Fluoro-18-deoxyglucose positron emission tomography computed tomography (FDG-PET/CT) and magnetic resonance imaging (MRI), including unenhanced single-shot spin-echo echo planar imaging (SS SE-EPI) and small paramagnetic iron oxide (SPIO) enhancement, were compared prospectively for detecting colorectal liver metastases. Twenty-four consecutive patients suspected for metastases underwent MRI and FDG-PET/CT. Fourteen patients (58%) had previously received chemotherapy, including seven patients whose chemotherapy was still continuing to within 1 month of the PET/CT study. The mean interval between PET/CT and MRI was 10.2+/-5.2 days. Histopathology (n=18) or follow-up imaging (n=6) were used as reference. Seventy-seven metastases were detected. In nine patients, MRI and PET/CT gave concordant results. Sensitivities for unenhanced SS SE-EPI, MRI without SS SE-EPI and FDG-PET/CT were, respectively, 100% (p=9 x 10(-10) vs PET, p=8 x 10(-3) vs MRI without SS SE-EPI), 90% (p=2 x 10(-7) vs PET) and 60%. PET/CT sensitivity dropped significantly with decreasing size, from 100% in lesions larger than 20 mm (identical to MRI), over 54% in lesions between 10 and 20 mm (p=3 x 10(5) versus unenhanced SS SE-EPI), to 32% in lesions under 10 mm (p=6 x 10(-5) versus unenhanced SS SE-EPI). Positive predictive value of PET was 100% (identical to MRI). MRI, particularly unenhanced SS SE-EPI, has good sensitivity and positive predictive value for detecting liver metastases from colorectal carcinoma. Its sensitivity is better than that of FDG-PET/CT, especially for small lesions.