A clinical interactive technique for MR-CT image registration for target delineation of intracranial tumors

Replacement of current CT-based, three-dimensional (3D) treatment planning systems by newer versions capable of automated multi-modality image registration may be economically prohibitive for most radiation oncology clinics. We present a low-cost technique for MR-CT image registration on a "first generation" CT-based, 3D treatment planning system for intracranial tumors. The technique begins with fabrication of a standard treatment mask. A second truncated mask, the "minimask," is then made, using the standard mask as a mold. Two orthogonal leveling vials glued onto the minimask detect angular deviations in pitch and roll. Preservation of yaw is verified by referencing a line marked according to the CT laser on the craniocaudal axis. The treatment mask immobilizes the patient's head for CT. The minimask reproduces this CT-based angular treatment position, which is then maintained by taping the appropriately positioned head to the MR head coil for MR scanning. All CT and MR images, in DICOM 3.0 format, are entered into the treatment planning system via a computer network. Interactive registration of MR to CT images is controlled by real-time visual feedback on the computer monitor. Translational misalignments at the target are eliminated or minimized by iterative use of qualitative visual inspection. In this study, rotational errors were measured in a retrospective series of 20 consecutive patients who had undergone CT-MR image registration using this technique. Anatomic structures defined the three CT orthogonal axes from which angular errors on MR image were measured. Translational errors at the target isocenter were within pixel size, as judged by visual inspection. Clinical setup using the minimask resulted in overall average angular deviation of 3 degrees +/-2 degrees (mean +/- SD) and translational deviation within the edges of the target volume of typically less than 2 mm. The accuracy of this registration technique for target delineation of intracranial tumors is compatible with practice guidelines. This method, then, provides a cost-effective means to register MR and CT images for target delineation of intracranial tumors.