Comparison of two inverse planning algorithms for cervical cancer brachytherapy

Abstract Purpose To compare two inverse planning algorithms, the hybrid inverse planning optimization (HIPO) algorithm and the inverse planning simulated annealing (IPSA) algorithm, for cervical cancer brachytherapy and provide suggestions for their usage. Material and methods This study consisted of 24 cervical cancer patients treated with CT image‐based high‐dose‐rate brachytherapy using various combinations of tandem/ovoid applicator and interstitial needles. For fixed catheter configurations, plans were retrospectively optimized with two methods: IPSA and HIPO. The dosimetric parameters with respect to target coverage, localization of high dose volume (LHDV), conformal index (COIN), and sparing of organs at risk (OARs) were evaluated. A plan assessment method which combines a graphical analysis and a scoring index was used to compare the quality of two plans for each case. The characteristics of dwell time distributions of the two plans were also analyzed in detail. Results Both IPSA and HIPO can produce clinically acceptable treatment plans. The rectum D2cc was slightly lower for HIPO as compared to IPSA (P = 0.002). All other dosimetric parameters for targets and OARs were not significantly different between the two algorithms. The generated radar plots and scores intuitively presented the plan properties and enabled to reflect the clinical priorities for the treatment plans. Significant different characteristics were observed between the dwell time distributions generated by IPSA and HIPO. Conclusions Both algorithms could generate high‐quality treatment plans, but their performances were slightly different in terms of each specific patient. The clinical decision on the optimal plan for each patient can be made quickly and consistently with the help of the plan assessment method. Besides, the characteristics of dwell time distribution were suggested to be taken into account during plan selection. Compared to IPSA, the dwell time distributions generated by HIPO may be closer to clinical preference.

cervical cancer BT, the clinical application of inverse planning is still not widespread, due to the small number of catheters and the limitation of catheter placement. Recently, the IPSA optimizer was improved by adding a special parameter to restrict the dwell time variance between adjacent dwell positions in a catheter. 19 Its effects are still under investigation. As an alternative to IPSA, HIPO also has not been fully studied. 5,20 So far, only Ref. 21 has compared HIPO with un-improved IPSA. Thus, it is necessary to investigate the algorithms with constraint optimization and more clinical cases in order to make better use of them for cervical cancer BT especially with small number of catheters. Although several studies have proposed methods and tools for quantitative comparison of multiple plans, [22][23][24][25][26] they are mostly focused on EBRT and have never been adopted in BT. There is a lack of effective methods to compare BT plans quantitatively, comprehensively and consistently.
This study compared the dosimetric outcomes and characteristics of dwell time distributions for the plans generated using IPSA and HIPO for cervical cancer BT. A special plan assessment method was applied for quantitative comparison of quality among different treatment plans.

| MATERIAL AN D METHODS
Twenty-four patients treated between January 2017 and December 2019 were selected from our institution's clinical database for this retrospective study. According to FIGO stage classification, 27 the local tumor stage of the patients was as follows: IB2 = 2, IIB = 6, IIIB = 3, IIIC1r = 8, IIIC2r = 5. All the patients underwent 45 to 50 Gy whole pelvic EBRT followed by five fractions of intracavitary/ interstitial brachytherapy (IC/ISBT) with prescribed dose (PD) of 6 Gy. Nucletron standard tandem/ovoid (T/O) applicators and interstitial needles were used to deliver the IC/ISBT treatment. According to the different tumor shapes, the patients were treated with different combinations of applicators and needles as follows: seven patients with one tandem two ovoids, seven with one tandem three needles, seven with one tandem two ovoids two needles, two with one tandem two ovoids three needles, one with one tandem two ovoids four needles. After the insertion of applications, all patients underwent CT scans using the Brilliance CT Big Bore (Philips, Amsterdam, Netherlands) with 3-mm slice thickness. These scans were transferred to the Oncentra Brachy v4.6 (Elekta Brachytherapy, Veneedal, The Netherlands), where high-risk clinical target volume (HR CTV), intermediate-risk clinical target volume (IR CTV), bladder, rectum, sigmoid, and bowel were contoured in accordance with GEC ESTRO recommendation. 28,29 The HR CTVs covered a wide range, between 22.6 and 140.8 cc (mean 68.0 cc). IR CTV was a 3-mm volumetric expansion of HR CTV while subtracting all OARs. We treated it as a target but also as a help structure to control high dose regions outside HR CTV during dose optimization. The dose volume constraints in this study followed NCCN clinical practice guidelines v. 3.2019 (see Table 1). 30 The median of the constraint ranges and the EBRT dose of 45 Gy/25f were adopted for determining the dose volume constraints per fraction of IC/ISBT. Direct applicator reconstruction was carried out on the CT images using multi-planar reconstruction (MPR). All treatment plans were planned using the Oncentra Brachytherapy planning system v4.6, with a 192 Ir source for a Flexitron afterloader unit. The activation step was set to 2 mm.

2.A | IPSA planning
IPSA provides a combination of source activation, dose normalization, dose optimization, and dose prescription. Thus, the optimization can be performed just after contouring and applicator reconstruction. Table 2 shows the initial optimization settings used in this study. HR CTV was identified as the Reference Target. Note that its minimum surface/volume doses (700 cGy) were set to be higher than the PD just for optimization, aiming to increase the coverage of the targets while keeping the dose to OARs unchanged as far as possible. When the plan was optimized, the dose to 90% of HR CTV (D 90 ) would be normalized to 100% of the PD (600 cGy). The optimization parameters were adjusted and the calculation was repeated if the clinical objective was not achieved.
In Oncentra Brachy v4.3 and above, the IPSA optimization engine introduced a special parameter, dwell time deviation constraint (DTDC), which allows restriction of the difference in dwell times between adjacent dwell positions within each catheter. The But studies have shown that a high value of DTDC may work against target coverage and OARs sparing. 19,31 In this study, the DTDC was set to 0.1.

2.B | HIPO planning
HIPO was only used for the optimization of the dose distribution in this study, hence the source dwell positions were set the same as those for IPSA. The optimization parameters are listed in Table 3. IR CTV and HR CTV were identified as the PTV and GTV, respectively. Similar to DTDC, the dwell time gradient restriction (DTGR) is a modulation restriction parameter for HIPO to restrict large fluctuations between dwell times in neighboring dwell positions. It is also a relative value between 0.0 and 1.0, reflecting the "weight" of its importance in the optimization solution space. 31 The higher the value, the smaller the fluctuation. 32 However, to minimize adverse impact on target coverage and OARs sparing, the DTGR was set to 0.1 as well. Moreover, HIPO enables manual control of the sampling point settings for regions of interest (ROIs). For a high optimization precision, we increased the number of sampling points proportionally to the volumes of targets and OARs.

2.C | Plan evaluation
The dose volume parameters recommended by GEC ESTRO GYN were analyzed for all plans, including D 90 (dose to 90% of HR CTV and IR CTV), V CTV,200 (the volume of HR CTV receiving 200% of the PD), D 2cc (minimal dose received by the most irradiated 2 cc volume of bladder, rectum, sigmoid and bowel). The conformity index (COIN) was used to evaluate how well the PD covers the target volume and excludes nontarget volumes, which was calculated as follows: 33 where V CTV,ref is the volume of CTV that receives dose equal to or greater than PD; V ref is the volume receiving the PD. As the high dose region is a cause of concern, we defined a factor, local- were also analyzed.
For each patient, the plan assessment method described in Ref.
22 was adopted to quantitatively compare which of the two plans has a better performance. The method consists of two parts. The first is a graphical analysis providing a set of radar plots to show each quality score intuitively. The second is a total plan score weighting all quality scores to evaluate plan quality entirely. The quality score of each dosimetric parameter mentioned above can be calculated according to the following expression: The dose-volume constraints used for this study (Gy where C j is the constraint value of objective j given in Table 4, and P j is the corresponding plan value. For targets, a high P j represents a high coverage, homogeneity or conformal index, resulting in a low S j .
Similarly, for OARs, a low S j means a low dose to the OAR. Each quality score is represented by a point along the angle bisector of the corresponding objective in the radar plot. The distance between the point and the radar plot center corresponds to the score value.
By connecting all the points, a polygon representing the plan quality is generated. The smaller the polygon area, the higher the plan quality.
The total plan score was defined as follows: where w j is the weight of objective j, which reflects its importance in clinical treatment. As shown in Table 4, the weights used for this study were defined by a group of two professional treatment planners and three radiation oncologists based on clinical practice. The set of weights represented our clinical preferences. The two-sided paired t-test was used to make statistical comparisons of different quality indices between the IPSA and HIPO plans.

| RESULTS
Both IPSA and HIPO were able to produce dosimetrically acceptable treatment plans. Table 5   Both plans were created for the same patient treated by combining the T/O applicator with four interstitial needles. Figure 3 presents two radar plots corresponding to two cases of the study. The innermost octagon of the plots represents the constraints of all the objective, and out of the octagon implies exceeding the constraints, by which planners or radiation oncologists would be able to easily analyze the plan properties. When the polygon is closer to the plot center, the corresponding plan is superior. Thus, for the left plot, the IPSA plan is better than the HIPO plan and for the right plot, the opposite is true. In addition to the radar plot, the total weighted score can be used for plan comparison more directly. That is, a lower score corresponds to a better plan. Figure Figure 5b shows that the dwell times obtained using the HIPO algorithm formed a wave distribution, The scoring parameters used in the plan assessment method.

A U T H O R C O N T R I B U T I O N S
Conception and design of the study -Qi Fu, Yingjie Xu, Jing Zuo.

CONFLI CT OF INTEREST
No conflict of interests.