BACKGROUND: The use of computed tomography (CT) or magnetic resonance (MR) to overlay or register uptake patterns displayed by single-photon emission computed tomography (SPECT) with specific underlying anatomy has the potential to improve image interpretation and decrease diagnostic reading errors. The authors have developed a method that will allow the selection of a region of interest on MR or CT images that correlates with SPECT antibody images from the same patient. This method was validated first in phantom studies and subsequently was used on three patients with suspected colorectal carcinoma. METHODS: Two patients were injected with the technetium-99m-labeled 88BV59 immunoglobulin G human antibody, and the third patient was injected with the iodine-131-labeled 16.88 immunoglobulin M human antibody. CT or MR scans were obtained before antibody infusion, and subsequent SPECT scans were obtained on the first or fourth day after infusion. A customized body cast with landmarks was used for each patient during the CT, MR, and SPECT scans to match slice positions for all scanning modalities. Corresponding fiducial landmarks were identified on axial images. A computer graphics program was written to match and overlay corresponding landmarks for each imaging modality. The image registration accuracy was measured by comparing fiducial marker separations (center to center) on the registered scans. This separation uncertainty was 1-2 mm for CT-MR and 3-4 mm for CT-SPECT phantom studies. RESULTS: For patient studies, the fiducial alignment uncertainty was 3-4 mm for axial CT-SPECT and MR-SPECT images, and 6-8 mm for sagittal CT-SPECT and MR-SPECT images. The accuracy of the anatomic alignment of the patient and image registration system was +/- 1 cm in the medial-lateral axis and +/- 2 cm in the cranial-caudal direction. CONCLUSIONS: This type of image analysis may resolve uncertainties with the anatomic correlation of SPECT images that otherwise may be regarded as questionable when SPECT is used alone for radioimmunodiagnosis.