Treatment Planning and Quality Metrics

14.1

Introduction

The overall goal of the treatment planning process is to produce the optimal dose distribution for the patient, taking into account the following factors:

• 1.

  Treatment intent (curative or palliative)

• 2.

  Stage of disease (extent of involvement)

• 3.

  Other therapies (chemotherapy, hormones, etc.)

• 4.

  Previous treatments

• 5.

  Reproducibility (immobilization, patient comfort)

• 6.

  Deliverability (collisions, modulation)

• 7.

  Safety (sensitivity to changes, robustness)

The process can be very complicated and involves a multidisciplinary approach. A workflow process map for external beam radiation therapy (EBRT) and brachytherapy can be found in the American Association of Physicists in Medicine (AAPM) white paper on incident learning. From the very beginning, the physician, radiation therapist, medical physicist, and dosimetrist must have clear communication regarding the factors mentioned above. As much information as possible should be documented in the medical record so that the sequence of events and current status of the planning process is clear to everyone. This begins with the immobilization and setup ensuring reproducibility and patient comfort. Again, clear communication is important. For instance, if a patient has had shoulder surgery and cannot comfortably raise that arm above the head, other immobilization strategies may need to be employed. Likewise, if a patient becomes agitated during the CT simulation because of claustrophobia, the therapist will need to communicate with the physician and nurse so that steps can be taken to reduce the patient's anxiety. This may involve medication or the use of relaxation techniques. If either method is required, will it be continued during treatments? This should be communicated and well planned. Several references describe the treatment planning process in Radiation Oncology. A schematic of the simulation and treatment planning process is shown in Figure 14.1 .

The ability to reliably reproduce the treatment over the entire course is the key to successfully delivering radiation therapy. The first step in achieving that goal is a careful setup and immobilization of the patient that maximize comfort while restricting motion. This is discussed in detail in Chapter 6 .

After the patient has been positioned, images can be obtained that are used to define the targets and critical structures needed to safely plan the treatment course. In many cases additional imaging studies are used in order to get information about soft tissue definition, functional information, or metabolic information that is unavailable from the primary image set. These image sets must be coregistered to preserve accurate geometric representation of the data.

Accurate contouring of the target(s) and organs at risk (OAR) is critical to developing a high quality treatment plan. If targets are contoured too liberally, the OARs may receive unnecessarily high doses. Likewise, if OARs are expanded to a degree that is inconsistent with the setup and localization technique being used, the dose to the target may be unnecessarily compromised. The expansion of OARs to account for setup uncertainty and patient motion and the creation of planning volumes is discussed in more detail in Section 14.4.

The physician treatment intent should contain the total planned dose and fractionation and should inform the priorities regarding dose goals and constraints. The stage of the disease will determine the extent of coverage, such as which nodal chains need to be covered. Again, communication is important. For instance, if the physician tells the dosimetrist that it is acceptable to compromise on target coverage to ensure that all OAR goals are met, a note should be made in the medical record so that another dosimetrist taking over the planning or a physicist reviewing the plan will also understand the goals.

A plan must be consistent with the capabilities of the delivery system in order to be delivered safely. For instance, if it was determined that the reproducibility of patient setup and isocenter placement is 3 mm for a given immobilization and imaging technique, the plan margins should reflect this (i.e., do not use 1 mm margins).

After the contouring is complete and the delivery method has been selected, e.g., 3D versus intensity-modulated radiation therapy (IMRT), the plan can be optimized. This can be a manual trial and error process for 3D or an inverse planned IMRT process. In either case, the desired dose goals must be clearly stated. The planner should be aware of the delivery equipment limitations to avoid creating a plan that is undeliverable. Some of these limitations (gantry speed, dose rate, minimum monitor units (MU), maximum MU, etc.) may be input into the treatment planning system (TPS) to prevent creation of plans that exceed limits. Collision detection is not implemented as of this writing. Plans with a high degree of modulation may push systems to their limit and, while possible to deliver, may lead to excessive interlocks.

After the plan has been optimized, the various dose constraints and goals are evaluated to determine if the plan is acceptable. If they are not, beam parameters such as gantry angles or number beams and dose constraints may be modified before another optimization is performed. There may be many iterations before an acceptable plan is achieved. A study by Nelms et al. showed a wide degree of variation of plan quality given the same planning input data. This shows the need for better tools to optimize plans and to provide guidance on what is achievable given a patient's unique anatomy and diagnosis. These will be discussed further in Section 14.4.

By using a multidisciplinary approach, with clear communication and documentation, and taking into account the factors mentioned above, the probability of delivering quality radiation therapy is increased. In the following sections each step in the treatment planning process will be examined. As mentioned, patient positioning and immobilization are covered in Chapter 6 , and CT simulation is covered in Chapter 13 and will not be covered in detail in this chapter. In addition, the specifics of IMRT and volumetric modulated arc therapy (VMAT) are discussed in Chapter 16 .

14.2

Patient Positioning and Immobilization

As described in Chapter 6 , the goals for patient immobilization include:

• 1.

  Reduce patient motion and improve day-to-day reproducibility of setup.

• 2.

  Improve patient comfort.

• 3.

  Accommodate the requirements of the physical devices.

• 4.

  Avoid normal tissue.

The reader is referred to Chapter 6 for further details.

14.3

Image Acquisition and Registration

Computed tomography (CT) is the primary imaging modality in radiation therapy. The CT simulation process is described in Chapter 13 . CT does not always yield the best soft tissue contrast for target and organ delineation even with the use of contrast agents. It is currently required for treatment planning to accurately calculate dose because it is the only imaging modality that provides density and therefore attenuation information. High density materials such as hip prostheses and dental implants can cause significant streaking artifacts. These can cause difficulty in the identification of targets and organs at risk. The impact on the dose calculation is minimal in the case of hip prostheses but has been shown to be significant for dental fillings. Magnetic resonance (MR) imaging may be useful in these cases. If available, megavoltage CT (MVCT) can be used to reduce the artifacts due to the decrease in the photoelectric effect at the higher energy. Newer CT scanners have the ability to reconstruct data outside the primary fan beam to create an expanded field of view (FOV). This can aid in contouring the entire external contour of the patient. The density in the expanded areas has been shown to be inaccurate, so care should be taken when used for planning.

Positron emission tomography (PET)/CT scans are often used to obtain metabolic information that would indicate tumor activity. The scans take a relatively long time to acquire, involve the use of radioactive material, have low signal-to-background ratios in some cases, and have poor spatial resolution. The use of PET/CT relies on the use of well-defined protocols for image acquisition, reconstruction, and segmentation.

Investigators are working on methods to correlate MR signals to density. If this can be done then MR could be used as the sole modality for planning, which would allow the clinicians to take advantage of the superior soft tissue definition of MR as well as to eliminate the use of ionizing radiation. MR imaging can be used for delineation of soft tissue, obtaining metabolic and functional information, and monitoring the response to treatment. The last use would take advantage of the lack of radiation dose, and thus the scans could be safely done at a higher frequency. Proton spectroscopy can sometimes be used to identify malignancies or necrotic tissue. MR simulation would have the disadvantages of potentially lower geometric accuracy, smaller FOV, much longer acquisition times, and higher cost.

The use of two or three imaging modalities for a single patient is not uncommon because the modalities have different strengths as described above. To accurately combine these images to obtain a complete anatomic, metabolic, and functional picture requires an accurate registration process. This can be particularly challenging if some of the imaging is not done with the patient in treatment position. Image registration algorithms are either rigid or deformable. Rigid registration is a simple geometric translation and rotation of one image set to align with another. It cannot account for differences in patient position or motion. Deformable registration maps one image set to another using a transformation matrix to produce a deformation map. This allows for differences in patient positioning, motion, organ size change, and organ shape change. The validation of deformable registration is very complex. AAPM TG-132, Use of Image Registration and Data Fusion Algorithms and Techniques in Radiotherapy Treatment Planning, is developing a report that will review existing algorithms and discuss issues related to implementation, methods to assess the accuracy of registration, issues related to acceptance testing, and quality assurance (QA). Various publications have looked at the accuracy of different algorithms. In general, no algorithm performs with the same accuracy in all situations so they must be evaluated for each clinical application.

14.4

Anatomy Definition

The definition of anatomic structures is crucial to achieving high quality radiation therapy delivery. The contouring of targets and OARs can be a time-consuming process in, for example, head and neck IMRT cases or relatively simple in the case of palliative spine treatments.

The issue of standardizing structures and reporting is the topic of two important reports: ICRU-83 on prescribing, recording, and reporting photon-beam IMRT and the American Society for Radiation Therapy (ASTRO) 2009 Recommendations for Documenting IMRT Treatments. Though these reports specifically name IMRT in their titles, they apply to other techniques as well. AAPM in June, 2014, formed AAPM TG-263, Standardizing Nomenclature, which should provide a report in the near future.

The structures outlined in Table 14.1 are a modification and update to ICRU-50 and -62, which defined these concepts. ASTRO's 2009 IMRT report recommendations call for the clinician to specify all the volumes in Table 14.1 except for RVR and TV, which are planning structures.

The process sometimes begins with the contouring of the external patient contour, though not all planning systems require this. Some systems do this automatically on the import of the images. The contour should be examined for accuracy because it will determine the depth of treatment and ultimately the MU. Typical areas of concern are around the mouth and nose where the contour can wrap inside the body, wires or markers placed on the skin, and immobilization devices. Most systems do not account for any structure outside the body contour, so the immobilization devices will have to be contoured as part of the body if the dosimetric impact of the devices is to be calculated. Some systems now have the ability to add the treatment couch to the image set to properly account for couch top attenuation. Some planning systems (e.g., TomoTherapy, Pinnacle) account for every structure in the scan FOV and therefore inherently calculate the attenuation of the immobilization devices. One caution when using a system with that capability is that some CT imaging artifacts create a bright ring around the edge of the FOV. This will be seen as a high density region, and the MU will be artificially increased. Careful evaluation of the image set should be done by zooming out to show the maximum FOV and switching to a lung window to make the ring more obvious. Some TPSs have the capability of filtering out low density artifacts outside the body contour by applying a threshold.

Contouring of targets is done by the physician, who will typically contour a GTV or CTV. The dosimetrist or physicist can then create a PTV using expansion criteria specified by the physician. Margin expansion will be discussed in more detail below. The physician may be aided in target delineation through the use of an atlas-based segmentation algorithm which uses deformable registration from an “expert” case. The final contours must be carefully reviewed and modified as necessary by the physician.

Contouring of OARs is typically done by both dosimetrists and physicians. Dosimetrists often contour bony anatomy, spinal cord, lung, and other easily identifiable structures such as kidney and liver. This can be done using manual techniques, atlas-based segmentation, auto-segmentation, or model-based segmentation. Auto-segmentation relies on following density boundaries and requires the user to place a seed point or bounding volume. These work well for organs with high contrast to surrounding tissue such as the lung. Model-based segmentation algorithms have the typical properties of an organ relating to size, shape, and density that are used to determine the contours. They can also use a seed point or bounding volume. Again, these automated contours must be reviewed and modified as necessary.

Inaccurate contouring can lead to suboptimal plans, delay the planning process by making the achievement of dose goals difficult or impossible, or cause geographic misses. Accurate target and OAR contouring is key to high quality planning. The importance of correct contouring is emphasized in the ASTRO report Safety Is No Accident and the safety white paper on peer-review in which the role of physician peer review of the contours is highlighted. It must be recognized that if contours are not reviewed until after planning (e.g., at new patient rounds or “chart rounds”), the threshold for rejection will be much higher because the planning process will have to be repeated. This has led some groups to explore peer review prior to planning, and this is advocated in Safety Is No Accident. Contouring atlases are available to assist in delineation of volumes. NRG Oncology has contouring atlases available on their website as well as links to atlases for the Radiation Therapy Oncology Group (RTOG), Gynecologic Oncology Group (GOG), and National Surgical Adjuvant Breast and Bowel Project (NSABP) at http://www.nrgoncology.org/Resources.aspx . There are other atlases that have been described in the literature. CTV definitions in atlases, for example, may vary based on staging.

One crucial parameter in treatment planning is the margin that is used to expand from the CTV to PTV. A margin that is too small risks underdosing the CTV. However, if 100% dosimetric coverage of the CTV in all patients were required, a very large CTV-to-PTV margin would be required resulting in higher OAR doses. Clearly some compromise is needed. A solution to this problem comes in the formulation of “margin recipes.”

14.5