Cumulative sum method in nonzero fixed action level setup correction strategy: An application of statistical process control for targeted prostate radiotherapy

Abstract

This study aims to incorporate a statistical process control tool, the cumulative sum (CUSUM) method, in the nonzero fixed action level setup correction strategy for the detection and correction of systematic shifts in prostate position during prostate image-guided radiotherapy. A daily nonzero action level setup correction strategy using CUSUM chart was proposed and tested retrospectively on 34 nonrandomized patients with prostate-implanted fiducial markers in situ plus daily localization via stereoscopic electronic portal imaging. A cumulative error index, which monitors a cumulative process index over time, was implemented in parallel with the localization offsets of the prostate. A value of this index exceeding a decision interval H indicated that a systematic change in the mean value of localization had occurred. When a value of the index exceeding H was recorded, the magnitude of the systematic shift was estimated and the subsequent prescribed offsets were corrected to account for the new setup positions. With the new protocol, the authors produced simulated treatment histories in which the detection of systematic shifts and the frequency of localization offset exceeding action level were evaluated. The protocol was examined against three levels of mean systematic shifts of localization for detection (μ(1)) of 3, 5, and 7 mm. The overall rates of detection of systematic shifts using this algorithm were relatively high. The level μ(1) = 3 mm resulted in the highest rate of detection of systematic offset change, in 42% of the charts. For larger values of μ(1), the overall detection numbers fell to 18% and 12% of the charts for μ(1) = 5 and 7 mm, respectively. The mean magnitude of change in systematic offset at time of detection was 3.1, 4.4, and 5.2 mm for μ(1) = 3, 5, and 7 mm, respectively. By adapting the setup position for any change in systematic localization offset over a course of treatment using the new protocol, it has been shown that the frequency of localization offsets exceeding action levels was reduced. For the optimal choice of intervention levels for correction and μ(1), the number of localization offsets exceeding action levels has been reduced by approximately half. The nonzero fixed action level setup correction strategy incorporating CUSUM charts has been shown to provide an objective method for (i) detecting a systematic shift of a patient localization offset, (ii) quantifying the magnitude of a systematic change based on previous observations, and (iii) correcting the prescribed offset by accounting for the systematic shift(s) in patient setup position.