Delivery of online adaptive magnetic resonance guided radiotherapy based on isodose boundaries.

Magnetic Resonance-guided Radiotherapy (MRgRT) allows direct monitoring of treated volumes. The aim of this study was to investigate the feasibility of a new gating strategy consisting in using an isodose as boundary. Forty-four patients treated for thoracic and abdominal lesions using MRgRT were enrolled. The accuracy of the new strategy was compared to the conventional one in terms of area improvement available for gating without compromising target coverage. A mean increase of 24% for lung, 15% for liver and 11% for pancreas was observed, demonstrating how the new method can be useful in challenging situations with low dose conformality.

Introduction

Recent technological developments in the field of radiation oncology led to the introduction of new hybrid systems, able to combine the delivery precision of a modern linear accelerator with the image quality of Magnetic Resonance (MR) imaging [1], [2], [3]. Besides enhanced positioning imaging quality, these systems allow to reduce inter-fraction treatment variability owing to a dedicated online adaptive workflow and to efficaciously manage intrafraction organ motion, required for real-time MR imaging and advanced gating strategies [4], [5].

Several recent studies quantified tumor motion during treatment and demonstrated the occurrence of effects which may happen during a treatment fraction, such as modified target trajectories with movement of the tumor outside the high dose region [6], [7]. For this reason, both passive and active strategies have been developed to safely manage tumor motion during therapy, reporting different levels of reliability and robustness [8], [9], [10], [11]. With the introduction of Magnetic Resonance-guided Radiotherapy (MRgRT), direct monitoring of tumor motion during treatment delivery has become possible: hybrid MR-Linacs allow the continuous acquisition of planar MR images during treatment with up to 8 frames/second on a single sagittal plane or with 5 frames/second on three orthogonal plans [5], [12].

As MR images are acquired in 2D-cine mode, in clinical practice often a tolerance region (boundary) is defined to stop the radiation beam as soon as the target structure moves beyond this region. This gating strategy can either be managed automatically or manually by a radiation therapist. A visual scheme of the gating functionality is reported in Supplementary materials. Automatic gating commonly uses Clinical Target Volume (CTV) as target structure and its isotropic geometric expansion, which frequently coincides with the Planning Target Volume (PTV), as boundary region [13], [14], [15], [16].

The possibility of indirect CTV monitoring using nearby anatomical reference structures (i.e. organs at risk, OARs) has also been explored with interesting results in selected clinical scenarios [17], [18]. Despite its wide diffusion, the strategy of using the CTV as target structure and its isotropic expansion as boundary can lead to suboptimal results, as the treated volume inside the prescribed isodose may not match with an isotropic structure. For example in pancreatic lesions, the anatomical proximity of several radiosensitive OARs to the tumor (e.g., duodenum, stomach and bowel loops) leads to highly irregular dose distributions cause of the anatomy and location of the target [19], [20]. In these cases, the use of an isotropic boundary can be misleading allowing beam delivery also when the tumor is located outside the high dose region; this can result in decreased target coverage and increased risk of toxicity.

To the best of our knowledge no alternative approaches to prevent such effect during MRgRT delivery are available. The aim of this study was to investigate an alternative strategy to manage beam delivery in cases where the high dose region significantly differs from the geometric expansion of the CTV. Therefore, the use of an isodose line as alternative boundary was compared to conventional isotropic boundaries for different stereotactic body radiotherapy (SBRT) treatments of thoracic and abdominal lesions.

Material and methods

Forty-four patients treated for thoracic and abdominal lesions were retrospectively enrolled in this study (17 lung, 17 liver and 10 pancreatic lesions). All the patients underwent a MRgRT treatment delivered in 3 to 8 fractions on 0.3 T MRgRT system (MRIdian, ViewRay, Mountain View, California, USA) from June 2019 to February 2020 and signed a specific informed consent for therapy delivery and use of therapy imaging for our study.

All MRgRT treatments were administered in breath hold conditions by monitoring the tumor position during the entire therapy by means of 2D-MR images acquired in cine modality with 4 frames/second. No GTV to CTV margin was foreseen and PTV was defined by adding a 3 mm isotropic margin to the CTV. Dose calculation was performed with a Monte Carlo algorithm with a grid resolution of 0.2 cm and the beam energy was 6 MeV [21].

MR images were acquired on the sagittal plane passing through the geometrical center of the GTV using a true fast imaging (TRUFI) sequence, with 5 mm slice thickness, 35x35 cm2 field of view, T2*/T1 image contrast and spatial resolution of 3.5x3.5 mm2 [22], [23].

Owing to automatic gating, the contours of the target structure were automatically deformed and projected onto the different cine frames as well as the boundary structure, which was rigidly transferred to the cine MR frames. The radiation beam was automatically turned off whenever a predefined percentage of the target structure (set equal to 5% in this study) moved out of the boundary. A visual scheme of the gating functionality is reported in Supplementary materials.

Treated volume was defined for each treatment plan as the volume enclosed by the Reference Isodose (RI) and represented the dose value taken into consideration to evaluate the target coverage of the treatment [24]. The RI was set equal to the 95% of the prescribed dose for the plans normalized at 50% of the target structure, while it was coincident with the prescribed dose for the plans normalized to an isodose volume, as suggested by ICRU 91[25]. Treatment plans were classified as function of their conformality, defined using the conformation number (CN) [26], [27].

Here, the PTV represents the Target Volume (TV), VRI represents the volume of reference isodose and TVRI the TV covered by the reference isodose, corresponding to the intersection between PTV and VRI.

Based on the CN value, treatment plans were classified in plans with low (CN ≤ 0.6), moderate (0.6 < CN ≤ 0.7) or high (CN > 0.7) conformality. Boundary analysis was performed comparing the area covered by the PTV in the gating plane with the area covered by the RI on the same plane.

The degree of overlap between PTV and RI on the gating plane was quantified in terms of Hausdorff distance (HD) and Dice index, calculated between the reference isodose and the PTV area using MIM software (version 7.0.4, MIM software Inc, United States) [28], [29]. The difference between the two boundaries was considered significant for Dice values lower than 0.9. In the evaluation of the importance of using a dosimetric boundary at different plan complexity levels, the correlation between Dice, HD and CN was quantified using Spearman’s correlation test [30]. All the statistical analyses were performed using R software (version 3.6.1, R Core Team, Austria).

Results

The observed dose distribution conformality was low in two of the selected plans, moderate in 44 and high in 18 cases. As regards the dose conformality, the lowest median CN values was observed in case of pancreatic (0.7) and lung lesions (0.7), while higher values were obtained of liver (0.7).

The median values and the corresponding ranges for the different parameters chosen to compare the PTV and reference isodose volume (RIV) boundary were reported in Table 1, separately for each anatomical site, while detailed values were reported as Supplementary material.

Table 1

Median values of the indicators chosen to compare RI and PTV for single anatomical site. Dose values corresponding to the RI, PTV and CN of the treatment plans were reported as SBRT plan indicators.

		Dosimetric Indicators			Geometric Indicators
		RI	PTV	CN	DICE	Δ Area (RIV-PTV)		HD
		Gy	cc			cm2	%	mm
Lung	Median	47.5	2.8	0.7	0.8	0.6	24	2.8
	Range	39.9 / 50	1 / 38	0.5 / 0.8	0.7 / 0.8	0.0 / 1.9	1 / 48	1.5 / 12.1
Liver	Median	47.5	14.0	0.7	0.9	0.9	15	3.0
	Range	22.8 / 51.3	4.3 / 75.3	0.6 / 0.9	0.8 / 0.9	0.3 / 2.9	2 / 30	2.8 / 6.0
Pancreas	Median	40.0	37.6	0.7	0.8	1.2	11	6.8
	Range	30 / 44	7.3 / 74.7	0.5 / 0.8	0.8 / 0.9	−1.9 / 3.6	−2 / 35	3.1 / 11.9

The area covered by RI was generally larger than the PTV area, with a mean increase of 24% for lung, 15% for liver and 11% for pancreatic lesions: only in three pancreatic cases an RI area smaller than PTV was observed, mainly due to the presence of radiosensitive OAR in close proximity to the target.

Concerning HD analysis, higher values were observed in pancreatic lesions (median value equal to 6.8 mm) with respect to liver (3.0 mm) and lung (2.8 mm), demonstrating a larger difference in using the two strategies in case of pancreas, as also supported by the lowest value in terms of percentage degree of overlapping (11% in pancreas, 15% in liver and 25% in lung).

Overall, on the basis of the geometric analysis performed, a significant difference in using the RI with respect the PTV as boundary was observed in 22 of the investigated cases, with a larger impact observed in case of pancreatic lesions (8/10) with respect lung (12/17) and liver (2/17).

Spearman’s correlation analysis showed a high correlation (R = 0.8) between the Dice index and the CN of the treatment plan, independently from the treatment site investigated, while no significant correlation was observed between HD and CN (R = -0.3).

Discussion

In this study, the use of a treated isodose boundaries as gating structure for MRgRT was investigated in comparison to conventional standard strategy consisting in using an isotropic expansion of the target structure.

The analysis highlighted limited differences between the two strategies for treatment plans characterized by dose distributions of high conformality (only 4/18 plans with CN ≥ 0.7 reported a Dice index < 0.9), while larger differences were observed in plans with low conformality (23/26 plans with a CN < 0.7 reported a Dice index < 0.9).

Gating efficiency is a relevant and contemporary topic in MRgRT, as time represents a key-factor to clinically manage as many patients as possible with these hybrid machines, as demonstrated in literature describing experiences in both low and high field units [31], [32].The accuracy of treatment delivery, using a geometrical boundary, was firstly studied by van Sörnsen de Koste et al observing that this strategy ensured high gating efficiency in 15 patients affected by abdominal and thoracic tumors, with the GTV inside the selected boundary for 67% to 87% of the total treatment time [33].To the best of our knowledge, no alternative gating strategy to geometric one gating has been reported in literature.

Using treated volume as boundary may further improve gating efficiency by stopping the treatment delivery only when the GTV moves outside the region receiving the prescribed dose, regardless of its geometric position with respect to PTV.

Dice analysis demonstrated relevant differences between the two strategies in the lung and pancreas, while this was not observed in the liver. Fig. 1 summarizes some real-life clinical scenarios in which the use of this strategy provides significant improvements in gating efficiency: the use of this strategy increases the area where treatment delivery is allowed, ensuring equivalent target coverage and increasing gating efficiency in lung lesions (Fig. 1a). Similarly, the application of the dosimetric boundary can avoid that target results are outside the high dose region, but the system does not stop the beam delivery, as in the described case of pancreatic lesions (Fig. 1b).

Fig. 1

Schematic impact using a reference isodose as boundary in the treatment of lung and pancreatic lesions. Top: sagittal plan views. Bottom: comparison of the area covered by reference isodose (red) and PTV (blue). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Despite the promising results, this exploratory study has different limitations, which should be considered prior to transferring this procedure to clinical practice. Firstly, MR gating implemented in low field hybrid units is performed in a single sagittal plane; all considerations reported do not account for the left–right direction, for which the use of larger CTV-PTV margins remains to date the only effective strategy for motion compensation, as evidenced in various low field MRgRT experiences [8], [34], [35].Secondly, this is a retrospective analysis using plans of patients treated with the geometric boundary strategy, which limits the clinical impact of these observations. Further prospective studies including larger cohorts of patients are recommended to quantify the gain in terms of gating efficiency and treatment time reduction, which could be achieved introducing the dosimetric boundary strategy in daily clinical practice.

In conclusion, the use of the dosimetric boundary could allow to identify a boundary region with the advantage of directly reflecting the actual dose distribution, ensuring a gating region which could allow reduced treatment time in case of lung and liver lesions and to optimize the beam delivery in case of pancreatic ones.

Declaration of Competing Interest

The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: Nicola Dinapoli: has received speaker honoraria by ViewRay company. Luca Boldrini: has received speaker honoraria by ViewRay company. Davide Cusumano: has received speaker honoraria by ViewRay company. Vincenzo Valentini: has received speaker honoraria by ViewRay company.