Couch modeling optimization for tomotherapy planning and delivery

Abstract We sought to validate new couch modeling optimization for tomotherapy planning and delivery. We constructed simplified virtual structures just above a default setting couch through a planning support system (MIM Maestro, version 8.2, MIM Software Inc, Cleveland, OH, USA). Based on ionization chamber measurements, we performed interactive optimization and determined the most appropriate physical density of these virtual structures in a treatment planning system (TPS). To validate this couch optimization, Gamma analysis and these statistical analyses between a three‐dimensional diode array QA system (ArcCHECK, Sun Nuclear, Melbourne, FL, USA) results and calculations from ionization chamber measurements were performed at 3%/2 mm criteria with a threshold of 10% in clinical QA plans. Using a virtual model consisting of a center slab density of 4.2 g/cm3 and both side slabs density of 1.9 g/cm3, we demonstrated close agreement between measured dose and the TPS calculated dose. Agreement was within 1% for all gantry angles at the isocenter and within 2% in off‐axis plans. In validation of the couch modeling in a clinical QA plan, the average gamma passing rate improved approximately 0.6%–5.1%. It was statistically significant (P < 0.05) for all treatment sites. We successfully generated an accurate couch model for a TomoTherapy TPS by interactively optimizing the physical density of the couch using a planning support system. This modeling proved to be an efficient way of correcting the dosimetric effects of the treatment couch in tomotherapy planning and delivery.


| INTRODUCTION
Carbon-fiber flat top couches are widely used for radiotherapy. 1 These couches have heterogeneous absorption properties when beams pass through the couch before entering the patient. [2][3][4][5][6][7][8][9][10][11][12][13][14][15][16][17][18] Several authors have reported that the failure to factor in couch attenuation for beams sent in the posteroanterior direction can cause a reduction in target volume coverage. [10][11][12] The American Association of Physicists in Medicine (AAPM) task group report 176 recommends that the beam intensity attenuation by the couch should be taken into account by the treatment planning system (TPS). 8 From its earliest version, TomoTherapy (Accuray, Sunnyvale, CA, USA) planning software has implemented this such that this virtual couch has appropriate predefined physical densities and is commissioned sufficiently. TomoTherapy, with about 50% rear dose contribution, the passing rate of gamma analysis deteriorated to 91.28% (3%/3 mm) even when using a two-dimensional array ion chamber device with angular dependence correction. They concluded that in pretreatment plan verification, the greater the dose contribution from the rear, the poorer the agreement between the measured dose and TPS. 19 Similar to their report, we have experienced discrepancies between actual measurement values and planned values in IMRT verification, especially when using the TomoDirect plan, which sends some fixed gantry-angle beams through the couch in the PA direction.
The dose output stability of the newest generation of TomoTherapy delivery systems was achieved with the addition of a dose servo-controlled system called DCS. Smilowitz et al. reported that the standard deviation in the monitor chamber daily output varies < 0.5%, making it unlikely that this is the cause of these dose discrepancies. 20 We suspected that default settings of the couch may be inaccurate. However, it is not currently possible to override the predefined physical density of the couch.
Here, therefore, we developed and validated a new couch modeling optimization for tomotherapy planning and delivery. Forward planning (included in TomoDirect) mode with a reference field size of 10 × 5 cm 2 was used for verification of the couch model in TPS in order to calculate actual couch attenuation. The beam angles used in the model validation were 0°and from 120°to 180°in 5°increments. The prescribed dose was 2 Gy per beam at the isocenter. The calculation grid size, field width, and pitch were 1.36 mm × 1.36 mm × 2.0 mm (equivalent to 1 voxel), 5.0 cm, and 0.500, respectively. The reason that we use a pitch of 0.500 is that the maximum pitch allowed for TomoDirect settings when using a where D is the absorbed dose.

2.B | Simplified couch modeling optimization in TPS
The To minimize the geometrical influence of patient and couch, we adopted an ultra-thin 1-pixel-thick slab in the virtual structures. The shape of these structures is shown in Fig. 1. The thick part of the center and thin part of both sides were separated.
Based on the ionization chamber measurements, we performed interactive optimization and determined the most appropriate physical density of these virtual structures in the TPS. We defined the dose difference by the following Eq: where D is the absorbed dose.

3.A | Ion chamber measurements for verification of TPS accuracy
As shown in Table 1, attenuation without couch correction ranged from 0.9% to 9.9%, depending on the beam angles. The greatest attenuation was observed at the gantry angle of 150°, where the beam is passing through the flat flex circuits. The conventional couch modeling narrowed the range to 0.2%-5.4%.

3.B | Simplified couch modeling optimization in TPS
Modeling of the TomoTherapy couch top in the TPS using different combinations of the center slab and both side slabs is presented in Table 1 for a 10 cm × 5 cm field size. Our modeling decreased the dose difference to < 1.0% when the center slab of the virtual structure was assigned a physical density of 4.2 g/cm 2 and both side slabs were 1.9 g/cm 2 (Fig. 4).

3.C | Validation of the couch modeling in an offaxis plan
As shown in Fig. 5(a) Table 2, the average gamma passing rate between the measurements and calculations is improved by around 0.6%-5.1%, depending on treatment site, and was statistically significant at all sites. As shown in Fig. 6(a), the lower the gamma passing rate without couch modeling optimization was, the greater improvement was obtained (R 2 = 0.912 P = 0.009). The largest improvement with APPA beams was a supraclavicular plan in a TomoDirect plan of 5.8%. Likewise, the largest improvement with APPA-modulated beams was a mediastinal TomoHelical plan of 5.1%. Figure 6(b) shows the results before and after couch optimization in the right supraclavicular area. The underdosed areas were improved by couch modeling optimization.

3.E | Surface dose evaluation for couch modeling
As shown in Fig. 7, due to passing through the couch, the depth of maximum dose was shifted from 11 mm to about 3 mm, and the surface dose was increased from about 35%-40% to 98%. The calculated PDD with couch optimization was good agreement (within 1 mm) with that of the default settings of the couch. at a 130°gantry angle with 6 MV photon beams. 14 Smith and colleagues examined the dosimetric properties of the iBEAM evo couch (Medical Intelligence, Schwabmünchen, Germany). 15 Their ionization chamber measurements showed beam attenuation ranging from 2.7% to a maximum of 4.6% for a 6 MV beam. As shown in Table 1, our results showed similar or slightly greater attenuation than those reported in their study, confirming that the accuracy of the couch modeling by the manufacturer was inaccurate.
We were able to obtain the best agreement between measured and calculated doses with the couch modeling optimization. The level of agreement was < 1.0% at the isocenter, and all dose T A B L E 2 Mean gamma passing rate with and without couch modeling optimization at each treatment site  21 This confirms that the combination of the couch structure and assigned physical density adopted in this study are reasonable for correcting the couch effects using the TPS.
Using the results of the ArcCHECK-generated patient QA plan, we found that the influence of the type of couch was greater with the plans having fewer beam ports, such as for supraclavicular nodal irradiation, or with intense AP/PA beams such as that for mediastinal irradiation for the purpose of avoiding normal lung. These results support the fact that the couch greatly affects the dose verification.
The few TomoHelical plans in which deterioration of the gamma passing rate was observed were a case in which the dose difference shifted to the plus side due to accuracy of the TPS so that it over- According to the result of Section 3.E, the difference of surface dose with and without couch modeling optimization was negligible.
Therefore, introducing this method to areas closer to couch is feasible.
A limitation of this study is that the model cannot accurately reproduce the geometric relationship between the couch and target.
Therefore, fluctuations in dose discrepancy were observed, especially at the interface of these structures. In addition, a relatively large dose fluctuation was found in beam angles that passed through the flat flex circuits. This might have been due to the failure to factor in the composition of the copper foil and optical cable assembly in the dose calculation. Nevertheless, it is < 2.0%, and therefore at an acceptable level.
In addition, it is necessary to be mindful of dose calculation grid size, as the traditional TomoTherapy TPS planning station downsamples the CT image size from 512 × 512 to 256 × 256 even when using the finest grid size, so the optimal physical density may slightly differ from our report. Therefore, it is recommended that facilities independently investigate this.
Indexed patient immobilization systems are now commonly used to establish reproducible patient positions relative to the couch, and employing such devices provides the best opportunity to accurately account for the couch top during the planning process. To ensure accurate measurement results for off-axis situations such as that mentioned in Section 2.C, it is necessary to use a robust immobilization system. Until now, these acceptance tests and the commissioning test did not include couch modeling validation. Therefore, extra testing by the user is essential for safe delivery of tomotherapy to patients.
This report is the first study to carry out couch modeling optimization for tomotherapy. The use of the planning support system made it easy to implement the virtual structure stably to the conventional couch. Our method in this study is very simple, and can be implemented easily at any site.

| CONCLUSION
We generated an accurate couch model for the TomoTherapy planning system by interactively optimizing the physical density of the couch using a planning support system. This modeling proved to be an efficient way of correcting the attenuation effects of the treatment couch in tomotherapy planning and delivery.

CONFLI CT OF INTEREST
The authors have no conflicts of interest to disclose.  Due to passing through the couch, the depth of maximum dose shifted from 11 mm to about 3 mm, and the surface dose increased from about 35%-40% to 98%. Calculated PDD with couch optimization was in good agreement (within 1 mm) with that of the default settings of the couch