Three-Dimensional Dose Optimization for Noncoplanar Treatment Planning with Conformal Fields.

Recent advances in imaging techniques, especially three dimensional reconstruction of CT images, have made precision tumor localization feasible. These imaging techniques along with developments in computer controlled radiation treatment machines have provided an important thrust in developing better techniques for cancer treatment. This often requires a complex noncoplanar beam arrangements and elaborate treatment planning, which, unfortunately, are time consuming, costly and dependent on operator expertise and experience. A reliable operator-independent dose optimization tool is therefore desirable, especially for 3D treatment planning. In this dissertation, several approaches (linear programming, quadratic programming, and direct search methods) of computer optimization using various criteria including least sire fitting on the 90% isodose to target periphery, dose uniformity, and integral dose are presented. All of these methods are subject to restrictions on the upper limit of the dose to critical organs. In the quadratic programming approach, Kuhn-Tucker theory was employed to convert the quadratic problem into one which permits application of the very powerful, revised simplex method. Several examples are used to analyze the effectiveness of these dose optimization approaches. The studies show that the quadratic programming approach with the criteria of least square fitting and critical organ constraints is superior in efficiency for dose optimization in 3D treatment planning, particularly for cases with a large number of beams. Use of least square fitting allows one to deduce optimized plans for irregularly shaped targets by employing a multi-isocentric technique. Our studies also illustrate the advantages of using irregular conformal fields, optimized beam energy, and noncoplanar beam arrangements in contrast to the conventional treatment which uses a symmetrical rectangular collimator, fixed beam energy, and coplanar beam arrangements. Optimized plans can be implemented in computer controlled conformal radiation therapy as well as in conventional treatment techniques. This work demonstrates a useful dose optimization routine to reduce the difficulty in using trial-and-error methods for 3D treatment planning. Personnel-intensive and time consuming processes can now be accomplished with ease in a reasonable time to yield an optimized plan for an individual patient. The dose distributions in these treatment plans are significantly better than those produced by conventional multiple beam techniques. With the clinical application of our optimization method, the quality of radiation therapy should be greatly enhanced.