PURPOSE: Our study was designed to investigate the value of different imaging techniques in the early detection of VX-2 carcinoma in rabbit livers by direct comparison of sonography, CT, and MR imaging with pathologic specimens. MATERIALS AND METHODS: VX-2 carcinomas were induced in 20 rabbit livers by direct inoculation of 0. 1ml of aliquot of the tumor tissue suspension (1 x 106/mm3) to hepatic parenchyma with an 18 gauge needle after laparotomy. One week after the inoculation of tumor cells, sonography with a 5 MHz transducer, CT after a bolus injection of contrast medium (3 mL/kg), and Tl-weighted MR imaging at a 2. 0T with spin-echo techniques (TR/TE:500/30 msec, 2000/30, 80 msec) as well as enhanced Tl-weighted MR imaging with GD-DTPA (0. 1 mmole/kg) were performed in each rabbit. RESULTS: At pathologic specimens, adequate induction of tumor nodules were 69 with 3--21 mm size (average 8. 3 mm). All 69 tumors had fibrous capsules and central necrosis pathologically. The sensitivities of sonography, contrast enhanced CT and MR imaging were 54%, 61%, and 87%, respectively. Among MR images, the sensitivities of Tl-weighted images, proton density images, T2-weighted images, and enhanced Tl-weighted images were 33%, 77%, 83%, and 81%, respectively. On sonograms and CT scans, VX-2 carcinomas show variable appearance regardless of the tumor size and histopathologic findings. On MR images, the tumors were ring shape with variable thickness on all pulse sequences. There was no significant correlation among the findings of three imaging techniques. CONCLUSION: MR imaging has the highest detectability for the VX-2 carcinomas in rabbit liver. Therefore, MRI may be the most valuable imaging technique for the detection of small hepatic tumors with extensive necrosis like VX-2 carcinomas.