AbstractID: 8896 Title: Evaluation of the NCI benchmark to
characterize IMRT planning systems for clinical trials
The NCI benchmark is designed to characterize IMRT planning systems for potential use
in clinical trials. It consists of intracranial anatomy with an annular hemi-cylindrical
CTV half-surrounding a cylindrical normal tissue structure (NT). The longitudinal axis
of both structures is oriented in the craniocaudal direction. Planning goals were to
deliver prescribed dose (PD) to 100% of CTV, and keep NT dose < 60% of PD. Plans
were generated with the NOMOS Corvus system using data from Varian linacs with 10 x
8 mm beamlets from a MIMiC collimator, 10 x 4 mm beamlets from a Beak-MIMiC
collimator, and Millennium MLC beams. To make plans compliant with goals, a region
of avoidance (ROA) was inserted between the structures. To keep maximum dose within
the CTV, an ROA was placed around the surface of the CTV opposite the NT. Plans
using a conventional couch angle of 180° were superior to those with multiple angles
because the structures were of constant diameter and their axis was parallel to the gantry
axis. All three beam collimating systems produced similar plans, partly because the
structures were symmetric and of constant diameter. Thus the benchmark was found
biased towards a single couch position, and limited in its ability to differentiate plans
obtained with diverse collimator data. To more fully characterize an IMRT system’s
ability to plan for realistic anatomy, it is suggested that future benchmarks contain
structures such as cones or ellipsoids with variable diameters, and that these structures
have oblique orientation relative to anatomic axes.