Real‐time intrafraction prostate motion during linac based stereotactic radiotherapy with rectal displacement

Abstract Background Kilovoltage Intrafraction Monitoring (KIM) is a method which determines the three‐dimensional position of the prostate from two‐dimensional kilovoltage (kV) projections taken during linac based radiotherapy treatment with real‐time feedback. Rectal displacement devices (RDDs) allow for improved rectal dosimetry during prostate cancer treatment. This study used KIM to perform a preliminary investigation of prostate intrafraction motion observed in patients with an RDD in place. Methods Ten patients with intermediate to high‐risk prostate cancer were treated with a Rectafix RDD in place during two boost fractions of 9.5–10 Gy delivered using volumetric modulated arc therapy (VMAT) on Clinac iX and Truebeam linacs. Two‐dimensional kV projections were acquired during treatment. KIM software was used following treatment to determine the displacement of the prostate over time. The displacement results were analyzed to determine the percentage of treatment time the prostate spent within 1 mm, between 1 and 2 mm, between 2 and 3 mm and greater than 3 mm from its initial position. Results KIM successfully measured displacement for 19 prostate stereotactic boost fractions. The prostate was within 1 mm of its initial position for 84.8%, 1–2 mm for 14%, 2–3 mm 1.2% and ≥3 mm only 0.4% of the treatment time. Conclusions In this preliminary study using KIM, KIM was successfully used to measure prostate intrafraction motion, which was found to be small in the presence of a rectal displacement device. Trial registration The Hunter New England Human Research Ethics Committee reference number is 14/08/20/3.01.


| BACKGROUND
Prostate cancer is estimated to have a low a/b ratio, indicating that hypofractionated treatment schedules may increase the effectiveness of treatment. 1 Delivery of higher doses through hypofractionation increases the risk of damage to healthy tissues surrounding the prostate, particularly the rectal wall. 2 The risk and severity of rectal toxicities have been correlated with the volume of rectal wall exposed to high doses of radiation. 3 The application of dose volume constraints in planning, along with daily image guidance to enable reduced margins, are the most effective ways to reduce rectal dose, but rectal displacement devices such as injected hydrogel and the Rectafix (Scanflex Medical AB, Tumstocksv€ agen, Sweden) rectal retractor are also useful in allowing for safe dose escalation. 4 The Rectafix system (Fig. 1) uses a rod inserted into the patient's rectum and then gently depressed posteriorly and fixed in place, guided by patient tolerance, thereby manually moving the rectum away from the prostate. The Rectafix provides an average increase in separation of 0.5 cm between the anterior rectal wall and posterior prostate border, and may assist in immobilizing the rectal wall by preventing changes in filling by gas or feces. 5 Intrafraction prostate motion has the potential to reduce the dose coverage of the prostate and to increase the dose received by organs at risk. A variety of methods exist for monitoring the position of the prostate during treatment, including megavoltage (MV) imaging, 6 ultrasound, 7 combined MV and kilovoltage (kV) imaging, 8 Calypso electromagnetic guidance, 9 the BrainLAB ExacTrac x-ray system, 10 the Cyberknife platform, [11][12][13] and Navotek radioactive fiducials. 14 Several of these methods require additional equipment not available on a standard linear accelerator, are costly, and require further expertise to implement.
Kilovoltage Intrafraction Monitoring (KIM) takes advantage of the gantry-mounted kV imager available on many modern linear accelerators to determine the position of the prostate in three dimensions from 2D kV projections using a probability density function. 15,16 The geometrical accuracy of KIM has been established, and the software has successfully been used to measure prostate displacement during treatment in noninterventional 17 and interventional studies. 18 In this study, KIM was used to perform a preliminary investigation of the magnitude of intrafraction motion in a series of clinical trial patients receiving Stereotactic Body Radiation Therapy (SBRT) to the prostate with a Rectafix in place.

2.A | Quality assurance of KIM software
The KIM software was used offline to analyze kV images acquired during delivery to determine prostate motions. Quality assurance (QA) of the KIM software was performed on the Varian Clinac iX machine.
The tests used were those suggested for KIM QA by Ng et al,19 which were based on the recommendations for Calypso QA by Santanam et al 20 The tests suggested were a static localisation test, a dynamic test, a latency test, and a treatment interruption test. The latency test and treatment interruption test were not performed as KIM software was not intended for real-time use or gating during this study. The pass criteria applied for both the static and dynamic tests was 1.0 mm mean difference and 1.0 mm standard deviation between the KIM software trajectory and the trajectory output by the motion phantom.
Static localisation tests were performed to ensure that KIM was able to trace static offsets correctly and that all directions of the software and phantom coordinate systems were in agreement. Static localisation tests were performed using the CIRS 801-P Virtually Human Male Pelvis phantom (CIRS Inc., Norfolk, VA, USA) with an insert containing three cylindrical gold fiducial markers with dimensions 0.9 9 3 mm. The phantom was offset AE5 mm in each of the anteriorposterior (AP), left-right (LR), and superior-inferior (SI) directions.
A pre-arc of images taken during a rotation of 120°prior to treatment commencement and one partial treatment arc were delivered to the phantom at each position while kV images were acquired at 10 Hz.
The position determined by KIM was compared to the known static shift and the mean difference and standard deviation were determined.

2.B | Patient and treatment details
The PROstate Multicentre External beam radioTHErapy Using Stereotactic boost (PROMETHEUS) clinical trial is a hypofractionated F I G . 1. The Rectafix system. The system consists of a rectal retracting rod attached to a vertical column and locked onto a baseplate. A leg rest is also provided.

2.C | IGRT and KIM acquisition
Patients treated on the Clinac were first aligned using cone beam computed tomography (CBCT). Two-dimensional kV images were then acquired using the kV imager mounted on the gantry perpendicular to the treatment beam. Images were taken at 125 kV, 80 mA and 13 ms at 5 Hz with a 6 9 6 cm field size and 180 cm imager source to detector distance (SDD). The 180 cm SDD decreases the effect of MV scatter on the kV images. Images were taken over a gantry rotation of 120°immediately prior to treatment, and then as the gantry rotated during delivery of VMAT treatment. This 120°p re-arc was necessary to allow an earlier version of the KIM software to build its probability density model to track the fiducial marker positions. KIM software has since been updated so that pretreatment CBCT imaging can instead be used to build the model. Imaging during treatment was enabled using the service mode of the on-board imaging software, which is not currently possible in the clinical mode. Images were saved using a research framegrabber computer and in-house software. No real-time IGRT was used for these treatments.
Patients treated on the Truebeam were first aligned using kV/ kV matching to assess for gross RDD error or bowel gas. A full fan spotlight CBCT was then acquired for further alignment purposes and to enable the KIM software to build its probability density function. Images were acquired during delivery of VMAT treatment at 0.33 Hz with a 125 kV, 80 mA and 13 ms beam. A field size of 5 9 5 cm and an imager SDD of 180 cm was used for these patients.

2.D | Analysis
Following treatment, the pre-arc/CBCT images and the kV projections taken during treatment were processed offline using the KIM software. The KIM software automatically segments the location of each fiducial marker on each 2D kV projection, then reconstructs the 3D position of the markers by taking a maximum likelihood estimation of a 3D probability density function. 21 The displacement of the prostate in each of the AP, LR, and SI directions was quantified as a function of time throughout each fraction.

3.A | Quality assurance of KIM software
The results of both the static and dynamic localisation tests appear in Table 1, which shows the mean difference (x) and standard deviation (r) between the expected position of the phantom and the KIM measured position. All static localisation tests passed the criteria of <1 mm mean difference and <1 mm standard deviation. All dynamic tests passed the criteria of <1 mm mean difference and <1 mm standard deviation, apart from the high frequency and erratic trajectories, which both failed due to a standard deviation greater than 1 mm in the AP direction. These two motion trajectories represent the most extreme prostate motion, with high frequency, high amplitude motion, and are, therefore, the most difficult traces for KIM to track accurately.

3.B | Patient motion results
The KIM software gives displacement of the prostate in the AP, LR, and SI directions as a function of time throughout each fraction.
These results were analyzed to determine the percentage of time the prostate spent within 1 mm, between 1 and 2 mm, between 2 and 3 mm, and ≥3 mm from its initial position in each of the three directions. These results appear in Table 2. Table 3 displays the average and standard deviation for prostate displacement across all patients and fractions measured.
The majority of motion measured by the KIM system occurred in the AP and LR directions. The prostate was greater than 1 mm from its initial position in the SI direction only 2.8% of the treatment time (see Table 2); however, the average displacement was greatest in the SI direction, while still being sub-millimeter (see Table 3).

| DISCUSSION AND CONCLUSIONS
The KIM software was successfully used to segment fiducial markers and determine prostate motion during prostate SBRT boost treatments in 95% of fractions for ten patients with a Rectafix in place.
As such, it would appear to be a feasible approach to deploy clinically on either a Clinac or TrueBeam linear accelerator. KIM performs better when the incoming images have higher image quality this was observed for patients with a smaller distance across the hips treated on the Clinac, and for all patients treated on the Truebeam.     Bowel and bladder preparation combined with pre-treatment and mid-treatment imaging to assist positioning and assess bladder and rectal fullness is likely to be sufficient, as such small displacements were observed in this cohort so that the additional imaging dose provided by kV imaging can be avoided. Future work in this area is required to determine if the motion reduction is due to the Rectafix alone, or if the pre-treatment bowel and bladder preparation regime followed by patients or fast treatment times has the most impact on intrafraction prostate motion. The prostate was two or more millimeters from its initial position only

CONFLI CT OF INTEREST
The authors declare that they have no conflicts of interest.