MRI only radiotherapy using synthetic CT images: Dosimetric accuracy and patient positioning strategies in an MRI only workflow

Sammanfattning: Background and purpose: Magnetic resonance imaging (MRI) only radiotherapy has, due to superior soft tissue contrast of MRI and the possible reduction of systematic uncertainties, grown into a promising technique. The purpose of this work is to investigate parts of an MRI only workflow with synthetic computed tomography (sCT) images generated from MRI. This was done through investigation of treatment planning abilities and patient positioning strategies in a suggested MRI only workflow for prostate cancer patients. Material and methods: A novel atlas based generation method, The Statistical Decomposition Algorithm (SDA), was used for sCT generation. sCT images of six patients were evaluated visually and using the DICE similarity index. Treatment plans (Intensity modulated radiation therapy (IMRT) and protons) were calculated based on conventional CT and sCT and compared using clinical dose- volume histogram (DVH) criteria and gamma evaluation. Patient positioning in the MRI only workflow were investigated through two studies, including five patients. The ability to identify implanted gold fiducials in MRI was investigated and a MV bone match procedure tested with digitally reconstructed radiographs (DRR). Results: The image analysis implicated minor problems with the bone surfaces in the generated sCT, and some disparities compared to CT were noted. The treatment planning study showed that treatment planning on sCT introduced a mean dose difference to PTV of -0.2% to -0.1% compared to CT. Gamma analysis showed a mean pass rate close to 100% for IMRT (1%/2mm) and 93% for protons (3%/2mm). The fiducial marker identification study showed that the fiducials could be identified, in all cases but one. In this case a marker was identified centimeters from the right position. The bone match showed differences of 0.7-0.9 mm between match performed with CT-DRRs and sCT-DRRs images towards MV images. Conclusion: Good agreement in terms of dose accuracy was found when dose distributions were compared between sCT and CT treatment plans. The image evaluation indicated disparities in the bone surface between sCT and CT, but this had only small effect on the resulting dose distributions for photon plans. Patient positioning strategies needs further work and guidelines have to be developed. The initial results of this study show a problem with transferring the fiducial marker positioning strategies in the conventional workflow to the MRI only workflow. The MV bone match can be performed and, although it may not be the preferred procedure for prostate patient, the study showed good results and can be usable for other diagnosis.