A New Approach to Reduce Number of Split Fields in Large Field IMRT

Article Outline

• Abstract
• Introduction
• Methods and Materials
  • Pre-IMRT procedure: 3D open field calculation and modification of field gantry angle, jaw position, and collimator angle
• Results
• Discussion
• Conclusion
• References
• Copyright

Abstract 

Intensity-modulated radiation therapy (IMRT) has been applied for treatments of primary head with neck nodes, lung with supraclavicular nodes, and high-risk prostate cancer with pelvis wall nodes, all of which require large fields. However, the design of the Varian multileaf collimator requires fields >14 cm in width to be split into 2 or more carriage movements. With the split-field technique, both the number of monitor units (MUs) and total treatment time are significantly increased. Although many different approaches have been investigated to reduce the MU, including introducing new leaf segmentation algorithms, none have resulted in widespread success. In addition, for most clinics, writing such algorithms is not a feasible solution, particularly with commercial treatment planning systems. We introduce a new approach that can minimize the number of split fields and reduce the total MUs, thereby reducing treatment time. The technique is demonstrated on the Eclipse planning system V7.3, but could be generalized to any other system.

Key Words: Large PTV, Large-field IMRT, Split fields

Back to Article Outline

Introduction 

Four head and neck and 6 pelvic patients with planning target volumes (PTVs) larger than 14 cm in width were included in this study. For each patient, 2 sets of plans were performed. Our conventional intensity-modulated radiation therapy (IMRT) approach uses 7 gantry angles and permits the treatment planning system (TPS) to split the fields when necessary. Our modified approach maintains the same number of gantry angles, but we performed a 3D “pre-IMRT” before optimization. The X jaws were fixed so that the field size was also fixed to be 13.9 cm, just below the threshold for a split field. For targets with a field size >13.9 cm, dose coverage is ensured by delivery from other gantry angles. The y jaws, collimator and gantry angles were optimized for target coverage as in conventional forward planning. The goal of the preplan is to have ∼90% of the PTV covered by the 80% isodose level and fully covered by the 75% isodose level. Two optimizations were then run. During the first optimization, the jaws were opened and were set by the algorithm. This acted as the control configuration, i.e., that which most clinics conventionally run. The second optimization was run with the X jaws fixed at 13.9 cm to acquire a final fluence map that was run through a segmentation algorithm. We compared both plans in terms of PTV coverage, as well as dose constraints for organs at risk (OARs). In addition, we delivered both plans to a phantom in clinical mode to assess the effect on treatment time. All plans were performed and delivered using sliding window.

The PTV coverage from the conventional and modified plans is essentially identical for both sites studied. The mean dose difference was <0.2% and V95, D5 are <0.01%. The difference in the gradient across the PTV was also <0.2% on all patients. Dose-volume histograms (DVHs) for OARs were within 2% based on the planning DVH constraint for all organs. The bone marrow DVH met the constraints and were within 5% between modified and conventional plans. On average, the total treatment time decreased by ∼2.7 mins (range 1.2–3.4 min) or ∼27% of the beam on time. The total monitor units (MUs) also decreased 18% and 17% for head and neck and pelvic patients, respectively. Planning time increased by less than 1 hour per patient in general.

Reducing split fields can lower the treatment time and increase patient throughput on a linac. Peripheral dose to the patient is reduced, resulting from the decreased number of MUs.

IMRT was successfully applied to head and neck, breast, lung, pelvis, and total body.¹, ², ³ Because some of these target dimensions were larger than the maximum limitation of multileaf collimator (MLC) opening on the Varian machine (Varian Medical Systems, Palo Alto, CA), the IMRT field was divided into several subfields that have field sizes <14 cm.⁴, ⁵ The intensity fluence map of these subfields were also allowed to overlap 2 cm across with each other. This dynamic “feathering” technique had one subfield intensity decrease gradually across the overlap region while the other subfield increased slowly so that the sum of the subfields at the overlap region was the same as the original fluence map. The drawbacks of splitting large field were to increase total MUs, double the number of treatment fields, increase treatment time and paperwork, and consequently reduce patient throughput. Kamath et al.⁶ created a new split field algorithm and reduced the total MUs by 26% compared with the CORVUS TPS. Unfortunately, the segmentation algorithm was implemented in their in-house planning system and had not been commercialized and could not have widespread use. As IMRT was applied to the treatments of primary head tumor with neck nodes, prostate cancer with pelvis nodes, and lung cancers with large tumors, the time spent treating patients with split fields is an issue in busy clinical settings.

This study presents a method to reduce the treatment time by minimizing the total number of split fields on the regular commercial IMRT TPS.

Back to Article Outline

Methods and Materials 

Four head and neck and 6 pelvic IMRT patients with PTVs >14 cm from each field's beams eye's view were chosen for this study to ensure all the fields would split into 2 or more subfields. All patients were treated in the supine position. Pelvis patients were treated using vac-locs to position their legs, and head and neck patients were immobilized using a head cup, immobilization mask, and shoulder restraints. The gross tumor volume, clinical target volume, PTV, and normal tissues (NTs) were contoured by the physician. To confine the dose within the PTV, an avoidance structure called NT was drawn as body contour without PTV plus 0.8-cm margin.

Treatment plans were performed using the Eclipse TPS version 7.3 (Varian Medical Systems) with analytical anisotropic dose calculation algorithm. All IMRT plans were delivered with a Varian Linac 2100CD, which was equipped with a 120-leaf pair MLC. All treatments were delivered in clinical mode at 400 MU/min and were monitored by the Varis record and verify system to mirror clinical operations.

A treatment planning procedure was designed to minimize the planning time and provide a process that could be generalized to other TPSs. An overview of the process used in this study is provided in Fig. 1.

• 
  • View Large Image
  • Download to PowerPoint

• Fig. 1. 

  Nonsplit field IMRT planning flow.

At the beginning of treatment planning, the treatment field angles and total number of the fields followed our regular 7-field IMRT planning field arrangement. Because the span of the MLC travel was smaller than PTV dimension, to prevent field splitting, it was necessary to limit the x-direction field size <13.9 cm. For head and neck patients, the gantry angles were 180, 235, 285, 330, 30, 75, and 125° based on the Varian IEC scale. The corresponding gantry angles for pelvis patients were 180, 240, 285, 325, 35, 75, and 120°. The collimator angle was initially chosen to be 0°, but this could be changed. One of the x jaws was fixed at each gantry angle, with the exception of gantry = 180, where neither jaw was fixed. We chose to fix the jaw that was farthest away from the PTV based on the path length of the field edge as the principal that radiation near this field edge will contribute less to the overall dose (due to attenuation and inverse square effects). The loss in coverage from fixing this jaw will be more readily compensated from other fields.

Pre-IMRT procedure: 3D open field calculation and modification of field gantry angle, jaw position, and collimator angle 

Three-dimensional open-field dose calculations were run before IMRT optimization, when the total number of the fields and the gantry angle and jaw position of each field were decided. The premise was that sufficient dose coverage was necessary from an open field to permit the optimization algorithm to produce an acceptable plan. During this “pre-IMRT” procedure, each field was weighted equally and the dose normalization point was fixed at the field isocenter. Guided by the beams-eye-view display of the TPS (Fig. 2), the jaw positions and collimator angles were iteratively adjusted to ensure the PTV was enveloped by the 80%–75% isodose volumes. This choice of a 75%–80% dose coverage to the PTV was determined empirically. Care was taken to strategically cover parts of the PTV that were missed from other gantry angles, thus ensuing that our goals were met. This “pre-IMRT” procedure serves to ensure that the optimization engine and leaf motion calculator have sufficient dose to modulate on a nonsplit distribution before optimization begins. We maintained the constraints that the field width would remain within 13.9 cm and the collimator angle rotation between sequential fields was <10°. We limited the collimator angle rotation to ensure that treatment delivery time would not be limited by waiting for the collimator to reach a new position. If it was impossible to have the 75% isodose curve cover the PTV, some of the field widths would not be limited to 13.9 cm, which enabled us to allow the field to cover the entire PTV, but would also result in a split field.

• 
  • View Large Image
  • Download to PowerPoint

• Fig. 2. 

  Example of collimator settings to avoid split field plan. A 19 cm wide PTV is planned with seven IMRT fields. The central figure shows an axial cut of the pelvis with the PTV in red. The yellow lines represent the field borders and central axis of each of the beams. The corresponding beams eye views are shown encircling the central figure. In these, the yellow frame represents the collimator settings. Although the PTV is not completely covered by any field other than the PA field, the parts that miss coverage from one field are covered by the others.

After the pre-IMRT procedure established the field widths and gantry and collimator angles, the IMRT optimization could proceed. Two IMRT plans—“nonsplit-field and split field”—were created to compare the PTV coverage and NT sparing between nonsplit-field and split-field plan. The optimization constraints remained the same for both split and nonsplit plans.

The PTV coverage on both split-field and nonsplit-field techniques were evaluated with D5 and V95 because the mean dose was similar for both plans. D5 is a dose value that covers 5% of the PTV volume that could indicate hot dose volume within the PTV. V95 is the percentage PTV volume covered by 95% of the prescribed dose, which could indicate the quality of PTV coverage. Critical organs and NTs sparing were analyzed based on the Dmax and mean dose.

The additional treatment planning time was recorded to include the time spent editing the jaw positions. Total MUs on both split-field and nonsplit-field techniques were recorded and compared. The higher total MUs would result in larger peripheral dose, i.e., the dose to NT outside the field. The treatment delivery time, defined from the start of the treatment data download from the record-and-verify system for the first field to the completion of dose delivery, was recorded. Thus, patient setup was not included within the treatment time evaluation, but this would be the same regardless of the treatment method.

To evaluate the parameters of interest (e.g., mean dose), differences between the conventional approach and our nonsplitting approach, we define a metric R as the ratio between split-field and nonsplit-field plans:

(1)

Target coverage was evaluated using the ratio of mean dose, the percentage of PTV covered by 95% of prescription dose (V95), and the high-dose volume that covers 5% of the PTV (D5). Normal tissue comparisons were made for maximum doses to NT, spinal cord, and brainstem doses for the head and neck patients. For pelvis patients, the ratio of maximum and mean doses to bladder, rectum, bone marrow, NT, and bowel were calculated. We also calculated the ratio of total MUs and delivery times.

Back to Article Outline

Results 

The PTV coverage between the 2 techniques was essentially identical, regardless of site. Specifically, the mean dose of the PTV was within 0.5% on both head and neck and pelvis patients. The ratio of V95 between all the head and neck and pelvis patients were within 0.05%, and the ratios of D5 between split-field and nonsplit-field were within 1%.

Figure 3A shows the ratio of the split-field to nonsplit-field maximum doses to the spinal cord, brain stem, and NT for the head and neck patients. In most cases, the differences are <2% and the maximum dose is lowered with our new technique. For all plans, split-field and nonsplit field, the clinical constraints (e.g., <45 Gy to spinal cord), were met.

• 
  • View Large Image
  • Download to PowerPoint

• Fig. 3. 

  The normal tissue Dmax ratio of H&N patients. The ratio higher than 1 means the split-field has higher Dmax dose than the nonsplit field.

Figure 3B shows the ratio of maximum doses to OARs for the pelvis patients. Most NT Dmax differences were within 1% and <1 Gy. Most NT dose ratios were higher than 2%, which meant the split field had higher NT dose than the nonsplit field.

Figure 4A shows that for the head and neck patients, most split-field plans have higher mean dose than the nonsplit-field to the OARs, exceeding 5% for parotid gland for patients 1 and 2. Figure 4B shows that for most pelvis patients, bladder, bowel, and bone marrow mean dose ratio were within 2%, but the NT mean dose ratio was close to 5%.

• 
  • View Large Image
  • Download to PowerPoint

• Fig. 4. 

  (A) The H&N normal tissue mean dose ratio of the split-field to the nonsplit field. The ratio higher than 1 means the split-field has higher dose volume than the nonsplit field. (B) H&N normal tissue mean dose ratio of the split-field to the nonsplit field.

The total MU ratio was consistently larger than one (mean = 1.18) and ranged from 1.12–1.32 for all patients. These values reflect a reduction in the beam on time. Reduction of MUs will in turn lead to reduction in peripheral dose because of leakage radiation from the head of the accelerator and an increase in efficiency of delivery.

The delivery times are listed in Table 1. Delivery times correlated with the MU reduction, but not linearly, because of other factors including collimator motions, gantry rotation, and field splitting. Specifically, the entire nonsplit-field had reduced the treatment time and the range from minimum 20% to maximum 60%. Delivery times were reduced by an average of 2.7 minutes.

Table 1. Head and neck and pelvis treatment delivery time (sec) and time ratio of the split-field to the nonsplit-field techniques

Patient	Delivery Time (sec)		Ratio
	Split-Field	Nonsplit-Field
Head and neck
1	530	378	1.40
2	600	410	1.46
3	528	440	1.20
4	595	413	1.44
Pelvis
1	525	433	1.21
2	559	420	1.33
3	569	447	1.27
4	665	412	1.61
5	609	419	1.45
6	623	418	1.49

Back to Article Outline

Discussion 

All plans were designed to have the same PTV coverage; therefore, the differences between the 2 planning techniques were evident only on the OARs. For most OARs, the difference in mean and maximum doses was approximately 2%. In general, split-field plans had a lower maximum point dose and a higher mean dose.

Conversely, most of the split-field plans had higher mean dose value than nonsplit-field plans. One of the reasons split-field plans had a higher mean dose was because they had a higher transmission dose through the MLC. Beyond the higher transmission dose, nonsplit-field would also limit the dose contributed to the NT because of total MU reduction that would reduce the peripheral NT dose. Note that this quantity is not calculated in the TPS but varies linearly with MUs.

For some structures, such as the parotid gland, the nonsplit-field technique resulted in the gland being shielded by the jaw, with a 1% transmission factor for some fields compared with a 3% transmission factor for an MLC, resulting in significant dose sparing to the parotid gland.

The total MUs were reduced by about 19% on average. Because there were fewer subfields, dose delivery is more efficient because the MUs that would have to be delivered in the overlap region are not split into two, but instead are delivered in 1 sweep.

Treatment planning time was extended by approximately 30 minutes to perform the pre-IMRT processing, which is manageable in most clinical environments. We anticipate that with experience, the planner could establish the appropriate jaw opening, field gantry angle, collimator angles, etc., thus reducing the additional planning time.

On average the treatment time was reduced by 29% or about 3 min. Much of the time saved was because of the reduction in the number of treatment fields from 14 to 8. To deliver the radiation field, the linac had to reload the treatment data from the record-verify system and move the MLC to the default position for the delivery. Repeating these steps took additional time to complete and increased the total treatment time as number of fields increased.

The reduction in total treatment time has several advantages. It would benefit patients who cannot easily stay in the same position during treatment. Faster delivery would minimize the potential for these patients to move during treatment and reduce the probability of geometric miss.

In addition, some vendors are now supporting the volumetric arc theory on the basis that it improves patient throughput and reduces peripheral dose by reducing the number of MUs to the patient. This technique moves in the same direction without requiring additional software and hardware. Increases in throughput for VMAT have not yet appeared in the scientific literature, but our technique can produce a 3-min reduction per patient, which translates into an additional 2 15-min patient slots per 10-hour work day.

Moreover, aspects of this technique could be applied to volumetric arc therapy planning. The pre-IMRT planning could help alleviate the planning time by generating a process for quickly choosing the collimator angles. However, such an approach is speculative at this time and further investigation is warranted.

Back to Article Outline

Conclusion 

The nonsplit-field technique successfully reduced the total treatment time to 3 min and treatment MU by 19% without significantly changing the dose distribution on the PTV and NT DVH. The PTV mean dose difference between nonsplit fields with split field was 0.12%, the PTV V95 difference was 0.04%, and the V5 difference was 0.15%. The NTs, Dmax, mean dose difference were <1% and 2%, respectively.

Patient treatment with the nonsplit-field technique can reduce the total body dose because of the total MU reduction and can also reduce treatment time. Although, the nonsplit-field technique was tested on the Eclipse planning system (Varian Medical Systems) in this study, the same process could be implemented on different planning systems such as ADAC and Xio because these 2 systems have similar optimization algorithms and processes. Indeed, with the nonsplit-field technique, the planning time on these other systems could be reduced even further than on Eclipse because fields must be split manually before the optimization.

Back to Article Outline

References 

1. Chao C, Cheng J, Low D, et al. Intensity modulated radiation therapy provides better tumor target coverage and normal tissue sparing in patients with nasopharyngeal carcinoma (abstract). Int. J. Radiat. Oncol. Biol. Phys. 1999;45:420
   • View In Article
   • Full-Text PDF (142 KB)
   • CrossRef
2. Hong L, Hunt M, Chui C, et al. Intensity-modulated tangential beam irradiation of the intact breast. Int. J. Radiat. Oncol. Biol. Phys. 1999;44:1155–1164
   • View In Article
   • Abstract
   • Full Text
   • Full-Text PDF (1127 KB)
   • CrossRef
3. Mundt AJ, Lujan AE, Rotmensch J, et al. Intensity modulated whole pelvic radiation therapy in women with gynecologic malignancies. Int. J. Radiat. Oncol. Biol. Phys. 2002;52:1130–1137
   • View In Article
4. Hong L, Alektiar K, Chui CS, et al. IMRT of large fields: Whole-abdomen irradiation. Int. J. Radiat. Oncol. Biol. Phys. 2002;54:278–289
   • View In Article
   • Full Text
   • Full-Text PDF (26 KB)
   • CrossRef
5. Wu Q, Arnfield M, Tong S, et al. Dynamic splitting of large intensity-modulated fields. Phys. Med. Biol. 2000;45:1731–1740
   • View In Article
   • MEDLINE
   • CrossRef
6. Kamath S, Sahni S, Palta J, et al. Algorithms for optimal sequencing of dynamic multileaf collimators. Phys. Med. Biol. 2004;49:33–54
   • View In Article
   • MEDLINE
   • CrossRef