Imaging of Pediatric Head and Neck Tumors: A COG Diagnostic Imaging Committee/SPR Oncology Committee/ASPNR White Paper

Introduction

Head and neck cancer comprises approximately 12% of all childhood malignancies.¹ Common head and neck tumors include those which can occur in other regions of the body, such as neuroblastoma, rhabdomyosarcoma, lymphoma, Ewing Sarcoma and osteosarcoma, as well as those arising in tissues specific to the head and neck, such as salivary gland malignancies, nasopharyngeal carcinoma, and odontogenic neoplasms.¹ The presenting signs and symptoms of pediatric head and neck tumors are variable. However, most head and neck tumors come to attention due to the presence of a visible or palpable mass, or a functional deficit that varies by location, such as epistaxis, difficulty swallowing, and malocclusion.

While there are a variety of staging systems for head and neck tumors, common themes including size, location, and trans-spatial involvement, and nodal and distant metastatic disease are important across diagnoses. While staging may be tumor-specific, there is a growing trend to increase the uniformity of techniques used to assess outcomes and response to therapy. Both tumor staging and outcomes assessment require high spatial and contrast resolution, shaping imaging recommendations.

This manuscript was funded in part by the National Clinical Trials Network Operations Center Grant U10CA180886. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. The recommendations for imaging reflect available evidence and consensus expert opinion. Consensus recommendations are guided by the following principles: 1) Use of hardware and software available to the majority of pediatric cancer centers across the world; 2) Facilitation of existing staging and tumor response systems, such as RECIST 1.1²; 3) Compliance with existing guidelines for image appropriateness, as outlined by the American College of Radiology Appropriateness Criteria.

Imaging Modalities and Relevance to Head and Neck Tumors

While the general advantages and disadvantages of each imaging modality are summarized in Table 1, the discussion below highlights the attributes of each modality which carry implications for imaging structures of the head and neck.

Table 1:

Summary of Imaging Modalities

Procedure Name	Advantage(s)	Disadvantage(s)
MRI	Excellent soft-tissue contrast Excellent evaluation of bone marrow High spatial resolution is achievable Advanced sequences (DWI) with quantification No ionizing radiation	Cortical bone evaluation may be limited Motion artifact vulnerability due to longer acquisition times Sedation may be necessary, especially in children MR artifacts such as susceptibility (metal, air, etc.) Gadolinium deposition and NSF concerns Contraindications in the setting of certain implanted devices (cochlear implant, pacemaker, etc.)
CT	Excellent imaging of cortical bone and teeth (high spatial resolution) Identifies areas of calcification Allows for 3D imaging and creation of 3D printed bone-borne surgical guides Sedation typically not required	Suboptimal contrast resolution of soft-tissue tumors Ionizing radiation exposure20 Metallic artifact Risk of nephrotoxicity with iodinated intravenous contrast administration, primarily in patients with pre-existing renal insufficiency
US	Inexpensive No ionizing radiation No sedation or general anesthesia required Excellent ability to discern cyst vs. solid nodule Complementary to CT and MR in evaluating lymphadenopathy and invasion of structures Can guide interventional procedures and biopsies	Operator-dependent Limited reproducibility of measurements, making accurate comparisons between studies challenging Artifacts, mainly posterior acoustic shadowing Limited lesion localization and visualization of deep soft tissue structures
PET	Provides information regarding metabolic activity of tumors Detects metastatic disease Tumor response assessment Post-treatment change vs tumor recurrence	Low spatial resolution Ionizing radiation exposure Metabolic activity can be nonspecific, particularly within lymph nodes and tonsils (especially in children)

^*All of the modalities can serve a role in diagnosis, staging, follow-up and surveillance.

MRI

MRI exhibits excellent soft tissue contrast on account of the variable signal intensity of various tissues within the body due to different density and local environment of protons. This contrast is exploited in certain “anatomic” sequences, such as T1 and T2 weighted images, to yield high resolution detail. Contrast can be further enhanced by administration of intravenous gadolinium-based contrast agents, with the added benefit of distinguishing between cystic and solid lesions. The excellent contrast and spatial resolution of MR is particularly useful in the assessment of tumors arising within the soft tissues of the head and neck, where tumor location has important implications for diagnosis, staging, as well as treatment.³ A prime example of this involves detection of perineural spread of tumor, particularly when neural invasion spans multiple spaces, such as intracranial and extracranial compartments.^4,5 Here, high-resolution, contrast-enhanced, fat suppressed MRI, in addition to comprehensive knowledge of anatomy, is required to evaluate for perineural spread. In addition to anatomic depiction, MR sequences have been designed to assess tissue composition such as cellularity, vascularity, and the presence of fat, providing additional information of relevance to diagnosis and treatment response assessment³ (Table 2).

Table 2.

Sample MRI Protocol and Sequences

Sequence Category	Planes	Comment
Fluid sensitive T2 T2FS Inversion Recovery (STIR) T2W Dixon	2 of 3 planes (axial, coronal, sagittal )	Slice thickness = 5 mm maximum Fat suppressed T2FS or STIR sequence extremely important STIR is preferable in the setting of metal hardware or other sources of magnetic field inhomogeneity Non-fat sat T2 exhibits greater signal to noise ratio Dixon exhibits the most uniform fat suppression
T1 Pre-contrast Traditional T1 T1 FS Dixon (water-only)	axial, coronal	Slice thickness = 5 mm maximum Dixon water-only provides uniform fat suppression A T1 non-FS is highly recommended to look for the presence of fat and marrow involvement
T1 Post-contrast Traditional T1 T1 FS Dixon (water-only)	axial, coronal	Slice thickness = 5 mm maximum Fat suppressed images required A post-contrast sequence the same as pre-contrast may avoid pseudo-enhancement21 High-resolution needed for detection of perineural spread
Diffusion EPI DWI Non-EPI DWI	Axial	B-values 0 and 700-1000, interpret with ADC maps. Can help increase conspicuity of pathologic lymph nodes or denote cellularity of tumor May assist in detecting marrow lesions or infiltration EPI technique requires less time than non-EPI Non-EPI techniques reduce artifact related to aerated paranasal sinuses/mastoid air cells and motion.22,23
Perfusion DCE ASL	Axial	Optional May provide useful diagnostic and prognostic information24,25
Vascular 2D-TOF without Time-resolved MRA	Axial, coronal	Consider if vascular malformation is in the differential TOF can be performed without contrast Time-resolved MRA requires contrast

^*Sample sequences are provided. At least one sequence of each type should be considered, though individual selections may reflect the specific tumor and the local institutional capabilities.

^**FSE=Fast Spin Echo, TSE=Turbo Spin Echo, IR=Inversion recovery, DWI = Diffusion weighted imaging, ADC = Apparent diffusion coefficient, TOF = Time of flight, FS = fat suppression, DCE = dynamic contrast enhancement, ASL = arterial spin labeling.

^***Dixon fat suppression, while typically uniform, carries a processing artifact known as fat-water swap, though this can be corrected using post-processing techniques.

^****Considerations to reduce metal artifact: lower field strength (1.5T > 3T), increase bandwidth, increase the number of signal acquisitions (NEX), decreasing matrix size, FSE > SE and GRE, STIR > T2FS.

^*****Slice thickness maximum should be 5 mm but younger patients may need thinner slices.

^******Coverage may depend on patient size and available receiver coils. Head or neck coils should be used depending on patient size and tumor location. However, maxillofacial imaging typically extends from the superior orbital brow through the mandible and neck imaging extends from the skull base through the clavicles. The inclusion of a sequence extending to the mediastinum is optional. Vascular studies should extend inferiorly to the level of the aortic arch.

^*******Timing and ordering of sequences may vary. The optimal time to image post-gadolinium administration is uncertain for head and neck tumors,²⁶ but imaging is often performed between 5-15 min after administration, allowing time for the acquisition of other sequences such diffusion. Other sequences should be performed prior to the administration of gadolinium, such as non-contrast TOF.

CT

Computed tomography is a fast, relatively motion-resistant technique capable of obtaining high spatial resolution images of the head and neck. However, soft-tissue contrast is poor relative to MRI, limiting its use as a primary diagnostic or monitoring tool. Nevertheless, there are certain tumors which uniquely benefit from CT imaging, namely those involving the maxilla, mandible, and paranasal sinuses. The intrinsic contrast of the jaws, teeth, and sinus walls relative to soft tissues and air, combined with the high spatial resolution and short acquisition times of CT, produce high quality images of these maxillofacial osseous structures.⁶ The excellent depiction of calcium by CT allows for detection of tumor mineralization and periosteal reactions, which may aid in the diagnosis and response assessment of cartilaginous or ossifying tumors.

For tumors arising in or predominantly involving the maxilla or mandible, pre-operative 3D planning is made feasible by the acquisition of thin-section volumetric CT, with isotropic voxels, allowing for 3D reconstruction of the bones and some soft tissues. While this dataset is available on the scanner post-acquisition, it is not routinely sent to PACS and may be deleted from the scanner. Therefore, these reconstructions must be requested during the ordering or protocoling of the examination (Table 3).

Table 3:

Recommended CT Protocol: Reconstructions and Multiplanar Reformats

Plane	Reconstruction	Contrast phase	Required/ Optional	Comment
Axial, coronal, AND sagittal	Soft tissue and bone reconstruction kernels at 2.5-3 mm slice thickness	With contrast	Required	Perform with metal artifact reduction algorithm using single or dual energy techniques if metallic structures would cause significant artifact precluding adequate imaging.
Axial	Soft tissue and bone reconstruction kernels at native resolution (≤ 1 mm, isotropic voxels)	With contrast	Optional	Needed if planning to use for surgical planning and navigation systems Multiplanar reformats and 3D renderings can be created

^*The region scanned for a maxillofacial CT scan is from the frontal sinuses through the mandible. Coverage for a CT of the neck soft tissue should extend from the skull base through the T1 vertebral body. As the neck is contiguous with the chest through the thoracic inlet, some CT neck protocols extend inferiorly to include the aortic arch if there is concern for mediastinal involvement. This may be avoided if chest imaging is to be obtained separately.

^**Intravenous contrast may not be needed when CT is ordered for evaluation of bony structures to complement an MR exam obtained without and with contrast or for lesions entirely in bone or sinus cavities, where the outlines of the tumor can be clearly defined.

^***While CT data are acquired in the axial plane, images can be reformatted in the coronal and sagittal planes.

US

Ultrasound can be easily implemented without the need for sedation or general anesthesia and for this reason it is often the first imaging modality employed in the evaluation of a neck mass in a young child. This technique is well-suited for detection and characterization of a mass as either cystic, solid or both. Color and spectral Doppler techniques can also help in differentiating a cystic vs. solid nodule as well as distinguishing a solid tumor from a vascular malformation. Solid nodules can occasionally be characterized as specific structures when certain sonographic features are present, such as in the case of an echogenic fatty hilum of a lymph node. The excellent spatial resolution for shallow structures using a high-frequency linear transducer may allow for detection of invasion of structures, providing supplementary information to CT and MRI.^7,8

Positron Emission Tomography using Fluorodeoxyglucose (FDG PET)

FDG PET, often combined with low-resolution CT for the purposes of anatomic localization and attenuation correction, provides assessment of the metabolic activity of tumors throughout the body. As it relates to head and neck imaging, PET can be used to assess nodal involvement, detect distant metastatic disease, assess tumor response, and help distinguish post-treatment change from tumor recurrence.

Selection of Imaging Modalities for Head and Neck Tumors

Due to the smaller numbers and diverse group of pediatric head and neck tumors, no specific guidelines exist regarding the optimal imaging approach. However, imaging of pediatric head and neck tumors mimics that in adults and reflects the modality specific imaging attributes described above.

In the setting of a palpable mass of the head and neck, US should be first line to characterize the mass as cystic or solid and to determine if its sonographic features are characteristic of a benign finding (Grade D, SOR 1.69, Strong). If further characterization is required, or if a malignant diagnosis has been made by histopathological examination, cross-sectional imaging is needed.

MRI is the test of choice for imaging the primary soft tissue tumor of the head and neck at all timepoints (Grade D, SOR 1.31, Very Strong).

Contrast-enhanced CT is second-line for primary tumor imaging and is often required for the purposes of surgical planning (Grade D, SOR 1.77, Strong). Intravenous contrast may not be needed when CT is ordered for evaluation of bony structures to complement MRI or for lesions entirely in bone or sinus cavities, where the outlines of the tumor can be clearly defined.

CT provides complementary information to MR for predominantly intra-osseous lesions by demonstrating bony changes and tumor mineralization, though marrow involvement is best demonstrated by MRI (Grade D, SOR 1.31, Very Strong).

When tumors of the head and neck encroach upon the skull base, both CT and MR are required for complete evaluation (Grade D, SOR 1.69, Strong).

Intracranial extension of disease should be imaged with MRI (Grade D, SOR 1.07, Very Strong). CT or MR angiography may be employed for vascular mapping prior to surgery or to assess the potential for embolization of hypervascular tumors.

Local and regional disease can also be assessed with MRI, though CT neck with contrast is considered acceptable for the evaluation of cervical lymphadenopathy.⁹ As with other tumors, PET/CT is best for the evaluation of distant metastatic disease, but may also complement anatomic imaging to identify local and regional spread.¹⁰

Chest CT is required for the evaluation of pulmonary metastatic disease (Grade A, SOR 1.15, Very Strong)(Table 4).^11-13

Table 4:

Image-based Staging Recommendations

Site of Disease	Imaging Modality Recommendation	Comment
Primary	MR WO and W contrast	CT with contrast may be complimentary to MR for intra-osseous tumors and may be needed to assess sinus involvement
Local and regional spread	MR WO and W contrast Neck CT W contrast AND PET/CT	MR or CT acceptable for evaluation of cervical lymphadenopathy PET/CT is complementary
Metastasis	PET/CT CT chest W	Initial chest CT w contrast. If only pulmonary nodules are found, this can be followed with CT chest WO contrast

^*US may provide complimentary information regarding tumor extent and nodal disease. Non-contrast MRI is preferred to CT when contrast agents are unable to be administered. PET/CT is the primary modality of lymph node evaluation in the setting of lymphoma.

The initial chest CT should be obtained with contrast to evaluate for mediastinal and hilar lymphadenopathy. If subsequent imaging of the chest is performed solely for detection of pulmonary nodules, non-contrast chest CT is sufficient, particularly during follow-up and surveillance with prior negative chest exams (Grade D, SOR 1.46, Very Strong).

The same modalities and parameters described above and in tables are used for follow-up and/or surveillance, with the type and frequency of exams varying based on initial staging as well as surgical and histopathologic findings. RECIST 1.1 is a commonly used imaging response criteria on follow-up imaging, whereby single linear measurements are compared over time.² Due to the varying orientations of tumors relative to traditional orthogonal imaging planes, a multiplanar approach is preferred to allow for meaningful and reproducible measurements. Follow-up and surveillance imaging may reveal sequelae of treatment, including post-surgical and radiation effects, which may complicate image interpretation. Here, MR sequences designed to assess tissue composition can aid in interpretation and are preferred over CT. For predominantly intra-osseous tumors, RECIST 1.1 does not apply. No standard response criteria have been validated for these head and neck tumors.

Challenges

It is common for metal to be within the vicinity of an imaged head and neck structure in the pediatric population. Sources of metal include dental restorations, orthodontic brackets, and surgical hardware (Table 1). These sources of metal are known to cause artifact on both CT and MR examinations. In MRI, the safety of these metallic objects must also be carefully considered. When possible, orthodontic hardware should be removed. For other sources of metal, various techniques have been developed for both CT and MRI to reduce metal artifact.^14,15

Motion artifact is particularly problematic with MRI, due to the longer image acquisition times compared to CT, as well as the physiologic motion due to swallowing. Access to anesthesiology services for sedation or general anesthesia may also limit the timing and scheduling of an exam. Patients with complex medical needs may require subspecialized care to obtain high quality images. Non-pharmacologic strategies have been employed to reduce motion and improve scan quality, including sleep-deprived MR, mock scans, and distraction techniques.^16,17 However, in the majority of pediatric patients under the age of 10 years, either moderate sedation or general anesthesia is required which carries inherent risks.¹⁸

Even when motion and metal artifact are minimized, image quality on all modalities is often limited by the spatial and contrast resolution. This is particularly true in the head and neck, where the anatomy is remarkably complex. There are several layers of deep and superficial fascia which separate the tissues of the head and neck into spaces. These spaces, while geographically near one another, are topologically distinct. Consequently, a relatively small area may contain several different tissue types; bone, teeth, nerve, muscle, mucosa, lymph, and glandular tissue can be found in a very small patch of anatomical space. This density of distinct anatomical spaces pushes the limits of modern imaging modalities.

Future Advancements

In addition to the general efforts to reduce artifact and improve image quality, there are novel advancements with significant potential to impact the management of pediatric patients with head and neck cancer. Quantitative MR imaging, including ADC, T1, and T2 mapping, may provide greater diagnostic and prognostic information. MR spectroscopy has the potential to evaluate metabolic changes in tumors. Various techniques utilizing artificial intelligence may also identify clinically relevant medical imaging data invisible to the human eye. Dual energy CT, particularly with photon-counting detectors, stands to drastically improve the soft tissue resolution of CT. Contrast-enhanced ultrasound and elastography provide potential imaging biomarkers of diagnosis and treatment response.¹⁹

Summary

While the goal of all imaging is to maximize spatial and contrast resolution and minimize noise, this is of utmost importance in the imaging of pediatric head and neck tumors. The spaces of the head and neck are complex, and a small volume of anatomy may contain several tissue types separated by thin anatomic boundaries. Violation of these boundaries by tumors is important in staging and surgical planning. The difficulty in imaging this intricate anatomy is compounded by the smaller size of pediatric patients as well as the presence of motion and metal artifacts in some cases. The imaging recommendations reflect the need for high-resolution imaging and reduction of image artifacts and noise. MR and CT protocols must be tailored to the patient, institutional capabilities, and specific disease processes. Future advancements are expected within the next 5-10 years and when possible, should be incorporated into pediatric oncologic imaging protocols when possible.

Abbreviation

CT

Computed Tomography

MRI

Magnetic Resonance Imaging

US

Ultrasound

PET

Positron Emission Tomography

FDG

Fluorodeoxyglucose