A comparison of 3-D data correlation methods for fractionated stereotacticradiotherapy

Purpose: Stereotactic radiosurgery is currently used to treat patients who are not good candidates for conventional neurosurgical procedures. For trea tments of nonvascular tumor cells, it appears that fractionation offers a r adiobiological advantage between tumor and normal tissues. Therefore, fract ionated stereotactic radiotherapy (FSR) is preferred because it minimizes n ormal tissue complications and maximizes local tumor control probability. W e have implemented a methodology clinically to perform the noninvasive pati ent repositioning technique. The 3-D data correlation method for high-preci sion and multiple fraction stereotactic treatments has been presented. Methods and Materials: Three different optimization algorithms (Hooke and J eeves optimization, simplex optimization, and simulated annealing optimizat ion) are evaluated to calculate the transformation parameters necessary for FSR. A least square object function is created to perform the 3-D data mat ching process. By minimizing the unconstrained object function value the be st fit can be approached for the reference 3-D data sets. Simulation shows that these algorithms deliver results that are comparable to the previously published correlation algorithm (1, 2) (singular value decomposition [SVD] method). The advantage for optimization algorithms is easily understood an d can be readily implemented by using a personal computer (PC). The mathema tical framework provides a tool to calculate the transformation matrix whic h can be used to adjust patient position for fractionated treatments. There fore, using these algorithms for a high-precision fractionated treatment is possible without an invasive repeat fixation device and has been implement ed clinically. A bite plate system was incorporated to acquire 3-D patient data. With a 3-D digital camera localization device, the patient motion can be followed in real time with the system calibrated to the isocenter. Results: Two types of data sets are utilized to study the correlation resul ts. One is using the digitized patient data which were retrieved clinically . The other is using the randomly generated data sets. Simulation errors fo r the optimization algorithms are all less than 1 mm in translation and les s than 1 degrees in rotation. Currently, FSR is performed using special des igned repeat fixation devices which assure reproducible patient position fo r multiple fractions of radiation treatment. Clinical results indicated tha t this technique provided excellent treatment results. Conclusion: Three optimization algorithms have been applied and evaluated i n calculating the transformation parameters between two 3-D contours or dig itized data points, The mathematical functions behind these optimization al gorithms are straightforward and can be easily implemented. When incorporat ed with the proper CT/MR image data with an electronic portal imaging (EPI) system, this process can possibly verify the patient's treatment position whenever there is doubt about the movement during the treatment procedure. (C) 1999 Elsevier Science Inc.