Evaluation of offline adaptive planning techniques in image‐guided brachytherapy of cervical cancer

Abstract Modern three‐dimensional image‐guided intracavitary high dose rate (HDR) brachytherapy is often used in combination with external beam radiotherapy (EBRT) to manage cervical cancer. Intrafraction motion of critical organs relative to the HDR applicator in the time between the planning CT and treatment delivery can cause marked deviations between the planned and delivered doses. This study examines offline adaptive planning techniques that may reduce intrafraction uncertainties by shortening the time between the planning CT and treatment delivery. Eight patients who received EBRT followed by HDR boosts were retrospectively reviewed. A CT scan was obtained for each insertion. Four strategies were simulated: (A) plans based on the current treatment day CT; (B) plans based on the first fraction CT; (C) plans based on the CT from the immediately preceding fraction; (D) plans based on the closest anatomically matched previous CT, using all prior plans as a library. Strategies B, C, and D allow plans to be created prior to the treatment day insertion, and then rapidly compared with the new CT. Equivalent doses in 2 Gy for combined EBRT and HDR were compared with online adaptive plans (strategy A) at D 90 and D 98 for the high‐risk CTV (HR‐CTV), and D 2 cc for the bladder, rectum, sigmoid, and bowel. Compared to strategy A, D 90 deviations for the HR‐CTV were −0.5 ± 2.8 Gy, −0.9 ± 1.0 Gy, and −0.7 ± 1.0 Gy for Strategies B, C, and D, respectively. D 2 cc changes for rectum were 2.7 ± 5.6 Gy, 0.6 ± 1.7 Gy, and 1.1 ± 2.4 Gy for Strategies B, C, and D. With the exception of one patient using strategy B, no notable variations for bladder, sigmoid, and bowel were found. Offline adaptive planning techniques can shorten time between CT and treatment delivery from hours to minutes, with minimal loss of dosimetric accuracy, greatly reducing the chance of intrafraction motion.


| INTRODUCTION
Modern three-dimensional image-guided intracavitary high-dose rate (HDR) brachytherapy is increasingly used in combination with external beam radiotherapy (EBRT) and/or chemotherapy to manage cervical cancer worldwide, with significant improvement of local disease control and survival reported. [1][2][3][4] The entire radiation treatment is typically delivered in 45 Gy for 25 fractions with EBRT followed by HDR in 4-6 fractions using the tandem and ring (T&R) or tandem and ovoid (T&O) applicators. Magnetic resonance imaging [5][6][7][8] or computerized tomography (CT) [9][10][11] are currently used in HDR treatment planning to define the applicator position and delineate the target and organs at risk (OARs).
There are several uncertainties in the course of HDR treatment which could result in deviations between the actual delivered and the planned doses, including source calibration, dose calculation accuracy, target and OARs delineation, inter-fraction, and intra-fraction motions, etc. Source calibration, dose calculation, and contour delineation uncertainties have been extensively studied in the literature, and are beyond the scope of this study. 12 During the course of the HDR treatment, since the target, OARs and HDR applicator are not a rigid system, their relative positions may change not only from treatment to treatment, but also between the image acquisition and treatment delivery. A recent failure modes and effects analysis (FMEA) study identified the potential for applicator movement as one of the most high-ranking failure modes in the HDR treatment. 13 Due to the steep dose gradient in HDR treatment planning, small changes in the relative position between regions of interest and the applicator could lead to marked differences between the actual delivered and the planned doses. In this work we will focus on the inter-and intrafraction motion uncertainties during the T&R HDR treatment of cervical cancer.
Currently popular clinic practices for HDR treatment include using a single plan to treat the patient throughout the course, or creating an online adaptive plan for each fraction. In the single plan strategy, the plan from the first fraction can be propagated to the remaining treatments under the assumption that interfraction motion can be ignored. The intrafraction motion between the applicator insertion and treatment delivery can be minimized from the second fraction onward for the single plan strategy. However, previous studies have shown that the interfraction motion of critical structures relative to the applicator may cause marked dose deviations between the planned and delivered dose. 14 Under the assumption that variations due to interfraction motion are much greater than those due to the intrafraction motion, online adaptive replanning on a per fraction basis have been implemented for HDR treatment. 15,16 Online adaptive planning techniques can eliminate the interfraction motion since a new CT image will be acquired for the treatment planning each day. However, the time between the image acquisition and treatment delivery can be several hours. Significant anatomic changes may occur during that time period, and could increase the uncertainty in dosimetry. Dosimetric comparisons between single planning and adaptive daily planning strategies have been investigated, and improved dose sparing for OARs has been found for the adaptive daily planning technique. 17 However, the dosimetric impact of week-to-week interfraction motion versus a few hours of intrafraction motion is still under debate. Recently, the results of a large, multi-institution study suggest that effects of inter-and intrafraction motion may not be as different as we once believed. 18 The purpose of this study was to evaluate adaptive offline replanning techniques which may potentially reduce the operating room to treatment completion time for the HDR treatments, thus minimizing both inter-and intrafraction motion.

| MATERIALS AND METHODS
where n is the number of fractions, d is the dose per fraction, and α/ β = 10 Gy for the HR-CTV and 3 Gy for OARs. Our in-house guideline for the combined EBRT and HDR treatments is to maintain the dose D 2 cc (the minimum doses to the highest irradiated 2 cc volume) <90 Gy for the bladder, and D 2 cc <75 Gy for all other OARs (rectum, sigmoid, and small bowel), 19 while keeping the HR-CTV coverage D 90 (dose to 90% of target volume) >80 Gy.

| 317
The results from both offline strategies together with single plan strategy were compared with the clinical online daily adaptive replanning strategy (see Fig. 2 for an illustration). 3 | RESULTS Figure 3 shows the contour variation of the target and OARs for two adjacent treatments of a representative patient after the rigid registration based on the T&R applicator. Since the registration is based on the T&R applicator, only small interfraction motion was observed for the HR-CTV. However, significant day-to-day variations of the OARs were found due to their filling changes.

2.A | Strategy A
Only one patient (Patient #7) marginally failed to meet the small bowel constraint due to the anatomy of that patient. Table 1    Even though very dramatic significant shape changes for all critical structures occurred during the treatment course (Fig. 3 showed that both CT-based and MRI-based scans at in cervical cancer brachytherapy are adequate for OAR DVH analysis, MRI remains to be the standard for HR-CTV definition due to its superior soft tissue contrast, and CT image can significantly overestimate the tumor volume. 20 We want to point out in our current clinic workflow, CT images are used to define both the HR-CTV and OARs. The evaluation of HDR treatment plan is largely driven by the OAR doses rather than the HR-CTV coverage.

| CONCLUSIONS
Offline adaptive planning techniques allow plans to be created prior to the treatment day insertion, and then rapidly compared with the new CT. Our study shows offline adaptive techniques offer similar plan quality as online adaptive strategy, while dramatically shortening the time between the CT acquisition and corresponding treatment delivery from hours to minutes, therefore improving patient experience, staff convenience, and reducing dosimetric uncertainty due to intrafraction motion.

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