A stereotactic brain biopsy system that is magnetic resonance (MR) imaging-guided has not been validated in dogs. Our purpose was to determine the mean needle placement error in the caudate nucleus, thalamus, and midbrain of a canine cadaver brain using the modified Brainsight stereotactic system. Relocatable reference markers (fiducial markers) were attached to the cadaver head using a dental bite block. A T1-weighted gradient echo three-dimensional (3D) sequence was acquired using set parameters. Fiducial markers were used to register the head to the acquired MR images in reference to a 3D position sensor. This allowed the planning of trajectory path to brain targets in real time. Coordinates (X, Y, Z) were established for each target and 0.5 microl of diluted gadolinium was injected at each target using a 26-gauge needle to create a lesion. The center of the gadolinium deposition was identified on the postoperative MR images and coordinates (X', Y', Z') were established. The precision of this system in bringing the needle to target (needle placement error) was calculated. Seventeen sites were targeted in the brain. The mean needle placement error for all target sites was 1.79 +/- 0.87 mm. The upper bound of error for this stereotactic system was 3.31 mm. There was no statistically significant relationship between needle placement error and target depth (P = 0.23). The ease of use and precision of this stereotactic system support its development for clinical use in dogs with brain lesions > 3.31 mm.