Simulation of 3D-treatment plans in head and neck tumors aided by matching of digitally reconstructed radiographs (DRR) and on-line distortion corrected simulator images

Background and purpose: Simulation of 3D-treatment plans for head and neck malignancy is difficult due to complex anatomy. Therefore, CT-simulation and stereotactic techniques are becoming more common in the treatment preparation, overcoming the need for simulation. However, if simulation is still performed, it is an important step in the treatment preparation/execution chain, since simulation errors, if not detected immediately, can compromise the success of treatment. A recently developed PC-based system for on-line image matching and comparison of digitally reconstructed radiographs (DRR) and distortion corrected simulator monitor images that enables instant correction of field placement errors during the simulation process was evaluated. The range of field placement errors with noncomputer aided simulation is reported. Materials and methods: For 14 patients either a primary 3D-treatment plan or a 3D-boost plan after initial treatment with opposing laterals for head and neck malignancy with a coplanar or non-coplanar two-or three-field technique was simulated. After determining the robustness of the matching process and the accuracy of field placement error detection with phantom measurements, DRRs were generated from the treatment planning CT-dataset of each patient and were interactively matched with on-line simulator images that had undergone correction for geometrical distortion, using a landmark algorithm. Translational field placement errors in all three planes as well as in-plane rotational errors were studied and were corrected immediately. Results: The interactive matching process is very robust with a tolerance of <2 mm when suitable anatomical landmarks are chosen. The accuracy for detection of translational errors in phantom measurements was <1 mm and for in-plane rotational errors the accuracy had a maximum of only 1.5 degrees. For patient simulation, the mean absolute distance of the planned versus simulated isocenter was 6.4 +/- 3.9 mm. The in-plane rotational error in both planes was <3 degrees with one exception. Three large field placement errors (two patients with 11.5 and 16.0 mm distances of the planned versus simulated isocenter, respectively and one patient with a 7 degrees rotational error) were detected and, as with the smaller errors, were immediately corrected. Conclusion: On-line image matching of treatment planning CT-derived DRRs and distortion corrected treatment simulator images is a precise and reliable method to reduce field placement errors in the simulation of complex 3D-treatment plans for head and neck malignancy and thus enhances accuracy in the first step of the treatment preparation/execution chain. However, out-of-plane rotational errors could not be assessed and assumedly they are comparatively small since due to rigid fixation, detected in-plane errors were small. (C) 1997 Elsevier Science Ireland Ltd.