Effect of intrafraction adaptation on PTV margins for MRI guided online adaptive radiotherapy for rectal cancer

Data of 30 patients with intermediate risk or locally advanced rectal cancer treated on a 1.5 T MR-Linac (Unity, Elekta AB, Stockholm Sweden) between October 2018 and March 2021 were analyzed. Twenty-two patients received short course radiotherapy (SCRT; 5 × 5 Gy) and 8 received long course chemoradiotherapy (LCRT; 25 × 2 Gy). Ethics approval was obtained and all patients provided written informed consent for use of their data. Patients were treated using an online adaptive workflow (Fig. 1A) [23]. Approximately a week before start of the treatment, a planning CT and MRI were acquired on which the elective target volumes and organs at risk (OAR) were delineated for treatment planning according to delineation guidelines [24, 25]. For SCRT patients the CTV consisted of the mesorectum, the presacral region and pelvic lymph node regions, while for LCRT patients the lymph node region included the obturator region if pathological lymph nodes in situ were identified. The PTV was generated by adding an anterior anisotropic margin for the mesorectum CTV and a uniform 5 mm margin for the lymph node regions as suggested by Valentini et al. [24]. A bladder filling protocol consisting of drinking 250 ml water 30 min prior to simulation and the radiotherapy session on the MR-Linac was advised. On the MR-Linac, during each fraction first a 3D-T2-weighted MRI was acquired for adaptation (MRI_adapt) and the planning CT was deformably registered to MRI_adapt after which the elective target volumes were re-delineated for SCRT patients. Based on these new delineations, a new plan was optimized. For LCRT the plan corresponding to the actual mesorectum shape was selected through the use of a library of plans [26], followed by a virtual couch shift to correct for set-up errors[23]. Prior to starting the irradiation, an additional MRI was acquired to verify the position of the structures (MRI_ver). Subsequently, all patients were irradiated with an optimized 9-field intensity modulated radiotherapy (IMRT) plan with beam avoidance angles for the cryostat pipe (gantry angles 8°-18°) and two high attenuation regions of the MRL treatment couch (100°–140° and 220°–260°) [27]. After irradiation, a post treatment MRI (MRI_post) was acquired.

All acquisitions were performed with Field of View (FOV): 400 × 448 × 249 mm³, repetition time (TR): 1300 ms, echo time (TE): 128 ms, MRI_adapt had voxel size of 1.2 × 1.2 × 1.2 mm³ and acquisition time of 6 min, while MRI_ver and MRI_post used 1.2 × 1.2 × 2.4 mm³ acquired in 3 min.

All available images acquired during 5 daily treatment fractions of the patients treated in a short course scheme and the first fractions of every week of patients treated in a long course scheme were used for this study. The time intervals between MRI_adapt and MRI_ver, and MRI_ver and MRI_post were determined. These intervals correspond to the time needed for recontouring and plan adaptation, and irradiation respectively. On all the images, the gross tumor volume of the primary tumor (GTV_prim) and the mesorectum clinical target volume (CTV_meso) were delineated retrospectively using the contouring toolbox in Monaco v5.40.01 (Elekta, Stockholm, Sweden) by 2 experienced radiation technology therapists (RTT) following delineation guidelines [24, 25]. For each fraction, delineations of MRI_adapt were copied to the MRI_ver and MRI_post, and manually adjusted. All scans of one patient were delineated by the same RTT. Delineations were verified and, if needed, corrected by a radiation oncologist with over 10 years’ experience. We indicate the mesorectum CTV as delineated on the MRI_adapt with CTV_meso,adapt in the remainder of this paper. We use a similar convention for the other scans.

The peritoneal cavity (bowel area) as delineated on MRI_adapt of the first fraction was used and adjusted if needed. The CTV of the elective lymph node regions was not included in this study, considering the intrafraction motion of these regions is expected to be small [9]. For the same reason, a 2nd adaptation is suspected to have no substantial effect.

For every fraction, the delineated structures (GTV_prim,adapt and CTV_meso,adapt) were expanded in 3D to new structures: PTV_prim and PTV_meso in steps of 1.0 mm. The expansions were obtained using a rolling-ball algorithm [28], where the expansions were simulated on the actual scan. For GTV_prim,adapt the margin was isotropic and for CTV_meso,adapt we used anisotropic margins with the anterior expansion 1.6 times larger compared to all other directions. This choice is motivated by the study of Nijkamp et al. on mesorectum shape variation [29]. For every expansion the coverage was determined between the PTV and the associated structure on MRI_post. CTV_meso and GTV_prim were analyzed separately and independently. The coverage was defined as the number of overlapping voxels of the PTV and corresponding structures on MRI_post as a percentage of the total number of voxels of the structure on MRI_post. For CTV_meso, we considered the margin adequate when on average over the 5 fractions, PTV_meso covered 98% of CTV_meso,post in 90% of the patients. For the boost to the GTV_prim the criterion was relaxed to 95% volumetric coverage of GTV_prim,post by PTV_prim in 90% of the patients. The reason for using a lower coverage criterion is because the GTV_prim would be irradiated as a boost on top of the irradiation of the CTV_meso, resulting in less steep dose gradients. Considering the heuristic choice of these coverage criteria, we also assessed the effect of different volumetric coverage criteria on the PTV margins. All expansion and coverages were calculated using in-house software Match42.

To study the effect of a 2nd adaptation after MRI_ver, the intrafraction motion during redelineation and plan adaptation was evaluated. The verification envelope (VE) is defined as an expansion around GTV_prim,adapt or CTV_meso,adapt. We consider the threshold for intrafraction motion during adaptation to be exceeded, when the VE does not cover the GTV_prim,ver or CTV_meso,ver for at least 95% and 98% respectively.

For an isotropic VE, we determined whether the intrafraction motion during redelineation and plan adaptation exceeded the threshold for GTV_prim. For CTV_meso the VE was anisotropic with a similar anterior expansion a factor 1.6 times all other directions. If the threshold was not exceeded, no 2nd adaptation would be needed and thus GTV_prim,adapt or CTV_meso,adapt was used for evaluation. If the threshold was exceeded, a 2nd adaptation was applied and GTV_prim,ver and CTV_meso,ver were used to generate a new PTV_prim and PTV_meso to represent this adaptation. The online adaptive workflow following this approach is shown in Fig. 1B.

We varied the VE between 0 and 10 mm with 1 mm increments and we determined for each size the required PTV margin and the frequency of 2nd adaptations needed. This process is summarized in Fig. 2.

To determine the potential benefit of a 2nd adaptation for organs at risk, we determined the overlap between the bowel area and the required PTV_meso when no 2nd adaptations are performed and compared this to the overlap between the bowel area and PTV_meso after 2nd adaptations.

As a surrogate for the time required for re-delineation during the first and 2nd adaptation, we consider the change in volume of the CTV_meso. This is a practical surrogate, since re-delineation of CTV_meso is in particular necessary due to changes in rectal filling, which can directly influence the volume. For the first adaptation we consider the volume change between the reference CT and MRI_adapt. For the 2nd adaptation we consider the change between MRI_adapt and MRI_ver. A paired sample t-test was carried out (in SPSS v27.0) to test for a significant difference (α = 0.05).

The median time between MRI_adapt and MRI_ver was 12 min (inter-quartile range IQR = 10–23 min) and between MRI_ver and MRI_post 12 min (IQR = 11–15 min). The margin required for PTV_meso without 2nd adaptations was 6.4 mm in the anterior direction and 4.0 mm in all other directions. This is indicated in Fig. 3 for a VE of 10 mm, in which case no 2nd adaptations are needed. For PTV_prim a margin of 5.0 mm was required.

The coverage for all patients (n = 30) is shown in Fig. 4. In 90% of the population, the target criteria of 98% and 95% coverage were reached. The coverage of the remaining 10% was somewhat lower, but still above 88% in all cases.

The percentage of 2nd adaptations and resulting PTV margins are shown in Fig. 3. The minimal feasible margin for PTV_meso when performing 2nd adaptations was 3.2 mm in the anterior direction and 2.0 mm in all other directions using a VE of 1.0 mm. To achieve this margin, 2nd adaptations needed to be performed in 36% of the fractions.. Adapting more fractions did not lead to further margin reduction. For GTV_prim, 2nd adaptations would be needed in 17% of the fractions at a VE of 4.0 mm to reduce PTV_prim to 3.5 mm. Table 2 shows the number of patients needing a 2nd adaptation in 0, 1,2,3 of 4 fractions. For CTV_meso 7 patients needed a 2nd adaptation in 3 or more fractions. One of these patients was treated with LCRT. For GTV_prim, only 2 patients with tumors located in the mid- and upper rectum respectively, required a 2nd adaptation in 3 or more fractions.

When no 2nd adaptations were performed, the median (IQR) volume of the bowel that overlapped with PTV_meso was 35.3 (25.2–52.4) cm³ as shown in Table 3. For the reduced PTV_meso that was possible after 2nd adaptations, 19.3 (12.6–34.7) cm³ of the bowel overlapped with PTV_meso.

The CTV_meso showed larger changes between the planning CT and MRI_adapt than between MRI_adapt and MRI_ver, as reflected by a median volume difference (IQR) of 26.1 (14.6–45.7) vs 7.2 (4.8–12.9) cm³; p < 0.05.

For different coverage criteria, PTV margins for a workflow without 2nd adaptations (blue line) and when the minimum feasible margin is reached when performing 2nd adaptations (red line) are shown in Fig. 5. For both scenarios the margin increases gradually as more volumetric coverage is required. Above 97% a steeper increase of the margins is seen for PTV_meso as compared to the lower coverage criteria.