Sensitivity of array detector measurements in determining shifts of MLC leaf positions

Abstract Using a MatriXX 2D ionization chamber array, we evaluated the detection sensitivity of systematically introduced MLC leaf positioning shifts to test whether the conventional IMRT QA method can be used for quality assurance of an MLC tracking algorithm. Because of finite special resolution, we first tested whether the detection sensitivity was dependent of the locations of leaf shifts and positions of ionization chambers. We then introduced the same systematic leaf shifts in two clinical intensity modulated radiotherapy plans (prostate and head and neck cancer). Our results reported differences between the measured planar doses with and without MLC shifts (errors). Independent of the locations of the leaf position shifts and positions of the detectors, for the simple rectangular fields, the MatriXX was able to detect ±2 mm MLC leaf positioning shifts with Gamma index of 3%/3 mm and ±1 mm MLC leaf position shifts with Gamma index of 2%/2 mm. For the clinical plans, measuring the fields individually, leaf positioning shifts of ±2 mm were detected using Gamma index of 3%/3 mm and a passing rate of 95%. When the fields were measured compositely, the Gamma index exhibited less sensitivity for the detection of leaf positioning shifts than when the fields were measured individually. In conclusion, if more than 2 mm MLC leaf shifts were required, the commercial detector array (MatriXX) is able to detect such MLC positioning shifts, otherwise a more sensitive quality assurance method should be used.

adjustments of the endorectal balloon in order to reduce prostate rotation and deformation. 6 Without endorectal balloon, the intrafraction prostate motion is sporadic, depending on the treatment duration, 7,8 and thus compensating the intra-fraction prostate motion is treatment modality-dependent. Based on the data collected from the Calypso system, if the treatment duration is less than 4 min, 2 mm planning margins in the longitudinal and vertical directions were proposed with the assumption of a minimal lateral intrafraction motion. 7 To further reduce the planning margin to account for interfraction prostate deformation or rotation larger than 3°(a typical range a robotic table can compensate for), some researchers proposed real-time adaptive planning, 9 which may encounter logistical and practical challenges. 10 Others proposed to shift MLC leaves to track changes of the prostate. To account for the independent prostate and pelvic lymph nodal movement, our group also proposed an MLC tracking method to compensate for interfraction prostate motion. 11 To account for intrafraction prostate motion, a real-time monitoring with frequent patient repositioning or with MLC tracking during treatment has been proposed. A prototype MLC tracking system has been developed and clinically tested by Keall group. 2,4 To experimentally validate the MLC tracking algorithm for interfraction or intrafraction motion, we designed this study to test whether a commonly used detector array device is sufficient to verify the MLC tracking algorithm.

2.A | MLC tracking algorithm
An MLC tracking algorithm was proposed by our group to compensate for independent movement of the prostate and pelvic lymph nodes (PLN). With this MLC tracking method, the displacement of the prostate was compensated without affecting the dose distributions to PLN. Briefly, the algorithm was implemented with an inhouse program that automatically identifies MLC leaf pairs that were collimated to the prostate in the planning CT and adjusts the positions of these leaf pairs for each segment of the IMRT plan to compensate for the interfraction prostate motion relative to the pelvic bones. Meanwhile, the MLC leaves that were conformal to the PLN were unchanged. Based on the magnitude and direction of the daily prostate movement, the algorithm adjusts the positions of selected MLC leaf pairs to follow the translational motion of the prostate for each beam. The algorithm assumes that the prostate is a rigid body and the rotational motion is negligible. Because the field size in unchanged and the changes in the off-axis factors contribute only a negligible amount to the dose distribution, this MLC tracking does not require a real-time dose calculation.

2.B | Testing fields
A single rectangular field (2.4 9 10.4 cm 2 ), one 5-field prostate IMRT plan, and one 9-field head and neck IMRT plan, were used.
The single rectangular field represents a "best case" for the Gamma index to detect the leaf positioning shifts, as opposed to the more complicated IMRT plans. It is also used to assess the effect of the MatriXX array resolution. The IMRT plans were created using direct machine parameter optimization (DMPO) with beam energy of 6 MV within the Pinnacle treatment planning system (Pinnacle 9.0, Philips Radiation Oncology System, Madison, WI, USA). The prescribed fractional dose for the two IMRT plans was 2 Gy to the clinical target volume (CTV).

2.C | MLC leaf positioning shifts
Within Pinnacle, the positions of the MLC leaves on one side of the leaf bank in each segment were shifted by AE1, AE2, AE3, or AE4 mm, while keeping the rest of the treatment plan parameters unchanged.
Nine plans were created for each patient: the original plan without leaf positioning shifts, and eight plans with increasing systematic shifts. All these plans were measured and compared.

2.D | Measurements
All dose measurements were conducted with the MatriXX 2D array, which consists of 1020 air-vented parallel plate ionization chambers on a 32 9 32 Cartesian grid (there are no ion chambers at the four corners of the array). The physical and dosimetric properties of this device have been reported previously. [12][13][14][15][16][17] The chamber center-tocenter separation is 0.762 cm, with a sensitive volume of 0.08 cm 3    locations between the two adjacent ion chamber columns at 1.143 cm and 0.381 cm away from the in-plane center of the MatriXX. This is also the reason we chose 2.4 cm as the field width (so 1.2 cm half width includes two ion chamber columns). We chose 26 pairs of MLC symmetrically with 4 mm leaf width, so the field height was 10.4 cm. Gamma index was computed between the measured planar doses with and without shifts for each field. A threshold of 10% of the maximum dose was used to exclude areas with low dose from Gamma index calculation. Planar dose distributions from measurements were interpolated to a resolution of 1 by 1 mm 2 within the OmniPro I'mRT analysis software to match those from the TPS. Table 1 shows Gamma indices of 3%/3 mm and 2%/2 mm com-

3.B | Effect of ion chamber resolution and position
The 2%/2 mm Gamma index passing rate was observed to be consistent at (82.7 AE 1.8) % regardless of the relative position of the À1 mm leaf positioning shift to the detectors.

3.C.2 | IMRT prostate plan at 0°fixed gantry angle
To eliminate the angular dependence of the MatriXX array, dose distributions were measured at 0°fixed gantry angle. Table 2 and  Table 2 shows that detecting a 2 mm MLC leaf positioning shift in prostate patient-specific IMRT QA with MatriXX requires a Gamma index of 3%/3 mm and a tight passing rate of 95% for individual fields. Table 2 and Fig. 2(b) also shows that the Gamma index of the composite field was generally greater than the average Gamma index of individual fields and thus less sensitive to leaf positioning shifts.   decrease). Figure 5 shows

| DISCUSSION AN D CONCLUSION
In this study, sensitivity of detecting MLC leaf positioning shifts was assessed using the MatriXX ionization chamber array, which is a con-    show that little improvement could be achieved from increasing the dose grid resolution.
A threshold of 10% of the maximum dose within the field was used to limit the points considered in the Gamma analysis. More specifically, Gamma passing rate represents the percentage of points T A B L E 3 The average and standard deviation (SD) Gamma indices of individual fields and Gamma indices of the composite field of 3%/3 mm and 2%/2 mm computed between the measured planar doses with MLC leaf positioning shifts up to 4 mm and the calculated doses without shifts of the head and neck IMRT plan with all beams at 0°gantry angle.

CONF LICT OF I NTEREST
The authors declare no conflict of interest.