The present proposal details an atlas-based methodology to autosegment brain structures using magnetic resonance (MR) images for the purposes of radiotherapy treatment. This may decrease the amount of time needed by physicians to segment structures in the brain. With decreased segmenting time, there may be greater incentive to use intensity modulated radiotherapy (IMRT) with inverse treatment planning, particularly at non-academic institutions. The use of IMRT may decrease patient morbidity through radiotherapy optimization planning which minimizes dose to normal structures while providing uniform dose coverage over the tumor. Dose escalation becomes possible when using IMRT, with the intent of improvement in local tumor control. MR brain images will be acquired from patients scheduled to undergo radiotherapy. Structures in thee images will be auto- and manually- segmented. While any structure in the brain could be auto-segmented using this method, the structures that will be segmented in this study include the eyes, optic nerves, optic chiasm, pituitary gland, and brainstem (spinal cord, medulla oblongota, pons, and mid brain). A radiotherapy treatment plan will be carried out on the computed tomographic scan fused with the MR scan and dose-volume-histogram data for the auto- and manually-segmented structures will be calculated. We propose to test the hypotheses that there is no difference between auto- and manually-segmented structures in position/shape and no clinical difference in radiation dose deposited. A statistical model will be used to determine factors which significantly impact auto-segmentation accuracy, such as structure size, location, and patient disease. This information will then be used to improve auto-segmentation constraints.
