Purpose: The goal of the study was to investigate the use of automated registration techniques for interpretation of volume MR and high resolution FDG-PET images that were obtained from patients with brain tumors.
Method: Twenty-one patients with brain tumors were studied on one or more occasions using MRI and high resolution FDG-PET. The data were aligned using automated volume- and surface-matching algorithms. Composite images comprising the resliced pre- and postgadolinium spoiled GRE, T2-weighted SE, and PET data were constructed to correlate intensities of regions on the PET images with regions that corresponded to normal gray matter, white matter, and gadolinium enhancement.
Results: The accuracy of registration between the MR and PET images was estimated to be within 1-2 mm based upon the distance between surfaces of the outside of the head. In 12 of the 24 examinations, there were diagnoses of recurrent tumor, with only 5 of these exhibiting regions of higher FDG uptake than normal gray matter. For 19 of the 24 studies, the anatomic context provided by the registered MR images was found to be important in distinguishing recurrent tumor from necrosis based upon FDG uptake.
Conclusion: The automated alignment was found to be an important factor in interpreting the high resolution PET images. This was particularly true for small lesions close to the cortex and for situations where FDG uptake had been reduced by prior treatment with radiation therapy.