Virtual Gastroscopy Using Spiral CT in Gastric Lesions

ABSTRACT

PURPOSE: To compare virtual gastroscopy using spiral CT with conventional endoscopy for the detection andevaluation of gastric lesions. MATERIALS AND METHODS: During a previous six-month period, 30 patients withpathologically-proven gastric lesions underwent conventional endoscopy and virtual gastroscopy using spiral CT.There were 18 cases of advanced gastric carcinoma, eight benign ulcers, and four submucosal tumors(two leiomyomas,two lymphomas). Source images of virtual gastroscopy were three-dimensionally reconstructed within an AdvantageWindows Workstation and virtual gastroscopy images were obtained using Navigator Software. On analysis, imageswere graded according to their quality (excellent, good, poor). Virtual gastroscopy images were interpreted by tworadiologists blinded to conventional endoscopic findings, and were subsequently compared with endoscopic findingsin terms of detectability and findings. In the cases of advanced gastric carcinoma, lesions were classifiedaccording to Borrmann's system. RESULTS: For virtual gastroscopy, overall image quality was excellent in 21cases(70%), good in five(17%), and poor in four(13%). Lesions were detected in 25 cases(83%). Among the 18advanced gastric carcinomas, virtual gastroscopy image quality was excellent in 14 cases(78%), good in two(11%),and poor in two(11%). Lesions were detected in 16 cases(89%). Two Borrmann type IV cases were not detected. Amongthe eight benign ulcers, virtual gastroscopy image quality was excellent in three cases(38%), good in three(38%),and poor in two(25%). The detection of lesion was possible in five cases(63%). In all submucosal tumors, virtualgastroscopy image quality was excellent. Lesions were detected in all cases. CONCLUSION: Virtual gastroscopyusing spiral CT is safe and noninvasive, and for the evaluation of gastric lesions may be complementary to axialCT. It successfully detects gastric lesions, and in depicting the pattern of gastric folds its image quality isexcellent.