Imaging for Target Delineation and Treatment Planning in Radiation Oncology: Current and Emerging Techniques

1. Introduction

Radiation treatment has improved not only with technical advances in the treatment delivery machines, but also to great extent by the implementation of new imaging modalities and advanced image sequences.

In Radiation Oncology, imaging is not exclusively used for diagnosis and tumor response assessment. It also plays an important role in the delineation of target volumes and organs at risk as well as in treatment planning. Physician contouring is performed according to the recommendations made by the International Commission on Radiation Units & Measurements [ICRU 62; 78]^1,2. According to the ICRU, the Gross Tumor Volume (GTV) is the visible or clinical demonstrable location and extent of the tumor. The Clinical Target Volume (CTV) is the tissue volume that contains the GTV and/or subclinical malignant disease. The Planning Target Volume (PTV) is a geometric volume generated by adding a margin to the CTV to account for treatment-related uncertainties such as set up errors and organ motion.

For treatment planning, the vast majority of radiation treatment centers are equipped with dedicated computed tomography (CT) scanners that are specifically designed for radiotherapy. CT enables accurate dose calculation in treatment planning systems by using the relative attenuation of a series of x-rays to calculate the absorbed dose to tissue³.

In addition to CT, Magnetic Resonance Imaging (MRI) has become increasingly important, given the advantages of enhanced soft tissue contrast and no radiation dose. Newer approaches with the combined MRI linear accelerator (MR-Linac) enable improved accuracy of patient positioning compared to on-board x-ray and similar to cone beam CT (CBCT), as well as real-time MR imaging during radiotherapy to track motion-induced target variability⁴. MRI is also of great importance for target definition, especially in cases when CT provides poor imaging contrast between the tumor and surrounding tissues.

In the next sections, we provide a comprehensive overview of the main imaging modalities currently used in Radiation Oncology as well as new and promising techniques for target delineation and treatment planning.

2. CT Approaches

Prior to receiving radiation therapy, each patient undergoes CT simulation, which is primarily driven by the necessity of CT for dose calculation^5,6. Because CT is used for simulation and treatment planning, it is the primary image set used for contouring.

A major advantage of CT imaging is its fast acquisition times, which minimizes motion artifacts from breathing or swallowing^7,8. CT is exceptionally effective at visualizing high-Z materials (e.g. bone) and has uniform distortion across the acquisition field. However, the utility of CT is limited by its poor visualization of soft tissue, which causes difficulties in distinguishing between tumors and the healthy surrounding tissues. To account for this, complimentary imaging that adds anatomical and functional information like MRI are often fused with the primary CT imaging.

Nevertheless, target volume definition is considered the largest uncertainty in modern radiotherapy⁹. Both imaging-related factors (i.e. modality, technique, and interpretation) and patient-related factors (i.e. organ motion and voluntary motion) can influence the accuracy of target definitions and are considered to contribute to both intra- and interinstitutional observer variations^9–11. Substantial inter-observer variations have been found for all treatment sites and all target volumes relevant to radiotherapy planning, of which some had significant impact on target dose coverage and OAR doses¹². Likewise, large inter-observer and interinstitutional contouring differences were found in a multicenter RTOG study of structure delineation for breast cancer on CT. They reported variations in contoured volumes with standard deviations of up to 60%, and also dosimetric consequences for OARs¹³. Errors in target definitions are systematic in nature, thus the effect remains constant during the treatment course and may, in worst case scenarios, lead to under-dosage of the target and poorer local control and survival outcomes^14,15.

Efforts to reduce observer variations in target definitions include improved imaging techniques, the use of contouring guidelines and atlases, standardization of training, auto contouring and peer review. The use of written guidelines has been found to significantly improve consistency in target delineation and, in some situations, to reduce dose to organs at risk¹⁶. Guidelines exist for tumor and nodal contouring on CT and are available through the Radiation Therapy Oncology Group Core Laboratory for numerous treatment sites including lung¹⁷, breast¹⁸, head and neck¹⁹ and prostate²⁰.

Training interventions have also led to significant improvements in inter-observer variability; target delineation courses have been shown to result in both smaller and more homogenous target volumes among radiation oncologists²¹. Training and teaching that includes anatomical lessons followed by contouring practice with individual feedback is considered most successful¹⁶.

Further improvement in target delineation can be achieved by a clinical peer review process on radiation treatment plans, including information on patients’ disease characteristics, diagnostic imaging review, and clinical and visual inspection of the patient²². While contouring guidelines have led to improved standardization in target delineation, the contributions of technological innovations in CT have further improved target delineation accuracy through improved tumor-to-normal-tissue contrast and motion management. In a new technique called four dimensional CT (4D-CT), a series of 3D CT images are correlated with the patient’s breathing cycle over time to produce a motion model of the tumor. 4D-CT simulation is now the standard of care at many hospitals for patients with lung, esophageal, and liver tumors due to improvements in motion management during radiation treatment, which ultimately improves tumor localization and reduces normal tissue doses²³.

4D-CT can be used in several ways to improve motion management. If the scan indicates a relatively small range of motion of the tumor through the breathing cycle, then the physician may elect to treat the patient free breathing but to account for the motion during treatment planning by expanding the CTV to create the internal target volume (ITV). The motion margin is typically determined by population-based studies. However, with 4D-CT, the physician can define a patient-specific ITV that includes the full range of motion of the tumor. This patient-specific ITV approach is used in combination with free breathing treatment for small ranges of motion indicated by 4D-CT imaging. However, if 4D-CT imaging indicates substantial motion, then strategies to reduce motion effects such as breath-hold and phase-based respiratory gating^24,25 are needed. Comparison studies of 4D-CT motion management versus conventional 3D-CT have shown improved target coverage and subsequent reduced normal tissue dose^26,27.

Other approaches for improving the accuracy of tumor delineation with CT are contrast administration and dual-energy CT (DE-CT). CT target delineation relies on inherent differences in the density and attenuation of tumors relative to surrounding soft tissue. However, when tumors such as liver tumors have similar attenuation characteristics as the surrounding tissues, intravenous (IV) contrast can be used to enhance the tumor visibility²⁸. Differences in the uptake, retention, and washout of IV contrast can be helpful for identifying malignant disease. The combination of baseline CT imaging without contrast and subsequent imaging over time with IV contrast, so called Dynamic Contrast Enhanced CT (DCE-CT)²⁹, can further increase tissue contrast and reduce target delineation variation in liver tumors.³⁰. In addition, for treatment sites where both organ motion and low tumor-to-healthy-tissue contrast is an issue (e.g. in esophageal cancers), improvement in target definition may be achieved by using contrast enhanced 4D-CT scans³¹.

DE-CT uses both low (typically 80 kVp) and high energy (typically 140 kVp) x-rays. Tissue attenuation is dependent on not only material characteristics, but also photon energy. Having two data points rather than one can improve the identification of tissue composition³². While DE-CT is mainly used for diagnostic applications, its use is becoming more common in radiotherapy departments³³. DE-CT offers the advantage of improved image quality through superior material separation and metal artifact reduction^34,35. The subsequent impact on dose calculation accuracy, particularly in brachytherapy³⁶, proton therapy³⁷ and tumor delineation accuracy^38,39, is being investigated.

3. MRI approaches

For target delineation, MRI is primarily used in cases where CT does not reveal sufficient contrast between the tumor and surrounding structures, like for primary brain tumors and cerebral metastases. In this case, all contouring can be performed on a co-registered MR image and transferred to the planning CT. In other disease sites, such as head and neck cancers, target delineation can also be performed primarily on MRI with adaptation on CT, as long as the patient setup is reproduced accurately enough to allow rigid registration between the CT and MR or a dedicate deformable image registration tool is available.

The additional information from MRI can lead to a change in volume of the targets. In case of brain tumors, MRI led to larger volumes⁴⁰, whereas for the prostate bed reduced volumes with MRI were reported⁴¹. No difference in volume was reported when using MRI or CT for delineation for pharyngo-laryngeal tumors⁴², lung cancer⁴³ and the lumpectomy cavity of the breast⁴⁴. However, even in cases of non-significant differences in target volume, a significant difference can appear regarding the outline of the structures. Inter-observer comparisons have revealed higher similarity in contouring the lumpectomy cavity in breast cancer on MRI than on CT⁴⁴. However, higher inter-observer agreement was reported for CT than MRI and delineation of prostate bed⁴¹ and lung cancer⁴³. No difference was found in head and neck cancer⁴⁵ and brain tumors⁴⁰.

Another approach to use MRI for target volume definition is with artificial intelligence, like machine learning and deep learning⁴⁶, but these methods require further validation and should only be used as guidance for physicians in target delineation.

While so far only the standard MR sequences like T1-weighted, T2-weighted and T1-weighted post contrast imaging are routinely used for delineation, other promising sequences such as diffusion-weighted imaging (DWI) are under investigation. Delineation on DWI was shown to correlate better with microscopic tumor extent than the contours drawn on CT and standard MR sequences in laryngeal cancer patients⁴⁷. The addition of this sequence led to an adaption of the target volume in 73% of locally advanced cervical cancer patients⁴⁸ and to a larger volume in head and neck cancer patients compared to CT, PET and T2 or a combination⁴⁹.

Furthermore, many groups have been involved in the application of MRI radiomics features extracted from various MRI sequences and have explored using these features as surrogates of a non-invasive predictive tool to guide diagnosis, tumor phenotyping, and therapy outcome across different disease sites. MRI radiomics models have been investigated to differentiate between benign and malignant disease and tumor grading in central nervous system (CNS)⁵⁰, liver⁵¹, prostate^52,53, bladder⁵⁴, and across other organ sites^55,56. However, this method needs further validation before it can be applied clinically.

4. PET/CT and SPECT/CT Approaches

PET and SPECT are functional imaging techniques that are commonly leveraged in Radiation Oncology to differentiate between tumor and healthy tissues. They are typically combined with a simultaneous CT to produce images that contain both anatomical and functional information, whereas only few centers so far offer a combined PET/MR. PET/CT and SPECT/CT have been shown to improve lesion contrast and detectability and tumor staging accuracy compared to PET, SPECT, or CT alone^57–59, which is particularly useful for target volume definition when there is poor contrast between the tumor and surrounding tissues on CT or MR⁶⁰.

The most common PET radiotracer is 18F-fluorodeoxyglucose (18F-FDG), which is widely used for target delineation in many types of cancers, including head and neck cancer, lung cancer and lymphoma, and can lead to significant adaption of the treatment volumes⁶¹. For example, Prathipati et al. compared target volumes for non-small cell lung cancer generated based on contrast enhanced CT vs. PET/CT and found that the primary volume was increased and decreased for 39% and 50% of patients with the addition of PET, respectively⁶². Further, additional involved lymph nodes were identified for 27% of patients, which suggests that PET/CT allows for more accurate detection and targeting of malignant masses and better sparing of non-cancerous tissues.

Still, 18F-FDG is a poor radiotracer for non-glucose-avid tumors such as low-grade prostate cancers and thyroid cancers⁶³. In addition, there tends to be high uptake of 18F-FDG in benign lesions with increased glucose metabolism, such as acute inflammatory lesions and post-surgical scar tissue, resulting in a high rate of false positives with 18F-FDG PET/CT imaging⁶⁴. Certain organs such as the brain and heart also require large amounts of glucose to meet their high metabolic demands, so the uptake of 18F-FDG in these organs can mask nearby lesions.

To overcome some of the shortcomings of 18F-FDG, a host of other positron-emitting radiotracers have been developed and evaluated for detection of cancerous masses, including 11C-choline. Choline is an effective radiotracer for identifying locally recurrent disease after radical prostatectomy^65,66 and for detecting lymph node involvement⁶⁷ or distant metastases. The addition of PET compared to CT alone can have a major impact on target volume or prescription dose, which changed in 56% of patients, as described by Alongi et al.⁶⁸. Major drawbacks of choline, however, include the short physical half-life of 11C (20 minutes), which limits its use to institutions that have their own cyclotrons as well as a reduced detection accuracy for microcarcinomas and micrometastases.

An emerging alternative to 11C-choline for prostate cancer is prostate-specific membrane antigen (PSMA) radiolabeled with 68Ga. Although its half-life is roughly 3 times longer than that of 11C (68 minutes), it is still inconvenient for transport among facilities. Nonetheless, 68Ga can be produced using a generator composed of the parent isotope 68Ge, which can be maintained in a hospital and eluted to obtain 68Ga as needed⁶⁹.

Another emerging radiotracer is 18F-fluoro-ethyl-tyrosine (FET), which is used for brain cancer. Due to the brain’s high metabolism of glucose, 18F-FDG is taken up by healthy brain cells, often masking malignant tissues in the brain in conventional 18F-FDG-PET imaging⁷⁰. In a meta-analysis of 5 studies, 18F-FET-PET exhibited a pooled sensitivity and specificity of 0.94 and 0.88, respectively, compared to 0.38 and 0.86, respectively, for 18F-FDG-PET in patients with primary brain tumors⁷¹.

PET is considered a quantitative imaging modality, measuring the tracer activity in each pixel. Quantitative accuracy requires the raw voxel values to be corrected for photon attenuation in the patient, physical decay of the radionuclide, dead-time losses in the detector, and scatter and random coincidences⁶. The uptake of PET radiotracers can be quantified using the standardized uptake value (SUV), which characterizes the measured activity in a voxel as a fraction of the total administered activity⁷². However, SUV values can be affected by many factors including reconstruction algorithms, partial volume effects, radiotracer distribution times, accuracy of patient body mass, and activity of radiotracer leftover in the syringe after injection, and are at best “semi-quantitative” ^6,73.

The relative SUV of the tumor and nearby organs is used to differentiate malignant from healthy tissues, but visualization of the borders of high-uptake regions are largely dependent on the observer and the window/level settings used to view the PET/CT image⁷⁴. To address this inter-observer variability, many clinics employ auto-segmentation techniques based on threshold SUV values to define tumor boundaries. Even still, there are various thresholding methods used, including defining any voxels with SUV > 2.5 as tumor⁷⁵, isocontouring based on a given percentage of maximum tumor SUV⁷⁶, and thresholding based on the signal-to-background ratio⁷⁷, whereas the reference tissue can vary as well. Schinagl et al. compared contours generated by each of these methods to manually drawn contours based on visual interpretation of PET/CT for head and neck cancers⁷⁸. They observed significant variation in target volume and overlap fraction among all segmentation approaches. They also found that the SUV > 2.5 threshold, which is the most commonly used threshold value for lung cancer delineation, grossly overestimated the target volume in 45% of head and neck cases. This result was attributed to the relatively high uptake of 18F-FDG in the muscular tissue in the head and neck compared to the relatively low 18F-FDG uptake in healthy lung tissue. This example illustrates that thresholding methods that may work reasonably well for one disease site may not simply be extrapolated to other tumor sites without proper validation.

Similarly, caution should be exercised when extending the concept of SUV to radiopharmaceuticals other than 18F-FDG. Several studies have demonstrated a significant difference in SUV values for emerging radiopharmaceuticals compared to 18F-FDG^79,80, suggesting that SUV values cannot be reliably compared between different radiotracers due to differences in tracer metabolism.

Although auto-segmentation of tumor volumes based on SUV values can speed up the segmentation process and provide a less subjective definition of tumor volumes, all contours used for radiation therapy planning should be reviewed and modified as appropriate based on an experienced radiation oncologist’s clinical judgment.

In addition to defining the gross tumor volume for radiation planning purposes, PET/CT can be used to identify tumor sub-regions based on differential radiotracer uptake. In a process known as “dose painting,” different doses can be applied to sub-regions thought to be at a higher or lower risk of recurrence in order to increase the probability of long-term survival while decreasing the probability of radiation-induced side effects. A number of feasibility studies have demonstrated the potential of this treatment method in head and neck cancers^81–83, although further investigation and long-term follow-up are needed before dose painting can become the standard of care.

Figure 1:

Example of a patient with T2 tonsil cancer of the right side receiving intensity-modulated radiotherapy with 66 Gy. Left: Axial slice of planning CT with tumor (light green) and organs at risk (orange / cyan: parotid glands, light blue: part of the submandibular gland, yellow: spinal cord) delineated; middle: corresponding slice of PET/CT with highly FDG-active primary tumor (window level PET: 0.5 – 10 SUV); right: treatment plan with gross tumor volume (GTV = tumor) in green, clinical target volume (= GTV + subclinical disease) in lilac and planning target volume (= CTV + geometric expansion) in red. High-dose area highlighted in orange/red.

Figure 2:

Example of the difference in contouring by Radiation Oncologists of a left skull base metastasis on 60 keV DE-CT (left: axial slice, upper right: sagittal view, lower right: coronal view).

Figure 3:

Example of a patient with T4 tonsil cancer of the right side (tumor delineated in green). Left: Poor image contrast between tumor and surrounding tissue on CT; middle: Improved image contrast on T2-weigthed MRI; right: DWI MRI for the same patient

KEY POINTS.

• CT is still the current state-of-the-art imaging method for target delineation and treatment planning in Radiation Oncology

• 4D-CT, DCE-CT and DE-CT can be used to account for motion artifacts and to increase tissue contrast

• Functional MRI and PET are promising imaging techniques to improve accuracy of target delineation

SYNOPSIS.

Imaging in Radiation Oncology has a wide range of applications. It is absolutely necessary not only for tumor staging and treatment response assessment after therapy, but also for the treatment planning process, including definition of target and organs at risk, as well as treatment plan calculation. In this review, we provide a comprehensive overview of the main imaging modalities currently used for target delineation and treatment planning and give insight into new and promising techniques.