Assessment of imaging studies used with radiosurgery: a volumetric algorithm and an estimation of its error (Reprinted)

The Gamma Knife has played an increasingly important role in the neurosurgical treatment of patients. Intracranial lesions are not removed by radiosurgery. Rather, the goal of treatment is to induce tumor control. During planning, the creation of dose-volume histograms requires an accurate volumetric analysis of intracranial lesions selected for radiosurgery. In addition, an accurate follow-up imaging analysis of tumor volume is essential for assessing the results of radiosurgery. Nevertheless, sources of volumetric error and their expected magnitudes must be properly understood so that the operator may correctly interpret apparent changes in tumor volume. In this paper, the authors examine the often-neglected contributions of imaging geometry (principally image slice thickness and separation) to overall volumetric error. One of the fundamental sources of volumetric error is that resulting from the geometry of the acquisition protocol. The authors consider the image sampling geometry of tomographic modalities and its contribution to volumetric error through a simulation framework in which a synthetic digital tumor is taken as the primary model. Because the exact volume of the digital phantom can be computed, the volume estimates derived from tomographic "slicing" can be directly compared precisely and independently from other error sources. In addition to providing empirical bounds on volumetric error, this approach provides a tool for guiding the specification of imaging protocols when a specific volumetric accuracy, or volume change sensitivity, for particular structures is sought a priori. Using computational geometry techniques, the volumetric error associated with image acquisition geometry was shown to be dependent on the number of slices through the region of interest (ROI) and the lesion volume. With a minimum of five slices through the ROI, the volume of a compact lesion could be calculated accurately with less than 10% error, which was the predetermined goal for the purposes of computing accurate dose-volume histograms and determining follow-up changes in tumor volume. Accurate dose-volume histograms can be generated and follow-up volumetric assessments performed, assuming accurate lesion delineation, when the object is visualized on at least five axial slices. Volumetric analysis based on fewer than five slices yields unacceptably larger errors (that is,. 10%). These volumetric findings are particularly relevant for radiosurgical treatment planning and follow-up analysis. Through the application of this volumetric methodology and a greater understanding of the error associated with it, neurosurgeons can better perform radiosurgery and assess its outcome.