A method for registration of single photon emission computed tomography (SPECT) and computed tomography (CT) images...

To describe a simple method of coregistration of nonanatomic liver SPECT and CT images acquired in separate sessions for three-dimensional (3D)-CRT planning utilizing dual radiolabeled and radiopaque body surface markers, and evaluate the accuracy of the registration on the patient surface.

Ten patients treated for liver metastases or hepatocellular carcinoma with stereotactic body radiation therapy or 3D-CRT were selected for this study to evaluate the SPECT∕CT registration process. All patients were positioned in a custom-molded vacuum bag on the flat table top. Nine radiopaque markers were taped to the abdominal surface in three axial planes at the level of the liver. Following CT imaging, the nine radiopaque markers were then labeled with radioactive tags, each containing 10 μCi of (99m)Tc, and SPECT images were acquired. The metric used to evaluate the registration was the fiducial registration error (FRE), defined as the root mean square of the distance between pairs of homologous markers on the CT and SPECT images. The evaluation of the registration accuracy was performed in two steps: first the minimum number of markers necessary to obtain a robust registration was optimized; second the FRE was calculated on the remaining set of unused markers. Additionally, the deformation of the patient's abdominal surface between CT and SPECT acquisition sessions was evaluated using the distances between all possible unused marker pairs on the CT and SPECT images separately. The root mean square of the CT-to-SPECT difference between those distances was used to define the deformation index (DI). The registration method was evaluated on all ten patients in addition to an anthropomorphic phantom study.

The minimum number of markers above which the registration was not improved by more than 1 mm was 4. The FRE, calculated over the 5 remaining markers, was 6.1 mm for the patient population and 1.8 mm for the phantom study. The DI was 5.0 mm on average over all 10 patients and correlated well with the FRE. The DI was 1.6 mm for the phantom study, which represented the imaging systems' resolution and the ability to place the CT and SPECT markers at the exact same location.

It is feasible to use radiolabeled and radiopaque dual body surface markers for registration of SPECT and CT images acquired in separate sessions allowing conformal avoidance of SPECT-defined functional normal liver. Point-based rigid registration accuracy on the patient surface of 6.1 mm can be achieved using 4 dual body surface markers. The main contribution to the registration error is the deformation of the abdominal surface, arising from the inability to setup the patient in the exact same position at different times on two different imaging systems, and to properly account for breathing artifacts on the CT and SPECT images.