Room: Exhibit Hall
Purpose: To develop a software application for computer assisted delineation of the clinical target volume (CTV) for glioma patients aiming to reduce inter- and intra-observer variability and decrease treatment planning time.
Methods: Automation of CTV delineation is a two-step process that includes automated segmentation of the brain structures and algorithmic expansion of the radiographically defined gross tumor volume (GTV) to encompass non-radiographically visible microscopic disease. Segmentation is based on a 3D convolutional neural network trained on the datasets for registered planning CT and diagnostic MR images where the GTV and critical structures were manually delineated. The surfaces of segmented structures serve as impenetrable barriers for tumor cells. After segmentation is complete, a contour of initial GTV expansion given by the radiation oncologists is adjusted by excluding voxels belonging to segmented structures from the volume within the radius of expansion. For each voxel its closest distance to the surface of the GTV is computed by projecting the voxel onto the GTV. If the projection crosses the barrier, the voxels are excluded from the CTV.
Results: The convolutional neural network was trained on 30 image sets and satisfactory results were obtained for segmentation of skull bone, surgical defects, optic chiasm, optic nerves, brainstem, corpus callosum, cerebellum, and brain hemispheres that define the falx cerebri. The algorithm for the GTV expansion was applied to images from the same set and the final CTV contours were presented to radiation oncologists for review.
Conclusion: The CTV delineation framework proposed in this study can be implemented as a tool to assist radiation oncologists in defining microscopic tumor extension. This will be especially valuable for physicians who are in practices with lower volume glioma patients as an aid to reproducibly delineate an optimal CTV of encompassing regions of high risk disease with minimization of unaffected normal tissues.
Not Applicable / None Entered.
Not Applicable / None Entered.