Automatic measurement of air gap for proton therapy using orthogonal x‐ray imaging with radiopaque wires

Abstract Purpose The main objective of this study was to develop a technique to accurately determine the air gap between the end of the proton beam compensator and the body of the patient in proton radiotherapy. Methods Orthogonal x‐ray image‐based automatic coordinate reconstruction was used to determine the air gap between the patient body surface contour and the end of beam nozzle in proton radiotherapy. To be able to clearly identify the patient body surface contour on the orthogonal images, a radiopaque wire was placed on the skin surface of the patient as a surrogate. In order to validate this method, a Rando® head phantom was scanned and five proton plans were generated on a Mevion S250 Proton machine with various air gaps in Varian Eclipse Treatment Planning Systems (TPS). When setting up the phantom in a treatment room, a solder wire was placed on the surface of the phantom closest to the beam nozzle with the knowledge of the beam geometry in the plan. After the phantom positioning was verified using orthogonal kV imaging, the last pair of setup kV images was used to segment the solder wire and the in‐room coordinates of the wire were reconstructed using a back‐projection algorithm. Using the wire as a surrogate of the body surface, we calculated the air gaps by finding the minimum distance between the reconstructed wire and the end of the compensator. The methodology was also verified and validated on clinical cases. Results On the phantom study, the air gap values derived with the automatic reconstruction method were found to be within 1.1 mm difference from the planned values for proton beams with air gaps of 85.0, 100.0, 150.0, 180.0, and 200.0 mm. The reconstruction technique determined air gaps for a patient in two clinical treatment sessions were 38.4 and 41.8 mm, respectively, for a 40 mm planned air gap, and confirmed by manual measurements. There was strong agreement between the calculated values and the automatically measured values, and between the automatically and manually measured values. Conclusions An image‐based automatic method has been developed to conveniently determine the air gap of a proton beam, directly using the orthogonal images for patient positioning without adding additional imaging dose to the patient. The method provides an objective, accurate, and efficient way to confirm the target depth at treatment to ensure desired target coverage and normal tissue sparing.

method provides an objective, accurate, and efficient way to confirm the target depth at treatment to ensure desired target coverage and normal tissue sparing. affecting surrounding healthy tissue is an important mission in radiation oncology. 1 The advantage of using proton therapy over photon therapy is that proton beams can deposit high dosage of radiation exclusively over a region at a certain depth below the skin surface and the normal tissue beyond the distal end of the beam receives minimal dose. [2][3][4] This physical characteristic of proton beams requires high precision in determining the target depth. 5 Any substantial changes of the target depth, such as the external body contour changes due to weight loss, may lead to potential target miss or significant damage to the normal tissues which would otherwise be spared. 6,7 To ensure accurate setup during each fractioned treatment and target depth, the air gap from the patient's skin surface must be measured and checked after patient positioning and prior to the beam delivery to avoid potential mistreatment.
With the exception of limited number of proton centers equipped with three-dimensional (3D) imaging positioning systems, 8 the two-dimensional (2D) orthogonal imaging-based system is still widely employed at many proton centers for the purpose of patient positioning. Unlike the 3D imaging system which can directly provide target depth information, with the 2D imaging system, the target depth information can only be indirectly derived by measuring the air gap between patient surface and beam nozzle prior to treatment.
Currently, measurement of the air gap is mostly manually conducted using a ruler, and the measurement can be subjective and inefficient.
It would be desirable to utilize the same imaging process for patient treatment position purpose to determine the air gap, at no expense of additional imaging time and dose, even for the proton machines equipped with only two-dimensional kilo-voltage (kV) imaging systems. The goal of this study is to develop an orthogonal imagingbased automatic method for the determination of the air gap at proton therapy treatment.

| MATERIALS AND METHODS
The method is based on the idea that coordinates of a linear object imaged on the orthogonal images can be reconstructed. 9 Since the air gap is measured from the patient skin surface and it is normally difficult to clearly identify soft tissue on x-ray images, a radiopaque wire can be placed along the patient skin surface during the patient treatment setup in the general location where the proton beam is aimed at (this wire will be removed prior to beam delivery). The developed method was first validated using a Rando ® phantom on a Mevion S250 proton therapy machine, as shown in Fig. 1(a), and subsequently validated in two consecutive treatment sessions for a proton patient.
An illustration of the setup of the x-ray sources and images, as well as the projection of the wire is shown in Fig. 2 C ¼ x a y b y a ; Exceptions to this algorithm are needed, for example, when y a , y b , or both are equal to zero. These exceptions are added into the program separately. We follow this algorithm for all points identified for the centerlines of wire projected on both films until we have reconstructed the centerline of the wire in three-dimensional space.
We align the points on the wire and find the shortest distance from the wire to the end of the snout or compensator tray. The code searches through the newly reconstructed points on the wire, calculates the distance from each of these points to the compensator tray, and returns the smallest value.
Using the above discussed algorithm, our computer program takes the two-dimensional LLAT and PA positioning images, as well as the planning RT Structure set and treatment RT plan as inputs. F I G . 2. Setup and matching of corresponding points for two orthogonal films and relative locations of the kV panels and sources. 9 Note. The actual coordinate system used in DICOM needs to be rotated to match the coordinate system shown above.
The outputs of the computer program consist of the coordinates of the points on the wire, the calculated air gap using the points on the wire, the planned air gap using the corresponding DICOM body contour sequence, and a three-dimensional plot of the body contour, the wire placed on top of it, the isocenter, and the compensator tray or snout. without the surface contour is given in Fig. 3(b). The derived air gap results are shown in

| CONCLUSIONS
An orthogonal image-based method has been developed to automatically and objectively derive the beam air gap values in proton treatment, without adding additional imaging time and doses. Knowing the desired air gap during treatment allows us to confirm that the target depth will not exhibit significant change and will be properly covered by the proton dose and that the normal tissues along the beam path will not be unexpectedly compromised. The backprojection algorithm and method described above is experimentally proven to be accurate. It is simpler, more efficient, and more easily accessible and reproducible than most other surface localization techniques currently available to proton centers equipped with only 2D imaging systems. Although a three-dimensional positioning imaging device may be the ultimate approach to automatically compute air gaps and therefore the target depth, this developed method can be useful and beneficial for both the patient and the radiation oncology team for the proton systems which still rely on two-dimensional imaging systems for patient treatment positioning.

CONFLI CT OF INTEREST
The authors have no relevant conflict of interest to disclose.