (10/41), 2.38% (1/42) vs. 60.98% (25/41), (59.21 ± 9.75) HU vs. (66.22 ± 10.33) HU and (65.02 ± 8.62) HU vs. (76.85 ± 11.89) HU, and there were statistical differences ( P<0.01 or <0.05); there were no statistical difference in the rate of calcification and the CT values of plain scan and arterial phase between the two groups ( P>0.05). Multivariate Logistic regression analysis result showed that the tumor location, growth mode, tumor-associated lymph node and ratio of long diameter to short diameter were the independent factors for differentiating GLM from GS ( OR = 34.385, 25.314, 0.023 and 97.700; 95% CI 2.848 to 415.171, 2.674 to 239.670, 0.001 to 0.637 and 3.113 to 3 066.549; P<0.01 or <0.05); when the model threshold was >0.647, the area under the curve was 0.988 (95% CI 0.934 to 1.000), with a sensitivity of 92.9% and specificity of 97.6%. Conclusions:When the tumor is prone to the upper part of the stomach, intraluminal growth, ratio of long diameter to short diameter >1.28, and the absence of the tumor-associated lymph node, GLM tends to be considered, on the contrary, it tends to be GS. Therefore, CT imaging features have certain value in differentiating GLM from GS before surgery.