The sensitivity of magnetic resonance (MR) imaging for detection of pulmonary metastases in 11 patients scheduled for thoracotomy and curative resection of metastases was evaluated with a prospective, controlled study. MR imaging performed at 0.5 T was compared with chest radiography, computed tomography (CT), and thoracotomy in 12 cases. (One patient had two separate occurrences of pulmonary metastases.) All images were interpreted in blinded fashion. When all MR sequences were interpreted together, MR imaging enabled correct identification of all patients with pulmonary nodules (100%). CT enabled detection of at least one nodule in all 12 cases (100%) by design; the sensitivity of chest radiography was only 64%. For individual nodules, MR imaging was at least as sensitive as CT (P2 less than .25 [two-sided value]) and significantly more sensitive than chest radiography (P2 less than .01). Among all MR sequences, short inversion time inversion-recovery sequences had the highest sensitivity for detection of individual nodules (82%).