Orthogonal Delivery to Improve IMRT Efficiency

Purpose/Objective(s) 

The delivery efficiency of IMRT is closely related to the number of segments used to deliver the prescribed dose distribution, or intensity map (IM). In current static leaf sequencing (SLS) method, to deliver an IM, MLC leaves move along one direction during the entire delivery process. This uni-direction delivery may not fully utilize the capacity of the MLC. In fact, in order to improve the efficiency of the IMRT delivery, it was proposed recently to rotate the MLC between the deliveries of the segments for an IM. In this work, we propose to optimally decompose an IM into two sub-IMs and use two orthogonal directions to delivery them to improve IMRT efficiency. More precisely, one sub-IM is delivered by moving MLC leaves, say, horizontally and the other is delivered by moving MLC leaves vertically after the MLC is rotated by 90 degrees.

Materials/Methods 

For a given IM M, we decompose it into two sub-IMs, λQ and R, to minimize the total complexity of both sub-IMs, where λ is an integer constant, and λQ and R are delivered orthogonally. The complexity of an IM determines the number of segments used in the delivery. We adopt the frequently used total sum of positive gradient to measure the complexity of an IM. Note that the IMs λQ and Q can be delivered by the same set of segments. Most of current SLS approaches are based on a method for reducing the intensity level of IMs, then compute a set of segments for the IMs with a smaller maximum intensity level. Our decomposition results in two “simpler” sub-IMs with smaller maximum intensity level, which, in turn, yields a more efficient delivery plan using fewer segments. Our algorithm is based on graph theory in computer science, by formulating the problem as computing a minimum s-t cut in a 3-D geometric graph.

To evaluate the performance of our method, we implemented our algorithm and experimented with the resulting software on 105 sets of clinical intensity maps obtained from the Department of Radiation Oncology of the University of Iowa. The clinical maps were derived from treatment plans designed for complex tumors surrounded by radio-sensitive critical structures observed in head and neck cancer patients, one of the more challenging IMRT planning scenarios. The number of segments, which was computed using Xia and Verhey's algorithm without considering interleaf motion constraint, was examined to measure the output results.

Results 

Experimental results showed that the reduction of the segments using our method was promising. 69 out of the 105 (65.7%) IMs had reduced numbers of segments for delivery. Our method produced 50% fewer segments with an average reduction of 20.09% comparing with the leaf sequencing method using a single direction for delivery. For the IMs with a total number of segments above the median, 67.3% of the IMs were improved, 76.92% of the top quarter were improved, and 90.0% of the top 10% of the IMs were improved. This shows that those IM's with the higher the number of segments had a greater chance of benefit when applying our method.

Conclusions 

Uni-direction delivery may not fully utilize the capacity of the advanced MLC's hence optimally decomposing it to two sub-IMs and delivering them in two orthogonal directions can help improve the IMRT delivery efficiency.