Despite generally good performance, mutual information has also been shown by several researchers to lack robustness for certain registration problems. A possible cause may be the absence of spatial information in the measure. The present paper proposes to include spatial information by combining mutual information with a term based on the image gradient of the images to be registered. The gradient term not only seeks to align locations of high gradient magnitude, but also aims for a similar orientation of the gradients at these locations. Results of combining both standard mutual information as well a normalized measure are presented for rigid registration of three-dimensional clinical images (MR, CT and PET). The results indicate that the combined measures yield a better registration function than mutual information or normalized mutual information per se. The accuracy of the combined measures is compared against a screw marker based gold standard, revealing a similar accuracy for the combined measures to that of the standard measures. Experiments into the robustness of the measures with respect to starting position imply a clear improvement in robustness for the measures including spatial information.