ADAPTIVE MODIFICATION OF TREATMENT PLANNING TO MINIMIZE THE DELETERIOUS EFFECTS OF TREATMENT SETUP ERRORS

Purpose: Using daily setup variation measured from an electronic porta l imaging device (EPID), radiation treatment of the individual patient can be adaptively reoptimized during the course of therapy. In this s tudy, daily portal images were retrospectively examined to: (a) determ ine the number of initial days of portal imaging required to give adeq uate prediction of the systematic and random setup errors; and (b) exp lore the potential of using the prediction as feedback to reoptimize t he individual treatment part-way through the treatment course. Methods and Materials: Daily portal images of 64 cancer patients, whose treat ment position was not adjusted during the course of treatment, were ob tained from two independent clinics with similar setup procedures. Sys tematic and random setup errors for each patient were predicted using different numbers of initial portal measurements. The statistical conf idence of the predictions was tested to determine the number of daily portal measurements needed to give reasonable predictions. Two treatme nt processes, were simulated to examine the potential opportunity for setup margin reduction and dose escalation. The first process mimicked a conventional treatment. A constant margin was assigned to each trea tment field to compensate for the average setup error of the patient p opulation. A treatment dose was then prescribed, with reference to a f ixed normal tissue tolerance, and then fixed in the entire course of t reatment. In the second process, the same treatment fields and prescri bed dose were used only for the initial plan and treatment. After seve ral initial days of treatments, the treatment field shape and position were assumed to be adaptively modified using a computer-controlled mu ltileaf collimator (MLC) in light of the predicted systematic and rand om setup errors, The prescribed dose was then escalated until the same normal tissue tolerance, as determined in the first treatment process , was reached. Results: The systematic setup error and the random setu p error were predicted to be within +/-1 mm for the former and +/-0.5 mm for the latter at a greater than or equal to 95% confidence level u sing less than or equal to 9 initial daily portal measurements. In the study, a large number of patients could be treated using a smaller he ld margin if the adaptive modification process were used. Simulation o f the adaptive modification process for prostate treatment demonstrate s that additional treatment dose could be safely applied to 64% of pat ients. Conclusion: The adaptive modification process represents a diff erent approach for use of on-line portal images. The portal imaging in formation from the initial treatments is used as feedback for reoptimi zation of the treatment plan, rather than adjustment of the treatment setup. Results from the retrospective study show that the treatment of individual patient can be improved with the adaptive modification pro cess. (C) 1997 Elsevier Science Inc.