A Personalized Anisotropic Margin for Cervical Cancer Radiation Therapy Under the Guidance of Daily Iterative Cone-Beam Computed Tomography (iCBCT)

Online adaptive radiation therapy (ART) eliminates interfraction uncertainties by adaption before each treatment session. However, intrafraction motions still exist and could become more severe due to long treatment time. Large isotropic margins can ensure clinical target volume (CTV) coverage but at the cost of more organs at risk damage. In this study, we proposed a novel personalized anisotropic margin search algorithm for cervical cancer radiation therapy under the guidance of daily iterative cone-beam computed tomography (iCBCT) to find the optimal margin values for each patient, which achieves the smallest possible planning target volume (PTV) and maintains CTV coverage. Twenty-two online Ethos ART treatment sessions were included for analysis. Two iCBCT scans were taken in each session. The first one (iCBCT(1)) was taken after positioning, and the second one (iCBCT(2)) was taken before beam delivery. Corresponding CTV1 and CTV(2 )were contoured in the two scans. In each session, minimal isotropic margins were first searched by iteratively increasing the magnitude until the resulting PTViso covers 99% of CTV2. Afterward, the margin values in all six directions were decreased iteratively until CTV2 coverage was smaller than 99% to get the personalized margin and target volume PTVint. In addition, the uterus was considered separately, and different margins were found for it and the remaining CTV, respectively, to reduce the target volume of PTV(sep )further. PTViso, PTVint, and PTVsep were compared in terms of CTV2 coverage and absolute volume. The algorithm successfully generated PTViso, PTVint, and PTVsep for all online ART treatment sessions. The mean +/- SD values for PTViso5mm, PTViso10mm, PTViso15mm, PTVint, and PTVsep were 1,074.0 +/- 78.1, 1,519.5 +/- 100.4, 2,006.4 +/- 122.5, 929.3 +/- 73.4, and 845.1 +/- 72.5 mL, respectively. The volume difference between PTVint and PTVsep was significant (p < 0.001). All the PTVs ensured an average coverage larger than 99%, and the differences between any two PTVs were insignificant. This study proposed a novel personalized anisotropic margin search algorithm for cervical cancer online ART. Compared to the conventional 5 or 10 mm isotropic margins, the personalized anisotropic margin reduced PTV volume by 13.5% and 38.8%, respectively; if the uterus was considered separately, the volume can be further reduced by 21.3% and 44.3%, respectively, while CTV coverage was still maintained. This algorithm could reduce target volume and potentially spare normal tissue better than isotropic margin expansion.