To clarify the situation, James Bedford from the Institute of Cancer Research and Royal Marsden NHS Foundation Trust in the UK has proposed a means of using ILS for aperture optimization. Bedford tackles the inverse planning problem for volumetric-modulated arc therapy (VMAT), though the concepts are equally applicable to IMRT (Phys. Med. Biol. 58 1235).

"ILS works well for fluence optimization, but we wanted to use it for aperture optimization," Bedford told medicalphysicsweb. "By converting fluence corrections into aperture shape corrections, the speed and accuracy of ILS can be used for aperture optimization."

Planning process

Bedford details the use of the AutoBeam inverse treatment planning system to optimize fluence, perform aperture sequencing and then optimize the apertures. For planning, the patient is represented by a 3D voxelized model containing volumes-of-interest, each of which has a prescribed dose and an importance factor.

ILS is used for fluence optimization, which involves distributing equally spaced control points around the VMAT arcs, and grouping these to deliver a series of intensity-modulated fluence maps around the patient. An initial uniform fluence is perturbed to meet specified dose-volume constraints. The optimized fluence is then sequenced into a starting set of deliverable beam apertures, a necessary step but one that impacts plan quality.

Bedford then describes a means of using ILS for aperture optimization, which is applied to restore plan quality. For each MLC leaf, at every control point, regions are defined on either side of the current leaf position. An iterative correction factor is computed and applied to each of these regions. If the dose beyond a leaf is too high, this leaf is closed slightly, and if the dose underneath a leaf is too low, it is opened up slightly. After each iteration, the weight (delivered monitor units) of each control point is adjusted and delivery constraints are imposed.

For this work, 60 fluence iterations and 40 aperture iterations were carried out per plan, using an Ultra 40 workstation with two 2.4 GHz dual-core AMD processors and 16 GB memory. A final two-arc plan with 360 control points was obtained in around 20 minutes.

Clinical examples

Bedford used the new heuristic to retrospectively plan five prostate cancer patients according to the PIVOTAL trial (which compares radiotherapy of the prostate and pelvis versus prostate treatment alone). Four planning target volumes (PTVs) were used, with 11 inverse planning objectives and a series of delivery constraints (including, for example, a maximum MLC leaf speed of 15 mm/s and a minimum planned gantry speed of 3.0 °/s). The resulting plans were recalculated in Pinnacle3 and then evaluated against the full clinical constraints.

Treatments were delivered over two arcs, each comprising 180 control points at 2° intervals. The first arc, which moves anticlockwise from gantry angles of 179° to 181°, deals with the overall structure of the fluence map; the second, which moves clockwise from 181° to 179°, deals with the finer structure.

To reveal exactly the impact of aperture optimization on the inverse planning process, plans produced using aperture and segment weight optimization were compared with those produced by segment weight optimization alone (using the apertures returned by the sequencing step).

For an example case, a beam's eye view of the first control point in each arc shows that the aperture for the second arc is narrower and further to the right of the beam's eye view than that of the first arc, as expected from the sequencing method used. The final dose distribution for a typical case conforms the dose closely to the target volumes and provides substantial sparing of the organs-at-risk.

In the patient cohort as a whole, all aperture-optimized plans met all of the clinical constraints (with the exception of three patients for whom bowel constraints were not met). Mean dose-volume histograms showed little difference between the two planning techniques for the critical structures. Aperture optimization, however, provided greater dose uniformity for all PTVs. Without aperture optimization, clinical constraints were not met for the PTVs.

Sinogram analysis

Bedford also used sinograms – plots of monitor unit distribution across the MLC leaf range – to assess the effects of aperture optimization. These revealed that ILS-based aperture optimization reduced MLC leaf motion, effectively smoothing the leaf trajectory, and minimized monitor unit variation between control points.

Sinograms for an example leaf pair in the pelvic nodes region during the first VMAT arc exhibited a saw-tooth shape after segment weight optimization alone. Adding aperture optimization significantly reduced this motion. Similar results were seen for a leaf pair in the prostate region.

Bedford concluded that ILS provides a powerful tool for aperture optimization in VMAT. He is currently adjusting the calculation of scattered radiation, to enable more accurate calculation of dose distribution during and after the optimization. "We are also working on portal dosimetry, so that we can more easily verify the resulting treatment plans and confirm the dose in vivo," he said.

Related articles in PMB
Sinogram analysis of aperture optimization by iterative least-squares in volumetric modulated arc therapy
James L Bedford Phys. Med. Biol. 58 1235
Comparative analysis of Pareto surfaces in multi-criteria IMRT planning
K Teichert et al Phys. Med. Biol. 56 3669
Intensity-modulated arc therapy: principles, technologies and clinical implementation
Cedric X Yu and Grace Tang Phys. Med. Biol. 56 R31