Delineation methods

The delineation of target volumes for NSCLC patients is problematic due to the extent of motion experienced throughout the breathing cycle. Various methods can account for this, such as MIP, AI end-of-expiration (EOE), end-of-inspiration (EOI) and ten-phase overlap approaches.Reference Reitz, Parda, Colonias, Lee and Miften² The ten-phase overlap approach delineates the target in each phase of the breathing cycle individually. This is clinically unsuitable as it can take ∼2·5 hours to perform.Reference Muirhead, McNee, Featherstone, Moore and Muscat¹⁰Table 1 summarises the relative advantages and disadvantages of the different delineation methods.

Table 1 Delineation methods within literature

Abbreviations: MIP, maximum intensity projection; AI, average intensity; IMRT, intensity modulated radiation therapy; 4D-PET-MIP, four-dimensional positron emission tomography maximum intensity projection; 4D-CT-MIP, four-dimensional computed tomography maximum intensity projection.

Bradley et al.Reference Bradley, Nofal and El Naqa⁹ noted that for stage 1 NSCLC, MIP was more reliable than either AI or the common 3D helical scan delineation types. However, it was noted that the accuracy of dosimetric calculations on MIP datasets is inferior to that of AI, as electron densities cannot be recorded. As such, MIP can be used to identify target position, whereas AI is required to represent the dosimetric outcomes of the patient's treatment accurately.

This was supported further by Muirhead et al.,Reference Muirhead, McNee, Featherstone, Moore and Muscat¹⁰ who found that MIPs performed on stage 1 patients were similar to that of the common ten-phase overlap approach, and thus should be used to decrease planning time. However, the study also showed that stage 2–3 disease resulted in MIP datasets to underestimate the target volumes. In this case MIP cannot distinguish boundaries between tissues with similar or higher Hounsfield Units, and therefore erroneously rejects these voxels.

Reitz et al.Reference Reitz, Parda, Colonias, Lee and Miften² contradicted both these studies, when reviewing IMRT planning of NSCLC patients in all four stages of disease. The authors concluded that MIP was superior to that of AI as it delineated smaller margins. However, the study consisted of only five patients, three of whom were stage 1. Larger scale study findings would be required to refute the other studies. Muirhead et al.,Reference Muirhead, McNee, Featherstone, Moore and Muscat¹⁰ noted that smaller margins created by the MIP were of concern, as volumes were not covered in all ten phases, thus signifying both a need for larger case size and for visual confirmation to check coverage of the entire breathing cycle.

It has been illustrated that positron emission tomography (PET) has improved the staging and treatment of NSCLC patients.Reference De Ruysscher, Nestle, Jeraj and MacManus¹⁵ Recently, Callahan et al.Reference Callahan, Kron and Schneider-Kolsky¹⁶ demonstrated how 4D-PET-MIP utilises the existing MIP technology and applies it to PET scans for the same purposes as used in CT. It was found that 4D-PET enabled more accurate positioning of target volumes in the scans of nine patients. The edges of the volumes were more defined, as was the overall position. The use of MIP to delineate the targets was found to be as accurate as the ten-phase overlap technique. Importantly, PET scans are more reliable near soft tissue structures, such as peripheral tumour placements, as the soft-tissue contrast is greater than in the CT-MIP technique. These findings were supported by a 2012 studyReference Hanna, van Sornsen de Koste and Dahele¹⁷ which concluded that 3D-PET did not accurately match 4D-CT-MIP delineations. Further investigations into the use of 4D-PET-MIP should be considered.

Overall, MIP proves an efficient method of delineating early stage target volumes. The drawbacks relating to dosimetric properties and accuracy during the latter stages of disease prove to be significant enough to warrant either the continued use of ten-phase overlap delineations, or an alternative method.

FB

Liu et al.Reference Liu, Balter and Tutt¹⁸ studied 152 patients of stage 2 and three NSCLC, in order to assess the 3D motion of target volumes. About 77% of patients did not receive surgery before planning and treatment, and 60–70% had lesions located within the central region. The motion detection in three dimensions was investigated as the authors noted sufficient literature stating target volume motion in the lung is anisotropic, disproving the use of universal margins. The breathing regularities of the patient, tumour size and location are directly linked to the degree of motion the target experiences (Table 2).

Table 2 Free-breathing planning and treatment methods within literature

Abbreviations: EOE, end-of-expiration; EOI, end-of-inspiration; PTV, planning target volume; MIP, maximum intensity projection; AI, average intensity.

Using EOE GTV delineation, FB 4DCT scans were monitored by Varian's RPM technology. The infrared markers on the abdomen allowed the system to track the breathing cycle effectively, measuring ten phases across the cycle. Visual verification was required once the rigid body deformation technique was applied to the fusion of the GTV phases. This allowed for the PTV to accommodate for tumour motion in each phase. The results showed a higher frequency of motion in the superior and inferior directions, and lower lobe lesions moved most often due to close proximity to the diaphragm. Additionally the smaller lesions were proved to move more than larger and superior lesions; this was linked to the effect of the diaphragm motion.

The report's findings were strengthened as it had the most significant sample size of the articles reviewed; 152 patients with 166 tumours made for reliable findings. It also reinforced the claims made by the authors of the effect and dependency of target motion according to anatomical and disease-specific characteristics. The position and motion of the targets is clearly relevant to the planning of many lung tumours. Many of the patients were of stage 3 or 4 NSCLC, whereas stage 1 and 2 are the most likely to have significant motion. A greater range of patient types could increase the validity of the results, to avoid skewing of results from the latter stages of the disease. Additionally, results could be separated into early and late stage disease. Future studies should investigate planning time, staff numbers and workflow changes required, as well as cost of technology and systems. Expected control and survival rates should also be compared with conventional methods of lung treatments.

Fritz et al.Reference Fritz, Kraus and Blaschke¹⁹ performed a study of single-dose fractionation of 40 stage 1 NSCLC patients that aimed to measure the side effects, control rates, survival rates and overall success associated with 4DCT and single-dose treatments. The authors found 4DCT to be more accurate than 3DCT for PTV delineation. Furthermore, accounting for motion in FB planning scans decreased cost compared with gated treatments, as no extra technology is required to be installed in the treatment area. The planning was performed through the use of three scans; EOI, EOE and mid ventilation. The combination of these three scans allowed for a projection of the extent of target motion during the breathing cycle.Reference Wolthaus, Sonke and van Herk⁷ A 10–15 mm margin was placed on the GTV. The results included 81% control rate at 3 years, with overall survival for 2 and 3 years at 66% and 53%, respectively. The cancer specific survival was 71% and 57% for 2 and 3 years post-treatment, respectively. The side effects to patients included a 75% rate of radiation pneumonitis, with no other significant toxicities or side effects. The strengths of this report include a significant patient sample size (n = 40), with similar type and stage of disease; excellent for result comparison. The target volume size was quite variable, which illustrated the abilities of the 4DCT and planning system to be applied to a variation of tumours. Patient follow up was also extensive and ongoing. The reporting of this data is critical when using IMRT where long-term effects are still to be determined. The use of 4DCT for single-fraction treatment of stage 1 NSCLC has not been reported elsewhere at the time of this review, therefore it is difficult to draw comparative conclusions.

Li et al.Reference Li, Li and Zhang²⁰ performed a study on 28 NSCLC patients, in an attempt to calculate the positional and volumetric differences of 3DCT and 4DCT planning methods. Three PTV's were created during the planning, one each for the different scan type and an additional one to measure the vector difference. The authors found that 4DCT planning has the potential to decrease the risk of target miss during treatment and normal tissue irradiation, by increasing the accuracy of margins for the patient. Using an individualised technique is superior as the motion of the target is accounted for, rather than a universal margin to encompass ‘possible’ motion.

The patients were grouped depending on the stage of their disease and location of the lesion. Varian's RPM system was used to monitor breathing cycles of patients, ‘binning’ ten phases of the scan. The physicians delineated the target volumes on the base slice of 50%, or end expiration. The results showed a 25–48% reduction of PTV size and a 30·8–48·4% decrease in normal tissue dose using 4DCT. The planning time for 4DCT was higher than that of 3DCT methods: 28 ± 15 minutes compared with 5–10 minutes, for GTV delineation. The report concluded that individualised margins are able to cover the majority of 4D target volumes; however this usually results in over-irradiating normal tissue. The study noted that to do this using 4DCT increases the workload of planning systems and patient exposure. Irregular breathing patterns of patients can also negate some of the benefits of 4D systems and a need for coaching or gated methods would be required. This can be problematic for NSCLC patients as the age and co-morbidities often found in patients can have significant effects on the breathing patterns.

The strengths of this paper are the comparable results over a wide variety of patient stages and pathologies. The position of the lesions, the stage of the disease and the multiple PTV delineations allowed for an extensive collation of data. The variations of motion in all three dimensions and the positional differences of lesions are also important to show the effects of 4DCT planning for the different presentations of disease.

Starkschall et al.Reference Starkschall, Britton and McAleer²¹ measured the differences between dose distributions and calculations of 3D and 4D CT planning systems for 15 NSCLC stage three patients. CT was combined with monitoring devices to form a 4D planning system, depicting ten phases of respiration. Deformable image registration was used on each of the ten phases, overlaying the EOE CTV volume over the nine other phases, in order to calculate an average dose distribution throughout the full motion of the target. The authors noted only a small increase in actual planning time, yet a significant increase in dose calculations, as there were ten datasets for the 4DCT compared with one for 3DCT. 4DCT PTV delineation resulted in an average reduction of 1·7% of volume compared with 3DCT. This illustrates the increased accuracy of the 4D scans, as there was an additional 1 mm margin added to the 3D scans for uncertainty. The strengths of this paper include the variability of target volume size and the use of deformable registration techniques, rather than rigid body methods. Rigid body is limited by approximations of the effect of motion, whereas the planning technique used in this study is similar to adaptive planning. The report was limited by the small case size (n = 15) and the use of 3D planning techniques in a 4D system. Although 4DCT was used, the custom 4D planning was effectively a copy of 3D methods transferred across using 4DCT-delineated volumes. The differences and limitations of gating methods compared with the 4DCT plan and the associated outcomes should be further investigated, as well as the effect of respiratory changes through treatment and the impact of those on the dose distribution results. By performing a 4D plan (i.e., using a 4D planning system) a more accurate representation of results could be provided. Control and survival rates would also be useful.

Reitz et al.Reference Reitz, Parda, Colonias, Lee and Miften² studied the use of simple IMRT for the planning of five NSCLC patients, ranging from stage 1–4 disease. Using FB 4DCT, the aim of the study was to quantify the feasibility of using five-field IMRT for NSCLC. The 4DCT scan was performed with an Anzai belt to monitor respiration during the scan. Target delineation was performed using maximum intensity (MIP) or mean intensity (MI or AI) methods.

The authors note that using EOE or EOI proves less accurate as it fails to account for ant-post and left-right deviations. It is unrealistic to perform delineations on ten phases of a patient's breathing cycle, due to time constraints in the clinical setting. Delineation using MIP proved more accurate and enabled margin reduction compared with that of AI plans. The margins for MIP plans were 0·8 cm while AI used a 1·5 cm margin.

The strengths of this report include the standard beam arrangement, dose tolerances and the use of tables and illustrations to concisely depict the gain of using MIP delineation over AI. The use of geometric equivalent uniform dose and TCP tabled information was beneficial, and proved one of the only pieces of literature to quantify the probabilities of the planning outcomes.

The limitations of MIP include total lung volume representation similar to that of EOE scans, not indicative of the actual lung volume during treatment. This provides far less lung volume during most of the treatment cycle (during FB). Additionally, electron densities cannot be gained from the MIP image, therefore using the AI data was required for both total volume calculations and accurate dosimetric readings.

Gating and tracking

Gated methods in combination with 4DCT scans provide more accurate alternatives to conventional 3DCRT planning of NSCLC. Commonly, a FB 4DCT is used to visualise the full extent of motion across the breathing cycle, and to use the most stable region of this cycle for treatment purposes. Treatments are delivered using either a breath-hold technique or FB, where the beam-on of the machine occurs at pre-determined stages of the respiratory cycle.Reference Washington and Leaver^5, Reference Underberg, Lagerwaard, Slotman, Cuijpers and Senan²² There are two approaches to gated treatments, where the breathing cycle is monitored by internal or external means. Internal gating requires invasive procedures for patients who are frequently unfit for such procedures. External gating can be performed using detection devices on the abdomen, spirometers and tracing methods.Reference Underberg, Lagerwaard, Slotman, Cuijpers and Senan²² Over 90% of lung patients can use external systems to achieve gating, compared with a significantly lower proportion for internal.Reference Keall⁸Table 3 summarises the studies relating to gated techniques.

Table 3 Gated planning and treatment methods within literature

Abbreviations: ITV, internal target volume; PTV, planning target volume.

The use of IMRT-gated treatments for NSCLC can increase treatment times by as much as 10 minutes.Reference Keall⁸ This is due to the beam-on time representing only 20% of the respiratory cycle, at the most stable point, usually end-expiration.Reference Muirhead, Featherstone, Duffton, Moore and McNee²³ Calculation and planning can be 15 times longer than the 3DCRT and non-gated approaches. With increased time, there is a risk the reduced patient comfort can cause movement. It is possible that cell-repair within the target may be possible when treatments such as stereotactic can reach 3 hours for a single treatment, with an average of 100 minutes. This raises questions as to the validity of such gated treatments if the time taken to accurately position, image and treat according to the breathing cycle can have a negative effect on the patient's quality of life.Reference Keall⁸

Underberg et al.Reference Underberg, Lagerwaard, Slotman, Cuijpers and Senan²² investigated the use of gated 4DCT planning and treatment for 31 stage 1 NSCLC patients. The aim of the study was to investigate the benefits of using isotropic margins and devising a method of future patient selection. A FB 4DCT scan was performed, using the Varian RPM device as a means to bin ten phases of the respiratory cycle. The extent of motion was observed from the maximum displacements on the 3D datasets. The ITVs included all phases of delineated GTV, therefore accounting for all motion. Isotropic margins were added to all three ITV's (10 mm, 3 mm, gating with 3 mm) to form variable PTVs for comparison. Eight to 12 non-coplanar beams were used for the dosimetry process; three separate fractionation schedules were used to view possible benefits for hypo-fractionated treatments. The report concluded that gating offered a 30% reduction of PTVs for 38% of targets and 50% in 15% of targets. The conclusions of this article were strengthened by an appropriate sample size and the patient selection methods. The reduction of margins illustrated by the normal tissue dose regions was effective.

It would be interesting to determine the impact of the technology on patient quality of life measures, such as side effects, control rates and survival rates as well as related costs, ease of use, staff and workflow requirements of RPM as well as impact on workload. The use of isotropic over individualised margins should be investigated further as literature has shown the latter to be more accurate.

A 2009 study investigated the use of SBRT using repeat scans and real-time tracking devices.Reference van der Voort van Zyp, Prévost and Hoogeman³ Using a breath-hold technique for the scanning of patients, repeat scans were used to confirm positioning of the markers inserted into the patient. Out of 70 patients with T-1 or T-2 stage NSCLC, two groups were formed: 59 received 60 Gy while the remainder received 45 Gy over three fractions. The control rate after 2 years was 96% and 78% accordingly, with overall survival at 62% and cause specific survival at 85%. Grade 3 toxicity was only experienced in 10% of patients. It was noted that the control rates were higher for those with T-1 disease, due to lower local recurrence risk and smaller target volumes. The paper highlighted the quality of life results, which has been overlooked in other literature. Control rates at 2 years proved high and in line with similar studies noted in the literature.