(Wednesday, 8/1/2018) 10:30 AM - 11:00 AM
Room: Exhibit Hall | Forum 6
Purpose: Tissue heterogeneity in the brain is generally ignored when using MRI. The existence of changes in dose distribution between Convolution and TMR 10 algorithms is known, but specific differences that may have clinical significance are not well-understood. We recently proposed a method for MRI-only stereotactic radiosurgery (SRS) planning; to improve it, we studied the dosimetry accuracy of synthetic CT images for convolution-based SRS planning that accounts for brain tissue inhomogeneity.
Methods: MR and CT images were acquired for a brain cancer patient. Synthetic CT images were generated from MR images using syngo.via RT Image Suite (Siemens Healthineers) using a fuzzy c-means method. Volumes and skull definitions were based on the MRI data set. This information used for planning on both synthetic CT and CT datasets using the convolution algorithm; the plan based on the TMR 10 algorithm was also generated using MRI dataset. Finally, all were evaluated for geometrical accuracy, isodose line coverage, and dose volume over two targets and brainstem.
Results: An average 3.87% overestimate for synthetic CT and 7.84% underestimate for CT in target and brainstem volume delineation compared with MRI. There was a 0.9% increase in max point dose inside the targets for synthetic CT, and 3% decrease in CT plans-compared with TMR 10-MRI. The average D100 for synthetic CT showed 0.9% decrease compared with 1.6% increase in CT-based compared to MRI plan. Overall, synthetic CT plans showed a 2.6% increase in D100 over CT images.
Conclusion: : Synthetic CT offered noticeable improvement in target and brainstem volume delineation. Using the CT plan as a reference, comparison of the convolution algorithm with TMR10-MRI showed a 1.6% increase in D100 convergence; the same algorithm with a synthetic CT plan showed a 0.9% decrease, possibly resulting from better co-registration.  
Keywords
Not Applicable / None Entered.
Taxonomy
Not Applicable / None Entered.
Contact Email
