Diagnostic performance of stomach CT compared with endoscopic ultrasonography in diagnosing gastric subepithelial...

To evaluate the diagnostic ability of multi-detector computed tomography (MDCT) compared to endoscopic ultrasonography (EUS) as a standard reference, and investigate the factors influencing the detection of small gastric subepithelial tumors (SETs) (<5 cm) on MDCT with stomach protocol.

We retrospectively investigated 70 patients who were suspected with gastric SETs on esophagogastroduodenoscopy (EGD), and underwent both EUS and computed tomographic (CT) scanning. EUS was performed by two gastroenterologists, and location, size, echotexture, echogenicity, and layer of origin were described when gastric SETs were detected on EUS. MDCTs were reviewed based on consensus of two radiologists blinded to the EUS result. Size, location, enhancement pattern, and contour of the lesion detected on CT were described. We calculated the diagnostic ability of CT compared to EUS with respect to detection of gastric SETs, and investigated the factors influencing detection of SETs on CT. We also used receiver operating characteristic (ROC) curve to obtain optimal cut-off size for predicting CT visibility of small SETs.

Of the 70 patients, who underwent both CT and EUS due to suspicious presence of SET on EGD, EUS detected 56 probable cases of SET and 14 cases of external compression. CT led to detection of 39 cases of SET out of the 56 cases. Sensitivity and specificity of CT was 69.6% and 100.0%, respectively. Positive predictive value (PPV) and negative predictive value (NPV) of CT were 100.0% and 45.2%, respectively. There was a significant difference in mean size of CT-detected lesions compared to CT-invisible lesions (14.36 mm vs. 8.52 mm, p < 0.001), but no significant differences in terms of layer of origin and location between these two groups (p > 0.5) were observed. The ROC analysis revealed that the optimal cut-off value, also referred to prediction of CT visibility, was 10 mm. Out of 70 cases, 26 cases (37.14%) were identified as external compression or insignificant lesions such as lipoma, hemangioma, lymphangioma, or gastritis cystica on CT, and do not require regular follow-up.

Stomach CT shows good feasibility with respect to depiction of small SETs, especially in cases where size is larger than 10 mm. Henceforth, it is proposed that stomach CT would be a complimentary or problem-solving tool for SET in evaluating the presence of external compression and characterization of tumors.