Purpose: To compare pathology macroscopic tumor dimensions with magnetic resonance imaging (MRI) measurements and to establish the microscopic tumor extension of colorectal liver metastases. Methods and Materials: In a prospective pilot study we included patients with colorectal liver metastases planned for surgery and eligible for MRI. A liver MRI was performed within 48 hours before surgery. Directly after surgery, an MRI of the specimen was acquired to measure the degree of tumor shrinkage. The specimen was fixed in formalin for 48 hours, and another MRI was performed to assess the specimen/tumor shrinkage. All MRI sequences were imported into our radiotherapy treatment planning Results: Between February 2009 and January 2010 we included 13 patients for analysis with 21 colorectal liver metastases. Specimen and tumor shrinkage after resection and fixation was negligible. The best tumor volume correlations between MRI and pathology were found for T1-weighted (w) echo gradient sequence (r(s) = 0.99, slope = 1.06), and the T2-w fast spin echo (FSE) single-shot sequence (r(s) = 0.99, slope = 1.08), followed by the T2-w FSE fat saturation sequence (r(s) = 0.99, slope = 1.23) Conclusions: MRI tumor dimensions showed a good agreement with the macroscopic pathology suggesting that MRI can be used for accurate tumor delineation. However, microscopic extensions found beyond the tumor border indicate that caution is needed in selecting appropriate tumor margins. (C) 2012 Elsevier Inc.