Original ArticleThe Clinical Implications of the Collapsed Cone Planning Algorithm
Introduction
It is recommended that the homogeneity of the dose distribution within a given target volume should be kept within +7% and −5% of the prescribed dose owing to the steep slopes of the dose-effect relationships for tumour control [1]. One step in achieving this degree of homogeneity is the accuracy of treatment planning systems (TPS) in calculating the delivered dose to the patient, both at the dose-specification point and in the surrounding tissue. It has been suggested that ideal levels for accuracy of dose planning algorithms are 2% in regions of low-dose gradient, and 2 mm in high-dose gradient regions [2].
In the two centres involved in the study, treatment plans are routinely calculated using the pencil beam (PB) algorithm on a Helax-TMS (version 6.1) radiotherapy TPS (Nucletron, Veenendaal, Netherlands) [3]. Despite meeting the above tolerances for treatment plans in regions of homogenous tissue density, previous studies have shown the limitations of the PB algorithm where heterogeneities exist 4, 5, 6, 7, 8, 9. These limitations are due to the one-dimensional density correction of the PB algorithm, which does not accurately model the distribution of secondary electrons in regions of tissue heterogeneity. Doses are scaled according to the radiological depth along a ray-line from the radiation source to a calculation point, not accounting for the effects of side and backscattered radiation.
The collapsed cone (CC) algorithm is now available for clinical use on Helax-TMS 10, 11, 12. This more complex dose model was first described in 1987. However, it is so computationally intensive that commercial TPS have only recently become powerful enough to allow consideration of its use in the radiotherapy clinic. In contrast to the one-dimensional correction of the PB algorithm, the CC uses full three-dimensional density scaling to model the effects of nearby heterogeneities on dose-calculation points. This results in increased accuracy, particularly within regions of low-density tissue and steep density gradients (such as interfaces between lung and soft tissues).
A number of previous studies have verified the accuracy of the ADAC Pinnacle (ADAC Laboratories, Milpitas, U.S.A.) CC algorithm in homogeneous and heterogeneous phantoms 13, 14, 15. Francescon et al. [16]used patient computed tomography (CT) data to compare the Pinnacle CC algorithm with a Monte Carlo-based dose model, which is currently considered to be the gold standard of dose calculations. It was shown that, for breast and mediastinum treatments, the results of the two calculation methods are comparable, with no systematic differences.
The increased accuracy of dose calculations using the CC instead of the PB algorithm on the Helax-TMS TPS has also been shown 11, 17. Morgan et al. [17]used tissue substitutes to compare measured and calculated doses using both algorithms in several situations. The CC algorithm modelled the dose reduction in lung tissue and the secondary build-up of dose downstream of lung tissue at interfaces between lung and soft tissue more accurately than the PB algorithm. Further measurements showed that neither algorithm correctly models the dose at interfaces between bone and soft tissue, but it was indicated that the CC algorithm is more accurate within bone.
We are unaware of any studies that have examined the clinical implications of the differences in theaccuracy of the available algorithms. Overestimation of the dose to the planning target volume (PTV) in regions of low-density heterogeneities by the PB model could lead to under-dosage of the PTV. This may have implications on tumour control. We have therefore carried out a retrospective study in which clinical PB plans were recalculated using the CC algorithm. The differences in the calculated dose distributions were evaluated in terms of dose coverage of the PTV and doses to organs at risk (OARs). We have also assessed the differences in absolute isocentric dose, and discussed the effect of the results in future treatment planning.
Section snippets
Method
A series of 20 consecutive patients with CT plans were selected (10 Ca oesophagus and 10 Ca lung). The oesophagus treatments were all two-phase plans, giving a total of 30 plans. These treatment sites were chosen, as the PTV includes areas of tissue heterogeneity and might be expected to show the difference between the PB and CC algorithms to maximum effect. Consecutive patients were chosen to avoid any selection bias. In most cases, a clinician defined the gross tumour volume (GTV), and the
Absolute Dose to the Isocentre
The monitor units needed to deliver the prescribed dose to the dose-specification point differed by up to 3.4% depending on the algorithm used. In all but four cases, more monitor units were needed to deliver the prescribed dose when using CC. This is due to the overestimation of the dose by the PB in heterogeneous regions, as shown in previous phantom studies [17]. Having shown that the CC algorithm is more accurate, it is clear that, in these cases, the delivered dose is too low when using
Discussion
The results indicate that, in many cases where heterogeneities are present, the actual coverage of the PTV is less uniform than the PB plan indicates. This results in considerable under-dosage of the PTV in most treatments within the thorax. There seem to be three main causes of this under-dosing: (1) inclusion of low-density normal lung tissue in the PTV, which receives less dose than the higher density GTV; (2) decrease in scattered dose from surrounding low density tissue; and (3) secondary
Conclusion
The clinical impact of the new generation of dose calculation models has yet to be fully assessed. From this small study, it has been shown that current international guidelines for PTV dose homogeneity cannot routinely be satisfied when using the CC model in regions such as the thorax. This is mainly due to the presence of low-density tissue in the PTV and the secondary build-up of dose. However, the suitability of these guidelines in such regions may be questioned. It is unclear how to deal
References (22)
- et al.
Accuracy of two- and three-dimensional photon dose calculation for tangential irradiation of the breast
Radiother Oncol
(1997) - et al.
Limitations of a pencil beam approach to photon dose calculations in the head and neck region
Radiother Oncol
(1995) - et al.
Impact of simple tissue inhomogeneity correction algorithms on conformal radiotherapy of lung tumours
Radiother Oncol
(2001) - et al.
Effect of low-density lateral interfaces on soft-tissue doses
Int J Radiat Oncol Biol Phys
(1997) - et al.
Pitfalls in the use of high-energy x-rays to treat tumours in the lung
Int J Radiat Oncol Biol Phys
(1990) - et al.
The influence of air cavities on interface doses for photon beams
Int J Radiat Oncol Biol Phys
(1993) - ICRU (International Commission on Radiological Units) Report 50. Prescribing, recording, and reporting beam therapy....
- Mayles W, Lake R, McKenzie A, et al., eds. Report No. 81. Physical aspects of quality control in radiotherapy. York:...
- et al.
A pencil beam model for photon dose calculation
Med Phys
(1992) - et al.
Application of AAPM radiation therapy committee task group 23 test package for comparison of two treatment planning systems for photon external beam radiotherapy
Med Phys
(1997)
The dosimetric verification of a pencil beam treatment planning system
Phys Med Biol
Cited by (22)
National dosimetric audit network finds discrepancies in AAA lung inhomogeneity corrections
2015, Physica MedicaCitation Excerpt :Pencil Beam [5,6] (PB) Collapsed Cone [7] (CC) Generally, dose calculation algorithms fall into correction-based equivalent path-length algorithms and model based algorithms [8] with electron transport either explicitly accounted for, or neglected.
Intensity-modulated radiotherapy for lung cancer: Current status and future developments
2014, Journal of Thoracic OncologyClinical implications of different calculation algorithms in breast radiotherapy: A comparison between pencil beam and collapsed cone convolution
2014, Physica MedicaCitation Excerpt :This approach, adopted in Pearson [12] and van Elmpt [28] studies for the treatment of lung cancer, showed that it's possible to obtain an equivalent uniform dose even with a minimal PTV coverage of 90% instead of 95%. Other authors [12,29] choose to deliver 95% of prescribe dose not to PTV but to CTV, in order to avoid any field size increase and arguing that, in lung tumors, if high density CTV moves into a part of PTV shown to be under-dosed, it will actually receive a higher dose than the planning isodoses indicate. As regard to breast treatment, a field sizes enlargement represent a major departure from clinical experience, with a significant increase of lung volume irradiated at all dose levels that might cause unacceptable toxicities.
Oesophageal Chemoradiotherapy in the UK-Current Practice and Future Directions
2013, Clinical OncologyCitation Excerpt :Although still not always possible to meet International Commission on Radiation Units and Measurements recommendations of target volume coverage, this algorithm is strongly recommended in oesophageal radiotherapy planning. In lung cancer cases, the clinical relevance of striving to achieve 95% coverage of the PTV has been questioned [5,35,36]. A ‘dense’ CTV will receive the required dose, regardless of its movement within a suitably sized PTV that extends into low-density tissues.
Dosimetric evaluation of a conformal seven-field coplanar technique for planning lung stereotactic body radiotherapy
2012, Clinical OncologyCitation Excerpt :The dose distribution was calculated and evaluated on all slices using the enhanced collapsed cone algorithm, a type B algorithm [20], with full inhomogeneity correction based on individual pixel Hounsfield values and a 0.25 cm dose calculation grid. This algorithm is recommended for dose calculation with low-density tissue inhomogeneities over a pencil beam (type A) algorithm, as the latter may lead to insufficient dose to the target volume [21,22]. Type A algorithms suffer from an inability to model lateral electron transport, giving less accurate results than type B, especially in the presence of lung [23,24].