Assessment of residual error for online cone-beam CT-guided treatment of prostate cancer patients.

PURPOSE Kilovoltage cone-beam CT (CBCT) implemented on board a medical accelerator is available for image-guidance applications in our clinic. The objective of this work was to assess the magnitude and stability of the residual setup error associated with CBCT online-guided prostate cancer patient setup. Residual error pertains to the uncertainty in image registration, the limited mechanical accuracy, and the intrafraction motion during imaging and treatment. METHODS AND MATERIALS The residual error for CBCT online-guided correction was first determined in a phantom study. After online correction, the phantom residual error was determined by comparing megavoltage portal images acquired every 90 degrees to the corresponding digitally reconstructed radiographs. In the clinical study, 8 prostate cancer patients were implanted with three radiopaque markers made of high-winding coils. After positioning the patient using the skin marks, a CBCT scan was acquired and the setup error determined by fusing the coils on the CBCT and planning CT scans. The patient setup was then corrected by moving the couch accordingly. A second CBCT scan was acquired immediately after the correction to evaluate the residual target setup error. Intrafraction motion was evaluated by tracking the coils and the bony landmarks on kilovoltage radiographs acquired every 30 s between the two CBCT scans. Corrections based on soft-tissue registration were evaluated offline by aligning the prostate contours defined on both planning CT and CBCT images. RESULTS For ideal rigid phantoms, CBCT image-guided treatment can usually achieve setup accuracy of 1 mm or better. For the patients, after CBCT correction, the target setup error was reduced in almost all cases and was generally within +/-1.5 mm. The image guidance process took 23-35 min, dictated by the computer speed and network configuration. The contribution of the intrafraction motion to the residual setup error was small, with a standard deviation of +/-0.9 mm. The average difference between the setup corrections obtained with coil and soft-tissue registration was greatest in the superoinferior direction and was equal to -1.1 +/- 2.9 mm. CONCLUSION On the basis of the residual setup error measurements, the margin required after online CBCT correction for the patients enrolled in this study would be approximatively 3 mm and is considered to be a lower limit owing to the small intrafraction motion observed. The discrepancy between setup corrections derived from registration using coils or soft tissue can be due in part to the lack of complete three-dimensional information with the coils or to the difficulty in prostate delineation and requires further study.