An anatomy-based beam segmentation tool for intensity-modulated radiation therapy and its application to head-and-neck cancer.

PURPOSE: In segmental intensity-modulated radiation therapy (IMRT), the beam fluences result from superposition of unmodulated beamlets (segments). In the inverse planning approach, segments are a result of ''clipping'' intensity maps. At Ghent University Hospital, segments are created by an anatomy-based segmentation tool (ABST). The objective of this report is to describe ABST. METHODS AND MATERIALS: For each beam direction, ABST generates segments by a multistep procedure. During the initial steps, beam's eye view (BEV) projections of the planning target volumes (PTVs) and organs at risk (OARs) are generated. These projections are used to make a segmentation grid with negative values across the expanded OAR projections and positive values elsewhere inside the expanded PTV projections. Outside these regions, grid values are set to zero. Subsequent steps transform the positive values of the segmentation grid to increase with decreasing distance to the OAR projections and to increase with longer pathlengths measured along rays from their entrance point through the skin contours to their respective grid point. The final steps involve selection of iso-value lines of the segmentation grid as segment outlines which are transformed to leaf and jaw positions of a multileaf collimator (MLC). Segment shape approximations, if imposed by MLC constraints, are done in a way that minimizes overlap between the expanded OAR projections and the segment aperture.
RESULTS: The ABST procedure takes about 3 s/segment on a Compaq Alpha XP900 workstation. In IMRT planning problems with little complexity, such as laryngeal (example shown) or thyroid cancer, plans that are in accordance with the clinical protocol can be generated by weighting the segments generated by ABST without further optimization of their shapes. For complex IMRT plans such as paranasal sinus cancer (not shown), ABST generates a start assembly of segments from which the shapes and weights are further optimized.
CONCLUSIONS: ABST is a fast procedure to generate a set of segments for IMRT planning. The plan is finalized by assigning weights to the segments or by direct optimization of segment shapes and weights. ABST allows us to avoid the step of translating optimized intensity maps to sequences of segments.