Patient data and treatment planning

From 2009 to 2012, 28 stage T1-2 prostate cancer patients were enrolled in the IRB approved institutional SBRT clinical trial. Patients receive 5 fractions of radiation with each fraction size of 7.4 Gy and the total dose of 37 Gy. The 5 treatments are scheduled to be delivered every other day. A minimum of 36 hours and a maximum of 96 hours should separate each treatment. No more than 3 fractions will be delivered per week. The total duration of treatment will be no shorter than 10 days and no longer than 18 days.

Patients are asked to have a full bladder during simulation and daily treatment. Patients are given instructions to drink 16–24 oz of water or other fluid 2–3 hours prior to treatment and to not urinate between this time and treatment as they are able. For rectum filling management, patients are instructed to take one tablespoon of Milk of Magnesia the night and one Fleet’s enema 2–3 hours before the simulation and each treatment.

Computed Tomography (CT) is the primary image set for treatment planning. The simulation is performed in the supine treatment position, with immobilization device (alpha-cradle) to minimize body motion and rotation. Axial cuts of 1.25 mm are acquired throughout the pelvis for the accuracy of marker identification and localization. Magnetic resonance imaging (MRI) is also acquired and fused with CT images to assist target delineation. MRI is acquired prior to Calypso transponder implantation to exclude artifacts interfering with soft tissue contouring. For fiducial markers, MRI is acquired in the same day of planning CT to minimize patient visit to the clinic.

CT and MRI image sets are registered for the delineation of target volumes and OARs. The definition of volumes is in accordance with the ICRU Report #50 and ICRT Report #62: Prescribing, Recording, and Reporting Photon Beam Therapy [10, 11]. The gross tumor volume (GTV) for the purposes of this protocol is the prostate only, and is defined by the physician based on the planning CT and MR along with clinical information. The clinical target volume (CTV) is the same as the GTV. The planning target volume (PTV) is defined as the CTV plus a 3 mm posteriorly and 5 mm in all other dimensions.

Imaging-guidance and dynamic tracking techniques

This SBRT protocol requires stereotactic treatment with the use of a 3-D coordinate system defined by implanted electromagnetic transponders or implanted fiducial markers. The initial localization and alignment is based on the center of mass of the transponders/fiducial markers. Prior to treatment, kV CBCT is also acquired for soft-tissue alignment verification and deformation review. Significant rotations or soft tissue deformations will be corrected at initial localization stage, and intra-fractional rotations are ignored. After initial localization is performed, all effort is made to initiate the treatment delivery as quickly as possible. Continuous tracking and adjustment of target position during treatment is achieved by translational shifting the center of mass determined via electromagnetic transponders or fiducial markers, using remote couch movement. Specifically, per-beam couch corrections is made, if necessary, based on the actual target motion provided by dynamic tracking with either Calypso or per-beam kV orthogonal imaging with OBI. A correction action is performed if the target migrated more than 3 mm in any of three orthogonal coordinates.

To analyze the IGRT effectiveness, the motion correction magnitude and frequency of each treatment fraction are pooled and analyzed. The overall drifting magnitude during treatment is also captured and used to perform the margin evaluation based on the margin recipe described by Van Herk et al. [12].

Dosimetric benefit of dynamic tracking and online correction is evaluated by comparing delivered dose distributions with and without such correction. The dose distribution without correction is generated in the same treatment planning system by accumulating doses without online correction from each beam. The difference in dosimetry between actual treatment with online IGRT and simulated treatment without per-beam online corrections is used to quantify the benefit of using dynamic tracking and correction. The dosimetric analysis was based on original planning CT and contours.

Treatment plan Dosimetry

IGRT analysis and margin evaluation

For the total of 28 patients, 25 were setup based on with Calypso and 3 were setup based on with OBI. Among the 25 Calypso cases, 4 patients had at least 1 or more fractions switched to OBI imaging due to technical issues. 5 (out of 28) patients had transponder/seed migration that was captured by Calypso/OBI or CBCT imaging and their corresponding locations were adjusted and monitored by CBCT. Figure 1 shows an example of such case. 3 (out of 28) patients (1 fraction each) had significant motion or soft tissue deformation that led to the treatment halt for that specific fraction. One had to wait till full bladder and empty rectum; one had muscle spasm that caused sudden large body motion and was able to maintain position after break; and one had unexplainable large body motion during treatment and was able to maintain position after communications and further instructions. Overall, the combination of Calypso, OBI and CBCT IGRT options not only supplements each other with their own unique advantages, but also helps to guarantee uninterrupted treatment when one option is not available.

Online correction frequency per fraction is summarized in the histogram in Figure 2. Nearly 70% of fractions are treated without corrections; and ~90% fractions have less than 2 online corrections performed during treatment. On average 0.5 corrections have been performed per fraction.

The correction magnitudes in three directions (vertical, longitudinal and lateral) are shown in the first three box plots of the Figure 3, and the combined vector length of corrections are shown in the “Combined” boxplot. In general the corrections are small in all directions, with the lateral direction having the smallest correction range, and the longitudinal (superior-inferior) direction having the largest correction range. The combined correction magnitude from all three directions is (0.22 ± 0.11) cm.

The total couch shift from the start to the end of each treatment represents the target position drift. Histograms and statistical test results of target drift in three directions are shown in Figure 4. The magnitude is larger than corresponding values in Figure 3 because it shows the cumulative drift of the target during each fraction. Largest drifts are seen in couch vertical (AP/PA) direction, whereas lateral drifts are the smallest. Although occasional drifts > 0.3 cm can be seen, the Wincoxon Signed-Rank test reveals that drifts in none of the three directions are statistically significant.

In additional to the statistics above, the necessary margin for this specific patient group is calculated based on the target drift data described above and the margin recipe by Van Herk et al. [12]. The directional margins for AP/LR/SI directions are 2.5/1.0/1.7 mm; and the isotropic margin is 2.5 mm. This calculated margin size confirms that the protocol margin selection is sufficient.

Dynamic tracking benefits analysis

Figure 5 shows the difference in multiple dosimetric parameters between the simulated treatment without online corrections and the actual treatment that includes the dynamic tracking and correction. In general, when looking at the cumulative dose over 5 fractions, the benefit from per-beam correction is small for all dosimetric parameters corresponding to the protocol constraints. The per-beam correction has very small impacts on CTV coverage: key CTV dosemetric parameters without corrections are all within ± 0.5 Gy (1.4%) from those calculated with per-beam corrections. Differences from no correction are < 0.2 Gy for PTV D1cc/Dmean and < 0.07 for both RTOG [13, 14] and Paddick [15] conformity indices. Slightly higher difference is seen for PTV D99%: 0.6 to 1.5 Gy reductions are found for 5 patients without online correction. For the bladder, the majority of patients have < 1%/1.5 cc difference between with and without corrections, with only 2 patients having differences of −3%/-4.1 cc and +1.5%/+3.7 cc respectively. Similar distribution has been observed for the rectum: the majority has < 1%/< 0.7 cc difference, with exceptions of 4 patients having differences ranging from −1.7%/-1.1 cc to +4.6%/1.6 cc. Very small differences (< 0.9 Gy/< 0.4 cc) are also seen for the femoral heads dosimetric parameters. In general, only 1–2 patients (4-7%) would see >1%/1 cc increase in OARs irradiated volumes if no online correction was made. The rigorous supportive measures on bladder and rectum management that patients follow on each fraction and precise initial soft-tissue based target alignment have helped to minimize the motion during the radiation treatment. The variation could be larger and more frequent without these steps. The small diffrence between treating with and without online IGRT can be attributed to two other important components in the protocol. One is the carefully constrcucted immobolization device that helps to minimize the inter-fraction and the intra-fraction external patient body motion and rotation. Therefore, the online IGRT can be focused on tracking and correcting internal organ motion. With patient education and instruction, the internal motion as obseverd, can be kept to a small range and overall target drfiting is small too.

On the other hand, the dynamic tracking helps to catch the significant motion in those 3 patients/fractions that requires beam off/treatment halt. Currently, there is no clear indication on complete prevention of such sudden motion, thus the dynamic tracking is still beneficial to ensure the patient is treated with high precision through each beam irradiation.

In this study all dosimetric analysis was performed on the planning CT and original contours. In reality patient’s anatomy is dynamic and could change from day to day; therefore more accurate dosimetric analysis could be achieved if daily anatomical information is taken into consideration through online imaging modalities such as CBCT. However at current stage, due to the suboptimal image quality of CBCT images and lacking of a reliable deformable image registration tool, accurate cumulative dose analysis of the bladder and the rectum using daily anatomy is still difficult to achieve. We are actively working towards this goal though in-house algorithm development and close collaborations with vendors.