OBJECTIVES: To prospectively compare the results of the pathologic examination including thin 3- to 4-mm thick serial slices of the surgical specimen in a well-defined anatomic part of the liver with those of 4 different liver imaging techniques. BACKGROUNDS SUMMARY DATA: Liver metastases (LM) of neuroendocrine origin are frequent and accuracy of pretherapeutic liver imaging in detecting small-sized LM is questioned, but no study has compared the different techniques used to image large liver specimens and meticulous pathologic examination. PATIENTS AND METHODS: Eleven patients with a metastatic endocrine tumor were selected before undergoing a right hepatectomy or a left hepatectomy passing along the median hepatic vein. Before the hepatectomy, patients underwent somatostatin receptor scintigraphy, computed tomography, magnetic resonance imaging, and ultrasonography (US). An intraoperative assessment (palpation and US) was also performed. Liver specimens were systematically cut into thin serial slices, 3- to 4-mm thick. A final histologic count of the total number of LM was obtained and compared with findings from each imaging technique. RESULTS: Compared with the final histologic count of LM, fewer than 50% of the LM were detected preoperatively. The accuracy was 24% for somatostatin receptor scintigraphy, 38% for computed tomography and US, and 49% for magnetic resonance imaging (the only imaging technique that detected half the number of LM). The size of the smallest LM was not greater than 2 mm in 54% of the patients. CONCLUSION: Half the number of LM from neuroendocrine tumors are undetectable on preoperative imaging.