High-precision prostate cancer irradiation by clinical application of an offline patient setup verification...

To investigate in three institutions, The Netherlands Cancer Institute (Antoni van Leeuwenhoek Huis [AvL]), Dr. Daniel den Hoed Cancer Center (DDHC), and Dr, Bernard Verbeeten Institute (BVI), how much the patient setup accuracy for irradiation of prostate cancer can be improved by an offline setup verification and correction procedure, using portal imaging.

The verification procedure consisted of two stages. During the first stage, setup deviations were measured during a number (Nmax) of consecutive initial treatment sessions. The length of the average three dimensional (3D) setup deviation vector was compared with an action level for corrections, which shrunk with the number of setup measurements. After a correction was applied, Nmax measurements had to be performed again. Each institution chose different values for the initial action level (6, 9, and 10 mm) and Nmax (2 and 4). The choice of these parameters was based on a simulation of the procedure, using as input preestimated values of random and systematic deviations in each institution. During the second stage of the procedure, with weekly setup measurements, the AvL used a different criterion ("outlier detection") for corrective actions than the DDHC and the BVI ("sliding average"). After each correction the first stage of the procedure was restarted. The procedure was tested for 151 patients (62 in AvL, 47 in DDHC, and 42 in BVI) treated for prostate carcinoma. Treatment techniques and portal image acquisition and analysis were different in each institution.

The actual distributions of random and systematic deviations without corrections were estimated by eliminating the effect of the corrections. The percentage of mean (systematic) 3D deviations larger than 5 mm was 26% for the AvL and the DDHC, and 36% for the BVI. The setup accuracy after application of the procedure was considerably improved (percentage of mean 3D deviations larger than 5 mm was 1.6% in the AvL and 0% in the DDHC and BVI), in agreement with the results of the simulation. The number of corrections (about 0.7 on the average per patient) was not larger than predicted.

The verification procedure appeared to be feasible in the three institutions and enabled a significant reduction of mean 3D setup deviations. The computer simulation of the procedure proved to be a useful tool, because it enabled an accurate prediction of the setup accuracy and the required number of corrections.