Major reduction of monitor units with the avoidance of leaf-sequencing step by direct aperture based IMRT optimisation

Abstract

Background and purpose. Compared with conventional 3D conformal radiotherapy (3D-CRT) the use of intensity-modulated radiation therapy (IMRT) has increased monitor units (MUs) in the delivery of prescribed dose to the patient and thus a potential risk of radiation-induced secondary cancer. Due to the elimination of the leaf-sequencing step in direct aperture based IMRT optimisation (DABO) the MUs in the beam delivery can be reduced. We compared MUs calculated by DABO with other IMRT techniques and 3D-CRT. Material and methods. Treatment plans for five head and neck cancer patients using dynamic IMRT technique (DMLC) and step-and-shoot (SMLC) technique (Varian Helios Cadplan), 3D-CRT (Varian Eclipse) and a home-made DABO were produced. The total number of MUs, dose coverage and standard deviation of prescribed dose in planning target volume (PTV) between different techniques were compared. Results. In all patients the PTV coverage and sparing of critical structures between the DABO, Helios DMLC and SMLC IMRT techniques was equivalent. Average MUs for beam delivery were 883 MU, 683 MU, 379 MU and 411 MU for DMLC, SMLC, DABO and 3D-CRT, respectively. Conclusions. The DABO IMRT technique is able to produce treatment plans equivalent in target coverage, dose uniformity and normal tissue sparing compared with the commercial IMRT techniques. The number of MUs with DABO were reduced with a factor of 2.3 (DMLC) and 1.8 (SMLC). The study suggests that with DABO IMRT technique the risk of secondary cancer can be reduced without compromise in the quality of treatment plans.

Intensity-modulated radiation therapy (IMRT) is becoming a routine technique in the treatment of several cancers due to improvements in dose conformity in the planning target volume (PTV) and normal tissue sparing. Unfortunately, the delivery of the IMRT treatment fields requires 2 to 5 times more monitor units (MUs) than 3D-conformal radiotherapy (3D-CRT) [1], [2]. An increase in the number of MUs is correlated with an increase in out-of-field radiation dose [3]. Hall and Wuu [4] estimated that if the number of MUs in IMRT is increased with a factor of 2–3 compared with 3D-CRT, IMRT is likely to almost double the incidence of second malignancies for patients surviving 10 years. The majority of second tumours are observed in the volume receiving <6 Gy [5] thus in the areas that are usually not in the treatment field. Therefore, minimizing dose to distant sites could be significant in reducing the risk of second tumours [6].

In most IMRT inverse planning systems the intensity distributions of the treatment fields are optimised. The intensity distributions are then reproduced with a multileaf collimator (MLC) using a separate leaf sequencing algorithm. Due to the separation of intensity optimisation and leaf sequencing, a large number of small subfields are required to realize the optimised beam intensities. Generally, as the number of subfields increases, the number of MUs and thus the out-of-field dose increases. A number of groups have investigated techniques to reduce the number of MUs needed for the MLC field segmentation [7], [8]. Also direct aperture based optimisation methods with IMRT have been studied in order to reduce the number of beam segments and number of MUs [9–11].

We present a direct aperture based optimisation (DABO) where the number of MUs required to achieve comparable dose distribution with commercial IMRT techniques is greatly reduced. The technique has also the advantage that the calculation of intensity distributions as an intermediate step is not required [12].

Material and methods

Direct aperture based optimisation (DABO)

The theoretical background of the present DABO technique has been already described [12–15]. Briefly, the algorithm for the control of MLC is based on the use of an explicit dose function and inverse calculation based on the MLC parameters. Since the calculation of intensity distributions as an intermediate step is not needed the physical limitations of the MLCs can be implemented. Furthermore, the head scattering, leakage and “tongue and groove” effect can be accurately modelled by the DABO method [12].

The DABO formalism is based on the use of pencil beam model for the dose calculation. The intensity distributions generated by the MLC are modified with the help of Heaviside function H. The dose model for one leaf pair can be modelled as

where x is a point in a patient space, D(x) the dose in the point x, w the width of the treatment field, δ t radiation beam-on time and h(x,u) the dose deposition kernel. H(u-a) and H(b-u) are the modified Heaviside functions for the left and right leaves [12]. Superposition of all leaf-pairs is used to calculate the total dose.

Prior to the direct aperture optimisation the treatment fields are set and dose limits for the PTV and organs at risk (OAR) are defined. Additionally, the user determines the minimum amount of MUs for each subfield and the maximum number of subfields for one field. Before the start of the optimisation, the predefined shapes of the apertures are generated using the beams-eye-view (BEV) of the PTV and the OARs for each beam angle. During the optimisation the feasible solution for the MLC leaf positions and the irradiation times of the subfields are calculated. The dose deposition kernel is computed for the subfields using triple Gaussian pencil beam and MLC scattering models. Nonlinear inequalities for leaf positions and irradiation times are solved by using the quasi-Newton method (L-BFGS-B) [12]. After the optimisation, a final dose calculation is performed using the optimised aperture shapes and aperture weights.

In the present investigation, a step-and-shoot MLC (SMLC) feature of the DABO optimisation was applied although the technique can be applied with the dynamic MLC (DMLC) treatment delivery [16]. Also the tongue and groove effect was not modelled by the optimization in this study. The DABO optimisation was implemented in the commercially available treatment planning system (Cadplan, Varian, Palo Alto, CA).

Treatment planning

Dose distributions of five successive patients with head and neck cancer were optimised with DABO and calculated. The plans were compared with treatment plans generated by conventional 3D-CRT, SMLC and DMLC techniques from Cadplan Helios (LMC-version 6.3.5). The IMRT plans were generated using Cadplan treatment planning system (version 6.47) with 4 MV photons and MLC with 40 leaf-pairs.

In all IMRT treatment plans (DABO, Helios DMLC and SMLC) equal number of beams arranged with the same beam angles were used. The number of treatment fields per patient ranged from 6 to 8. The 3D-CRT plans were generated by using the standard technique of the clinic.

For dose objectives, the minimum dose to the PTV was set to be greater than 98% and the maximum dose less than 105% of the prescribed dose. The OARs specified were spinal cord, parotid glands, mandible, mucous membrane and lungs (if the target volume extended to the lower neck). The maximum dose for the spinal cord was set to 48 Gy. For other OARs dose volume constraints were used. The additional constraints in the DABO optimisation were a minimum of 3 MUs for one subfield and a maximum number of 9 subfields for a treatment field. The plans were generated using a leaf position tolerance of 2 mm and the minimum leaf gap of 5 mm. With Helios SMLC the leaf motion calculation was performed with 9 intensity levels to ensure the adequate agreement between optimal and actual fluences [17], [18].

Treatment plan analysis

Dose-volume histograms (DVH) of the PTV and the OARs were calculated to compare the dosimetric quality of the 3D-CRT and IMRT plans. PTV coverage was estimated with D95 (the minimum dose covering 95% of the PTV). The dose uniformity was assessed by calculating the standard deviation (SD) of the doses within the PTV. The number of total MUs obtained for each method was compared. Additionally, the number of subfields created by SMLC and DABO was examined.

Results

Dose distributions with different IMRT techniques were almost identical which can be seen in Figure 1. The sparing of critical structures was also comparable as seen from DVHs in Figure 2. For simplicity, the DVHs of SMLC are left out from the graph. The average maximum and the average mean PTVs doses for DMLC, SMLC, DABO and 3D-CRT were 108.0%, 111.4%, 109.3%, 111.1% and 99.2%, 99.7%, 99.0% and 99.3% respectively. All the planned doses were normalized so that the median dose was 100%, corresponding to 56 Gy.

Figure 1.  Comparison of isodose distributions for a head and neck patient using DMLC, SMLC, DABO and 3D-CRT. The median doses of the PTVs were set to 56 Gy (normalized to 100%).

Figure 2.  DVH for PTV, spinal cord and mandibula between DMLC (

), DABO (▵) and 3D-CRT (♦). For simplicity, the DVH of SMLC is not presented.

The total MUs of all the treatment plans are presented in Table I. On average, the MUs created by DABO were reduced by a factor of 2.3 and 1.8 compared to Cadplan Helios DMLC and SMLC respectively. The number of MUs was smaller with DABO than with 3D-CRT by a factor of 1.1.

Table I.  Comparison of the number of MUs needed to deliver a dose of 2 Gy to PTV and the corresponding SDs and D95s.

	MUs				PTV COVERAGE
Patient					SD				D95 (Gy)
#	DMLC	SMLC	DABO	3D-RT	DMLC	SMLC	DABO	3D-RT	DMLC	SMLC	DABO	3D-RT
1	934	680	278	462	3.8	4.2	4.0	5.6	52.4	52.0	52.4	50.4
2	638	489	287	450	3.2	3.6	4.1	7.7	53.9	53.4	53.8	49.3
3	996	717	561	444	4.4	5.6	5.4	7.0	51.0	50.1	51.2	48.8
4	831	699	253	236	2.0	2.9	3.2	6.8	54.4	53.6	53.5	50.5
5	1018	830	516	462	6.0	6.2	7.4	7.8	50.0	50.1	48.8	47.2
AVRG:	883	683	379	411	3.9	4.5	4.8	7.0	52.3	51.8	51.9	49.2

The evaluated dose uniformity (SD) and PTV coverage (D95) are illustrated in Table I. The SD was smallest for DMLC as expected. The D95 values were comparable with the different IMRT techniques and largest for the 3D-CRT. The differences between the IMRT and 3D-CRT treatment plans were clear for the advantage of IMRT plans. The number of subfields generated by DABO was on the average 53% smaller than with Helios SMLC (Table II).

Table II.  Comparison between the number of subfields between SMLC and DABO.

	SUBFIELDS
Patient #	SMLC	DABO
1	93	48
2	44	22
3	49	32
4	59	25
5	88	32
AVRG:	67	32

Discussion

Although the risk of secondary malignancy after head and neck radiotherapy is small, the risk is increasing with longer survival times of the cancer patients, especially in children treated with radiotherapy. In the investigated commercial IMRT techniques the number of MUs was more than two-fold compared with conventional 3D-CRT and DABO (Table I). The increase in MUs leads to higher head leakage and scatter [3]. This is expected to increase radiation induced malignancies [19]. The present investigation supports the view that with the DABO technique the MUs and related risk of secondary malignancy stay at the level typical with 3D-CRT.

The results indicate that all IMRT techniques yielded a more uniform coverage of PTV than a 3D-CRT technique. With respect to PTV coverage the DMLC technique performed better than the step-and-shoot delivery methods (SMLC and DABO) (Table I). The main drawback of the DMLC technique is the highest MUs among the tested techniques. The reduction of MU (57%) by DABO was similar with other direct aperture based optimisations. Reductions of MUs by 80% [11], 33% [20] and 42% [10] are reported when compared to Corvus and Eclipse treatment planning systems or pencil-beam based optimisation, respectively.

The average number of subfields with DABO technique was 32, which is markedly less than 67 with Cadplan Helios SMLC. This is due to the separation of the optimisation into the optimisation of the beam intensities and leaf sequencing step. With other direct aperture based optimisation methods the number of subfields has been 34 [10] and 21 [11] when investigated with a head and neck patient examples.

We conclude that the target coverage and the normal tissue sparing are comparable for DABO and conventional IMRT plans. However, the number of MUs are 45 and 57% lower for the DABO method when compared to SMLC and DMLC, respectively. A linear risk estimate would mean that a risk of radiation-induced second cancer is decreased with a factor of about 2 with DABO compared to the conventional IMRT techniques. Using the aperture based IMRT optimisation the risks of radiation induced second tumours can be kept at the level typical with 3D-CRT without compromise in the quality of treatment plans.

Acknowledgements

This research was supported by the Finnish Funding Agency for Technology and Innovation and Varian Medical Systems Finland Ltd.