Su1293 NON-INVASIVE DIAGNOSTIC METHOD OF GASTRIC SUBMUCOSAL TUMORS USING CIRCULARITY OF LESIONS: A RETROSPECTIVE ANALYSIS IN 44 CASES

Abstract

Gastrointestinal submucosal tumor (SMT) shown as a hypoechoic mass lesion in endoscopic ultrasonography (EUS) mainly includes gastrointestinal stromal tumor (GIST), leiomyoma and schwannoma, and it is difficult to distinguish them by using imaging modalities. Although EUS-guided fine-needle aspiration is a standard option to diagnose SMTs, that is more or less invasive and sometimes less satisfactory due to insufficient amount of obtained tissue. Therefore, a new non-invasive and more reliable diagnostic method is still desired. We hypothesized that gastric SMTs could be discriminated by comparing a shape of lesions in EUS images, because leiomyomas generally appeared craggy whereas the others were roundish. Therefore, this study aimed to investigate diagnostic ability in circularity of SMTs, a surrogate indicator of roundness calculated as four pi times the area divided by the perimeter squared (0–1, 1 = a true circle). A total of 44 gastric SMTs 20–45 mm in diameter, which had been surgically removed after evaluation with EUS and finally diagnosed as GIST, leiomyoma or schwannoma between January 2010 and December 2018, were retrospectively collected at two institutions. In each lesion, one EUS still image showing the maximal area was selected. Then, two endoscopists who did not know histological diagnosis evaluated the circularity in three times per lesion by using ImageJ (ver. 1.50e; National Institutes of Health, USA), and mean circularity were compared among the three groups. Subsequently, a cut-off value was calculated by investigating a receiver operating characteristics (ROC) curve in groups with a significant difference. We also investigated an interobserver agreement between two evaluators. In 44 SMTs, GIST, leiomyoma, and schwannoma were 31, 8, and 5, respectively. Mean circularity in these three groups were 0.86 ± 0.74, 0.69 ± 0.82 and 0.89±0.57, respectively, with a statistical significance (p <0.01). The mean circularity in leiomyoma was significantly lower than that of GIST and schwannoma (p<0.01) whereas there were no significant difference between GIST and schwannoma (p = 0.054). When a cut-off value was set at 0.82, the area under ROC curve for predicting leiomyoma was 0.935 with 81.9% of sensitivity and 93.7% of specificity. Regarding the interobserver agreement, a kappa value was 0.78 (p <0.01). This retrospective pilot study suggests that circularity of SMTs calculated by using EUS still images may be useful as a non-invasive and reliable diagnostic indicator to distinguish leiomyoma from other mesenchymal tumors with high accuracy and reproducibility. To reconfirm and establish this diagnostic strategy, a large-numbered, prospective study is required.