Physical and clinical aspects of the dynamic intensity-modulated radiotherapy of 21 patients

Purpose: To describe the physical and clinical aspects of the dynamic intensity modulated radiotherapy of 21 patients. Methods and Materials: Dynamic, intensity modulated radiotherapy (IMR) was given to 21 patients with advanced or recurrent disease. 13 patients were immobilized with head screws, and 8, with non-invasive thermoplastic masks. The system was selected because it was designed de novo from a well established simulated annealing optimization model (SA), and with stringent leakage requirements and rapid leaf transit time for a multi leaf collimator (MLC). The system included a 6 MV linear accelerator (linac), an MLC, a quad processing computer system with SA software, a computer MLC controller with inclinometers and interlocks to stop radiation upon potential MLC or linac gantry fault detection, and immobilization devices attached to CT and treatment tables. The MLC was built around a 2 x 20 array of leaves with 9 half value layers of attenuation of the primary beam (99.8%). Over a trillion (240) beamlet patterns were dynamically changeable per deg. of linac gantry rotation. With all leaves shut, transmission was within a secondary collimator standard of < 0.5% of the primary beam. MLC control was via touch screen computer, and a disk drive which read beam pattern sequences from a disk generated by the planning system. Planning included 3D CT and magnetic resonance localization of regions of interest (ROI). The SA cost function incorporated idealized dose-volume parameter sets of up to 21 ROI/patient. Relative importance and spatial pre-eminance of each ROI were quantified into the constraint set, together with an instrument data file (IDR) built from depth dose and crossplot data of 8 x 8 to 20 x 200 mm field sizes and patterns measured with small diodes in a water tanc phantom. Planner output included dose volume histograms, tabulated dosimetry statistics, 2D dose distributions, and 3D translucent renderings of patient surfaces with underlying colored isodose surfaces surrounding tumors. Plans were approved after refining and reannealing input constraints to provide doses closest to those originally targeted. The approved plan was copied to a disk, and the disk was then inserted in the MLC controller for treatment delivery. Dose accuracy was tested with film, TLD, diodes and ionization chambers embedded in anthropomorphic and polystyrene calibration phantoms. Optical densities of dose calibration and verification films were obtained with a 300 point/'' scanner operated with photographic workshop software. Absolute dose distributions were extracted from film files with NIH Image software. Alignment of invasive immobilization was based on alignment of lasers on lateral cross hairs and an anterior 2 x 2 cm field light on an anterior field outline, respectively, embedded in a device fixed over the patient's head. Alignment of patients immobilized in face masks was based on lines and Barium paste dots placed on face masks. Marker alignment was repeated before each treatment to verify that patients were reproducibly positioned among CT and treatment sessions. Positional stability was radiographically documented by obtaining anatomy and fiducial markers near delineator wires on lateral CT scout views and comparing them to anatomy and lead fiducial markers on port films of 5 x 21 cm fields defined by the linac's secondary collimators. Different adjacent table positions were set up by adjusting the table such that side wall laser lights precisely filled a gap between white lines on a digital scale unit placed in the cranio caudal direction. The scale unit was zeroed at the origin of the treatment coordinate system and moved to positions indicated on planner printed treatment sheets. Patients were monitored in vivo with film and TLD under bolus on the surface expected to receive maximum surface dose. The minimum median, mean and maximum tumor volumes of the 21 patients were 2, 52, 68, and 200 cc, respectively. Prescribed doses (Rx) varied from 6 to 50.4, in fractions of 1.8 to 6 Gy. Results: Thus far 204 treatments have been well tolerated. Positional reproducibility among was ± 1 mm. Plan verification showed that for 18 of the 21 patients, delivered isodose surfaces were within 2 mm of planned surfaces, and delivered absolute doses, within ± 5 % of planned. Verifications of 3 patients with large and complex tumor geometry showed larger dose discrepancies that were believed to arise from imperfection in the dose calculation and verification procedures. Delivered distributions compared favorably to those of most other external beam treatment modalities. Improvements to external beam treatments offered by the IMR system are its abilities to provide highly conformal doses with detailed dose strategies for multiple ROI, and to irradiate multiple targets irrespective of their geometry and without having to change collimators as in radiosurgery. Recent replacement of the linac with a high dose rate linac and upgrades to the IMR system have increased dose calculation accuracy and ease of treatment. Conclusion: Heretofore unattainable conformity of dose distribution to large, complex, and convoluted targets is now a reality. The IMR system studied here made other external beam radiotherapy, including most radiosurgery with circular collimators, seem primitive. In conclusion, robust development of optimized dynamic photon pencil beam IMR is suggested