A Scoring Model for Radiologic Diagnosis of Gastric Leiomyomas (glms) with Contrast-Enhanced Computed Tomography (CE-CT): Differential Diagnosis from Gastrointestinal Stromal Tumors (Gists)
Jian-Xia Xu,Qiao-Ling Ding,Yuan-Fei Lu,Shu-Feng Fan,Qin-Pan Rao,Ri-Sheng Yu
DOI: https://doi.org/10.1016/j.ejrad.2020.109395
IF: 4.531
2020-01-01
European Journal of Radiology
Abstract:Objectives: To investigate CT findings and develop a diagnostic score model to differentiate GLMs from GISTs. Methods: This retrospective study included 109 patients with pathologically confirmed GLMs (n = 46) and GISTs (n = 63) from January 2013 to August 2018 who received CE-CT before surgery. Demographic and radiological features was collected, including lesion location, contour, presence or absence of intralesional necrosis and ulceration, growth pattern, whether the tumor involved EGJ, the long diameter (LD) /the short diameter (SD) ratio, pattern and degree of lesion enhancement. Univariate analyses and multivariate logistic regression analyses were performed to identify independent predictors and establish a predictive model. Independent predictors for GLMs were weighted with scores based on regression coefficients. A receiver operating characteristic (ROC) curve was created to determine the diagnostic ability of the model. Overall score distribution was divided into four groups to show differentiating probability of GLMs from GISTs. Results: Five CT features were the independent predictors for GLMs diagnosis in multivariate logistic regression analysis, including esophagogastric junction (EGJ) involvement (OR, 367.9; 95 % CI, 5.8-23302.8; P = 0.005), absence of necrosis (OR, 11.9; 95 % CI, 1.0-138. 1; P = 0.048) and ulceration (OR, 151.9; 95 % CI, 1.4-16899.6; P = 0.037), degree of enhancement (OR, 9.3; 95 % CI, 3.2-27.4; P < 0.001), and long diameter/ short diameter (LD/SD) ratio (OR,170.9; 95 % CI, 8.4-3493.4; P = 0.001). At a cutoff of 9 points, AUC for this score model was 0.95, with 95.65 % sensitivity, 79.37 % specificity, 77.19 % PPV, 96.15 % NPV and 86.24 % diagnostic accuracy. An increasing trend was showed in diagnostic probability of GLMs among four groups based on the score (P < 0.001). Conclusions: The newly designed scoring system is reliable and easy-to-use for GLMs diagnosis by distinguishing from GISTs, including EGJ involvement, absence of ulceration and necrosis, mild enhancement and high LD/SD ratio. The overall score of model ranged from 1 to 17 points, which was divided into 4 groups: 1-7 points, 7-10 points, 10-13 points and 13-17 points, with a diagnostic probability of GLMs 0%, 45 %, 83 % and 100 %, respectively.