BackgroundThe purpose of this study was to evaluate the impact of markerless on-board kilovoltage (kV) cone-beam computed tomography (CBCT)-based positioning uncertainty on determination of the planning target volume (PTV) margin by comparison with kV on-board imaging (OBI) with gold fiducial markers (FMs), and to validate a methodology for the evaluation of PTV margins for markerless kV-CBCT in prostate image-guided radiotherapy (IGRT).MethodsA total of 1177 pre- and 1177 post-treatment kV-OBI and 1177 pre- and 206 post-treatment kV-CBCT images were analyzed in 25 patients who received prostate IGRT with daily localization by implanted FMs. Intrafractional motion of the prostate was evaluated between each pre- and post-treatment image with these two different techniques. The differences in prostate deviations and intrafractional motions between matching by FM in kV-OBI (OBI-FM) and matching by soft tissues in kV-CBCT (CBCT-ST) were compared by Bland-Altman limits of agreement. Compensated PTV margins were determined and compensated by references.ResultsMean differences between OBI-FM and CBCT-ST in the anterior to posterior (AP), superior to inferior (SI), and left to right (LR) directions were − 0.43 ± 1.45, − 0.09 ± 1.65, and − 0.12 ± 0.80 mm, respectively, with R2 = 0.85, 0.88, and 0.83, respectively. Intrafractional motions obtained from CBCT-ST were 0.00 ± 1.46, 0.02 ± 1.49, and 0.15 ± 0.64 mm, respectively, which were smaller than the results from OBI-FM, with 0.43 ± 1.90, 0.12 ± 1.98, and 0.26 ± 0.80 mm, respectively, with R2 = 0.42, 0.33, and 0.16, respectively. Bland-Altman analysis showed a significant proportional bias. PTV margins of 1.5 mm, 1.4 mm, and 0.9 mm for CBCT-ST were calculated from the values of CBCT-ST, which were also smaller than the values of 3.15 mm, 3.66 mm, and 1.60 mm from OBI-FM. The practical PTV margin for CBCT-ST was compensated with the values from OBI-FM as 4.1 mm, 4.8 mm, and 2.2 mm.ConclusionsPTV margins calculated from CBCT-ST might be underestimated compared to the true PTV margins. To determine a reliable CBCT-ST-based PTV margin, at least the systemic error Σ and the random error σ for on-line matching errors need to be investigated by supportive preliminary FM evaluation at least once.