OBJECTIVE To study the usefulness of three-dimensional spiral CT (3DCT) in the diagnosis of advanced gastric cancer (AGC). METHODS Between June 1999 and December 2000, 54 patients with AGC were consecutively examined. On the 3D Virtuoso workstation, source images were uploaded to create a 3DCT volume block that was then processed with volume rendering technology (VA30C) to achieve virtual-reality endoscopy (VE), clipped volume block (CVB), and ray sum (RS). After the above scanning, all the patients were examined by a two-phase enhanced spiral CT (2DCT). The visualization, manifestation, and Borrman's classification of lesions in VE, CVB, RS, and 2DCT were evaluated and correlated with gastroscopic, surgical, and pathological findings. Respiratory artifact and gastric residue were also observed. RESULTS (1) CVB showed the excellent visualization in 88.9% of cases, in contrast to VE and RS (50.0% and 38.9%) (P < 0.01). (2) The accuracy in evaluating mucous membrane, ulceration, lumen, wall, cardia, pylorus, and extension of the tumor were more than 90.0% except mucosa by RS (84.4%) and ulceration by VE (87.5%) or RS (81.6%) which was significantly different from CVB (96.0%) and 2DCT (96.1%) (P < 0.05). VE demonstrated an accuracy of 95.8% in diagnosis of mucosal abnormality. (3) The correct Borrman's classification was obtained in 83.3% cases by VE, 79.6% by CVB, 72.2% by RS, 88.9% by 2DCT and 85.2% by 3DCT with significant difference between 2DCT and RS (P < 0.05), but not between 3DCT and 2DCT (P > 0.05). (4) In addition to 2DCT which had no step-like artifacts, they were invisible in 53.7% of VE, 40.7% of CVB, and 81.5% of RS, with RS showing the least artifacts among 3DCT (P < 0.01). A few of gastric residues caused by pre-scanning intake of water to swallow effervescent agent could be found on 3DCT images which caused no evident influence on diagnosis. CONCLUSIONS Additional information on the diagnosis of AGC can be obtained by use of 3DCT, especially the visualization of a lesion in clipped volume block and the observation of mucosa in virtual-reality endoscopy.