A novel unidirectional intensity map segmentation method for step-and-shoot IMRT delivery with segment shape control

Abstract

In intensity modulated radiation therapy (IMRT), intensity maps are computed from prescribed doses to target volumes, adding dose restrictions to the surrounding tissues. Those intensity (fluence) maps are discretized into matrices of natural numbers and translated to sequences of multileaf collimator (MLC) leaf movements, which will finally deliver the computed x-ray intensities. A unidirectional leaf sequencing algorithm that controls the shape of the segments and reduces leaf motion time for step-and-shoot dose delivery is presented. The problem of constructing segments in controlling its shape was solved by synchronizing right leaves motion. This is done without increasing the number of segments, or the total number of monitor units, and taking into account the unidirectional leaf motion and the interleaf collision constraints. The method was tested using random matrices and a clinical case planned with the PCRT 3D(R) planning system. Compared to other unidirectional leaf sequencing methods, the proposed algorithm performs very similarly. But, in addition, the segment shape control produces segments with smoother outlines and more compact shapes, which may help to reduce MLC-specific effects when delivering the planned fluence map. Finally, as a result of imposing unidirectionality, this algorithm can be extended for sliding window segment generation.

Reducing the number of segments in unidirectional MLC segmentations

“…However, it will not be considered in this work, because there are studies reporting that for IMRT plans with five or more beams and a large number of segments, the tongue-and-groove effect on the IMRT dose distribution is clinically insignificant due to the 'smearing' of dose in tissue (Deng et al 2001). Besides, the tongue-and-groove constraint added to solve it would yield to segmentations with around 10% in average more segments (Kalinowski 2006, Artacho et al 2009, depending on the algorithm and case.…”

“…The experiments were performed: (1) with two clinical cases planned with the PCRT 3D R (Técnicas Radiofísicas S.L., C/Gil de Jasa, 18E, 50006 Zaragoza, Spain, www.trf.es) treatment planning system; (2) using the RP technique (Siochi 1999), the OTNMU and ONS algorithms described in Artacho et al (2009, p 577), and a bidirectional method based on a branch-and-cut (BC) strategy included in the PCRT 3D R software; and (3) in a Siemens ONCOR TM Impression Plus linear accelerator for obtaining beam delivery times. The experiments were performed under the following conditions: (1) unidirectional segmentations with RP, OTNMU and ONS are done in both directions, from left to right and from right to left, and the best solution is selected; (2) none of the methods use the tongueand-groove constraint for the reasons explained in section 3.1, and if only the ICC is included in the RP and OTNMU, the TNMU obtained is optimum (Kalinowski 2006, Artacho et al 2009; (3) all the results were obtained with maxReduction = 100%, so the objective was the maximum possible NS reduction; and (4) maxd was set to 1 for the RP and 2 for the other unidirectional methods, because the ONS and OTNMU obtain steeper segmentations in the 3D space.…”

“…In this note, we will focus on the MLC step-and-shoot technique (Galvin et al 1993), which has many segmentation methods either unidirectional (Bortfeld et al 1994, Artacho et al 2009, and the rod pushing (RP) technique described in Siochi (1999), where leaves move only in one direction, or bidirectional (Galvin et al 1993, Xia and Verhey 1998, Siochi 1999, Dai and Zhu 2001, Crooks et al 2002, Kalinowski 2006.…”

“…Unidirectionality inherently minimizes the leaf motion time for a given set of segments (Siochi 1999, p 673), and some unidirectional segmentation methods like the RP or the OTNMU (Artacho et al 2009, p 577) are optimum, or close to the optimum, regarding the TNMU (Kalinowski 2006, Artacho et al 2009. However, the generally larger NS in unidirectional segmentations (Kalinowski 2006, Artacho et al 2009 compared to bidirectional ones can seriously affect delivery times, especially if the V&R cycle of the MLC used is high (Dai andZhu 2001, p 2113), since there are more segments to deliver and it is the only variable far from its optimum value. Therefore, a very valuable improvement in unidirectional segmentations would be the reduction of the NS, so that its main disadvantage diminishes or disappears.…”

Fixed number of segments in unidirectional decompositions of fluence matrices for step-and-shoot IMRT

“…In this work, we propose a method for post-processing MLC segmentations that can be included in two-phase step-and-shoot IMRT treatment planning systems and allows us to a priori fix the NS. This method is applicable to unidirectional MLC segmentations (Siochi 1999, Artacho et al 2009, where leaves are moved in a single direction. Unidirectional leaf movement makes these segmentations very suitable inputs for our method, since the leaf arrangement provides highly correlated adjacent segments, as can be seen in figure 1.…”