Purpose: Although computed tomography (CT) is widely used for diagnosing liver metastasis from colorectal cancer, diagnostic accuracy of CT is not satisfactory. Magnetic resonance (MR) imaging and F-18 FDG PET has been reported to be superior to CT. However, studies on direct comparison of PET and MR are scarce. We compared the diagnostic accuracy of FDG PET and MR in detecting liver metastasis from colorectal cancer. Materials and Methods: Among 363 colorectcal cancer patients who underwent F-18 FDG PET (ECAT, Siemens-CTI, Knoxville; Gemini, Philips, Milpitas, U.S.), 26 patients (M:F=17:9, age=62¡¾11) underwent MR to evaluate suspicious metastatic liver lesions. Finally, 35 liver lesions detected by CT from 26 patients were enrolled for analysis. PET and MR results were compared with pathologic reports, clinical findings or follow-up results. Results: Of the 35 lesions, 18 lesions (51.4%) were diagnosed as liver metastases, while remaining 17 (48.6%) as benign. The sensitivity and the specificity of PET were 94.4% and 94.1%, respectively, compared to 100% and 82.4% for MR. MR and PET was concordant in 30 lesions (85.7%): 17 metastatic (94.4%) and 13 benign (76.5%) lesions. ROC curve analysis revealed maximal SUV of 3.1 as the optimum standard in differentiating metastatic from benign liver lesions (AUC=0.897, p<0.001, sensitivity 83.3%, specificity 94.1%). For small lesions less than 1cm in diameter (n=20), diagnostic accuracy of PET was comparable to that of MR. Conclusion: F-18 FDG PET showed good diagnostic performance in detecting liver metastasis from colorectal cancer, which was comparable to MR. (Nucl Med Mol Imaging 2006;40(5):249-256)