Validation of a method for automatic image fusion (BrainLAB System) of CT data and 11C-methionine-PET data for...

(a) To implement a fully automatic method to integrate (11)C-methionine positron emission tomography (MET-PET) data into stereotactic radiation treatment planning using the commercially available BrainLAB System, by means of CT/MET-PET image fusion. (b) To validate the fully automatic CT/MET-PET image fusion technique with respect to accuracy and robustness. (c) To give a short glance at the clinical consequences for patients with brain tumors.

In 12 patients with brain tumors (9 meningeomas, 3 gliomas), CT, MRI, and MET-PET were performed for stereotactic fractionated radiation treatment planning. The CT and MET-PET investigations were performed using a relocatable mask for head fixation. Fifteen external reference markers (5 on each lateral and 5 on the frontal localizer plate) that could be identified in CT and MET-PET were applied on the stereotactic localizer frame; the marker positions were exactly defined for both investigations. The MRI/CT fusion was done completely automatically. The CT/MET-PET fusion was performed using two different methods: The gold standard was the CT/PET fusion based on the reference markers, and the test method was the automatic, intensity-based CT/PET fusion, independent of the external markers. The markers visible on CT and transmission PET were matched using a point-to-line matching algorithm. To quantify the amount of misregistration, the two fusion methods were compared by calculating the mean value of deviation between corresponding points inside a cubic volume of interest of > or =512 cm(3) defined within the cranial cavity. The gross tumor volume (CT/MRI) outlined on CT and T1-MRI with contrast medium was compared with the gross tumor volume (PET) defined in the reoriented MET-PET data sets. The clinical impact of MET-PET in tumor volume definition for stereotactic radiotherapy will be discussed.

The fully automatic integration of MET-PET into stereotactic radiation treatment planning was successfully realized in all patients investigated. Mean deviation of the intensity-based automatic CT/PET fusion compared with the external marker-based gold standard was 2.4 mm; the standard deviation was 0.5. The algorithm's robustness was evaluated, and the discrepancy of fusion results due to different initial image alignments was determined to be below 1 mm inside the test volume of interest. In patients with meningiomas and gliomas, MET-PET was shown to deliver additional information concerning tumor extension.

The precision of the automatic CT/PET image fusion was high. A mean deviation of 2.4 mm is acceptable, considering that it is approximately equal to the pixel size of the PET data sets. MET-PET improves target volume definition for stereotactic fractionated radiotherapy of meningiomas and gliomas.