Adaptive radiotherapy based on statistical process control for oropharyngeal cancer

Abstract Purpose The purpose of this study is to quantify dosimetric changes throughout the delivery of oropharyngeal cancer treatment and to investigate the application of statistical process control (SPC) for the management of significant deviations during the course of radiotherapy. Methods Thirteen oropharyngeal cancer patients with daily cone beam computed tomography (CBCT) were retrospectively reviewed. Cone beam computed tomography images of every other fraction were imported to the Velocity software and registered to planning CT using the 6 DOF (degrees of freedom) couch shifts generated during patient setup. Using Velocity “Adaptive Monitoring” module, the setup‐corrected CBCT was matched to planning CT using a deformable registration. Volumes and dose metrics at each fraction were calculated and rated with plan values to evaluate interfractional dosimetric variations using a SPC framework. T‐tests between plan and fraction volumes were performed to find statistically insignificant fractions. Average upper and lower process capacity limits (UCL, LCL) of each dose metric were derived from these fractions using conventional SPC guidelines. Results Gross tumor volume (GTV) and organ at risk (OAR) volumes in the first 13 fractions had no significant changes from the pretreatment planning CT. The GTV and the parotid glands subsequently decreased by 10% at the completion of treatment. There were 3–4% increases in parotid mean doses, but no significant differences in dose metrics of GTV and other OARs. The changes were organ and patient dependent. Control charts for various dose metrics were generated to assess the metrics at each fraction for individual patient. Conclusions Daily CBCT could be used to monitor dosimetric variations of targets and OARs resulting from volume changes and tissue deformation in oropharyngeal cancer radiotherapy. Treatment review with the guidance of a SPC tool allows for an objective and consistent clinical decision to apply adaptive radiotherapy.

Fractional doses of 190, 180, and/or 165 cGy were also prescribed for additional target volumes of various risks containing microscopic diseases. All plans were generated from treatment planning CT acquired on a SOMATOM Definition AS CT scan (Siemens, Munich, Germany) typically with 205 mA, 120 kVp, 1.3 × 1.3 mm axial resolution and 3-mm-slice thickness. Varian Eclipse treatment planning system (TPS) (Eclipse 13.7, Varian Medical System, Inc.) was used for planning and optimization.
All the patients were immobilized with thermoplastic head masks and treated on Varian TrueBeam machines with image guidance of the on-board kV-CBCT. CBCT images were acquired at each fraction prior to irradiation using a standard Head CBCT mode with parameters of 100 kV, 15 mA, Full fan. and half trajectory. CBCT volume (resolution: 0.05 × 0.05 × 0.2 cm) covers anatomy from brainstem to neck with a scanning length of approximate 19 cm. The daily CBCT is subsequently registered to planning CT for patient setup with 6 DOF couch shifts. The registration is focused on alignment of GTV/ CTV target, and reviewed/approved by physicians following a standard verification protocol which includes on-line review prior to delivery of first fraction and off-line review before procedure of following treatment fractions. In addition, physicists also review the registrations on a weekly base and communicate with therapists and physicians for any concerns on the setup errors. were also calculated using the deformed structures in CBCT and planned dose projected to CBCT, under the assumption that daily volumetric dose distribution had not significant change. 10 Thereby the change in volumes and DVH metrics for selected structures during the treatment course were reconstructed for statistical analysis.

T-tests between target volumes (GTV) in CT plans and fractional
CBCTs showed no significant volume changes until the 13th fractions (or 17 days from first treatment) on the whole cohort of patients under study. Using the data from the first 17 days, we calculated average and standard deviation (STD) of each dose metric, and created control chart for the metric with upper control limit average variations of volume (UCL: average + 3 × STD) and lower limit (LCL: average-3 × STD). Upper control limit and LCL can be specific on patient cohort used, clinical procedure followed, and structures of concern applied. Table 2 shows the mean, UCL and LCL of target and OAR dose metrics based on our patient cohort of this study.

| RESULTS
The change in the shape and volume of each structure was observed from the translational deviation of the center of a structure over the course of treatment. Summarized in Table 1      The clinical concern is lower tumor control and higher normal tissue complications associated with anatomic changes. Very few studies presented the quantitative correlation of the changes in tumor Zhang et al. 19 retrospectively investigated some patients in their clinic with and without an intervention of ART. Greater than 5% difference in the dose to the CTV or <0.75 Pearson correlation coefficient of the CTV was found to be the action level related to the clinical decision made for ART. 20 This study has made one step forward for developing the clinically relevant criteria, although the study does not have meaningful outcomes data to support their action levels. In addition, their decision-making criteria for ART were solely based on the dosimetric or geometric deviations of the target, which is normally less of an issue than those of the OARs. 11 Our study using statistical process control is the approach to the clinically tangible criteria that would allow for the decision-making based on the anticipated outcomes of concern. Thresholds of both high limit (HL) and low limit (LL) are generated for each dose metric of targets and OARs of a patient. Furthermore, a global HL and LL could be established by building SPC in a patient cohort. Optimal UCL and LCL should be developed with more patients and best be correlated with the clinical evidence. The SPC could provide a guidance for the action level when resimulation and replanning is expected.
As the start point to implement the procedure, clinicians can test their own patient cohort. Dose metrics for the structures of concern should be monitored to understand any trending. A clinical procedure can be worked out with the data and experience they have  19 The control chart provides a statistical yet objective baseline for clinicians to consider any necessary interventions along the course of treatment. This method can ultimately be refined to associate the action levels in dosimetric changes with the clinical outcomes.
The majority of this study involved deformable registration of daily CBCT with planning CT, from which the volume of structures on the treatment day was derived. Accuracy of the registration varies with the degree of anatomical changes and the quality of CBCT.
The commercial deformable image registration in Velocity has been assessed by various studies using virtual phantoms, thorax phantom, and patient data. 20,21 We also visually checked anatomical alignments from the registrations for at least one CBCT in a week treatment delivery. Nevertheless, separate quality assurance of the CBCT and CT deformable registration is necessary and an automated method may aid in the procedure. 22 In the current workflow of AMN, the change of dose distribution in each fraction is ignored when DVHs are reconstructed for evaluation. The assumption that the dose volume did not change with respect to the treatment plan during the course of treatment played a critical role in the reconstruction of DVHs of structures at treatment. However, the assumption itself is not always valid throughout the treatment course. As the patient anatomy changes, the external body surface, as well as the internal relative location of organs, deviated away from the treatment plan. This interfractional change of the patient anatomy altered the HU/density distribution in the spatial domain of the treatment plan. Therefore the actual fractional dose distribution, which was calculated based on the underlying anatomy, could be different from the planned dose volume. The impact of the change of dose distribution can be significant. To improve the accuracy of the adaptive DVHs, dose volume needs to be recalculated using the anatomy information in the CBCT. One of the problems preventing us from performing dose recalculation directly on CBCTs was the lack of CBCT HU/density calibration curve. Future work is underway to implement CBCT-based dose recalculation using CBCT to synthetic CT conversion 23 or CBCT HU override. 24

| CONCLUSIONS
Daily CBCT could be used to monitor dosimetric changes in both targets and OARs due to the volumetric changes and organ deformation during the course of oropharyngeal HN cancer radiotherapy.
We developed a method with statistical process control, which can be used to establish the clinical criteria for ART by analyzing the correlation of dosimetric changes with the outcomes data. Treatment review with guidance of an SPC tool enable radiation oncologists to objectively and consistently identify the fractions with dosimetric changes that are of clinical significance.

CONFLI CT OF INTEREST
None of the authors has conflict of interest or funding to disclose related to the work of this publication.