Functional magnetic resonance imaging of head and neck cancer: Performance and potential

Abstract

Functional magnetic resonance imaging (MRI) of tumors of the head and neck usually encompasses diffusion-weighted imaging (DWI) and intravenous (IV) contrast T1 dynamic perfusion imaging (DCE-MRI or PWI). Both techniques can characterize different tissues by probing into their microstructure, providing a novel approach in oncological imaging. In this pictorial review, we will cover the important technical aspects of DWI and PWI, the pathophysiological background and the current applications and potential of these functional MRI techniques in the imaging of head and neck cancer.

Introduction

Functional magnetic resonance imaging (MRI) of tumors of the head and neck includes diffusion-weighted imaging (DWI) and intravenous (IV) contrast T1 dynamic perfusion imaging (dynamic contrast-enhanced (DCE)-MRI or PWI). The implementation of these techniques is currently limited to a small number of head and neck imaging academic practices. The cost of software packages for performing perfusion analysis, and the time and expertise required for this prohibits the widespread use of the technique in community/private practices. However, DWI requires little time and is incorporated into the basic manufacturers' packages, so that it is more widely used. This article provides an overview of both these techniques and their potential benefits as problem-solving tools in head and neck cancer imaging.

DWI: important technical aspects

The basic principles of DWI have been extensively covered in the literature.¹ Clinical applications for DWI MRI in the head and neck region have been limited because echoplanar diffusion-weighted MRI in this region has several inherent drawbacks, including susceptibility of artifacts due to the anatomic heterogeneity of the area, the presence of dental work, as well as adjacent air and bone.² Important technical considerations in applying DWI include inferior saturation block to minimize pulsation artifacts, adding a posterior saturation band to eliminate the susceptibility from the posterior neck soft tissues–fat interface, manual shimming to avoid air containing or moving parts (improving homogeneity) and reducing fat shift and distortion artifacts. A shim should cover the spine and muscles of the neck while avoiding large areas of air. The phase encoding direction should be anterior to posterior (A-P) to minimize distortion. Moreover, imaging is best carried out in two groups in the upper (face) and lower neck with two different shimming blocks to compensate for the shape and thickness of different tissue parts (Figure 1).

Figure 1.

Important technical aspects in DWI: Saturation block to minimize pulsation artifacts (white arrow), adding a posterior saturation band to eliminate susceptibility from air and bone posterior to neck soft tissues (black arrow) and imaging carried out in two groups in the upper (face; black arrowheads) and lower neck (white arrowheads), with two different shimming blocks to compensate for different tissue part thickness.

DWI: diffusion-weighted imaging.

An echoplanar imaging EPI-DWI is applied, with an apparent diffusion coefficient (ADC) calculation usually carried out using a mono-exponential model [S = S0 × exp (−b × ADC)], where signal intensity (SI) is measured with at least two different b-values (s/mm²). For instance, 0, 50, 100, 500,750, and 1000 b-values (s/mm²) are plotted for estimated SI (ADC average; Figure 2). The influence of b-value choice on the calculated ADC has been studied by Thoeny et al.³ and Vandecaveye et al.⁴ Using a large range of b-values (0, 50,…,1000) leads to different ADCs with low (0, 50, 100) and high (500, 750, 1000) b-value settings providing different information, and reflects contributions from diffusion and perfusion. Vandecaveye et al.⁴ showed that the choice of different sets of b-values to calculate ADC can result in different interpretation of the data. Using ADC b0–1000 showed a sensitivity (SN) of 83% and specificity (SP) 94%, ADC b0–100 showed SN 79% and SP 67%, probably due to contamination by micro perfusion, and ADC b500–1000 showed SN 72% and SP 67% due to the influence of noise. Care should be taken in placing the region of interest (ROI) measurement in the tumor, to include the solid portions and exclude necrotic or cystic regions in reference to T2-weighted and gadolinium-enhanced T1-weighted MRI. A crucial consideration is that imaging should be done ahead of biopsy, not only to avoid the induced edema which will affect the measured ADC values, but also because of the risk of hemorrhage (iron content) inducing a local susceptibility artifact, thereby destroying the signal in this area.

Figure 2.

ADC calculation according to the mono-exponential curve model. The signal intensity measured at different b-values are plotted against the respective b-values (0, 50,100, 500, 750 and 1000 s/mm²) and used to calculate the ADC value.

ADC: apparent diffusion coefficient.

DCE T1-weighted MRI: background and important technical considerations

This technique provides a non-invasive assessment of microcirculatory characteristics of lesions by quantifying the time course of contrast enhancement of DCE-MRI.⁵

An IV injection of gadolinium-based contrast medium (gadopentate dimeglumine sodium) at a rate of 3 ml /s, and a dose of 0.1 mmol/kg (up to 20 ml contrast medium) is usually delivered by power injection through 18–20-gauge needle, followed by a 20 ml post-contrast saline flush. A fast 3D T1-weighted gradient-echo SPGR (e.g. Vibe, FLASH, LAVA) MRI axial scan of 3–4 mm thickness corresponding to the same locations and thickness of axial T1 and T2-weighted anatomical images is carried out, with timings adjusted to acquire four pre-contrast runs followed by sequential post-contrast runs. For semiquantitative analysis, a temporal resolution of about 8–10 s through the area of interest in the head and neck is sufficient, whereas for model-based quantitative analysis (Tofts) a higher temporal resolution of 3–4 s is essential. Depending on the amount of wash-out to be visualized, the scan often takes up to 300–500 s (usually with 30 to 40 scan acquisitions; time points) and the curve acquired can then be analyzed.^5,6

DCE-MRI curve semi quantitative analysis

DCE-MRI is based on rapid diffusion of a low molecular weight diffusible MRI contrast medium (CM), a gadolinium chelate-based agent (Gd-DTPA), between the intra and extravascular compartments, resulting in a T1 relaxivity SI change (C) over time (t), (C/t). The measured T1 SI relates to the vascular density (S) within the lesion and is affected by regional blood flow (F) and permeability (P). The rate of enhancement relates to vascular permeability and leakiness (interstitial environment diffusibility and temporary retention of CM).⁷ Similar to DWI, DCE-MRI should be carried out before biopsy to avoid induced alterations in the tissue environment.

Data obtained directly from a time–signal intensity curve measure the relative SI (RSI) post-contrast relative to pre-contrast (SI post − SI pre/SI pre), which is plotted against time (t) to get an RSI/t curve. The maximum SI (C-Peak) measures the degree of enhancement, whereas time to peak (TTP) and initial slope (IS) reflects the rate of enhancement (Figure 3). Both the SI maximum and the slope of enhancement shows a significant correlation with micro-vessel density and median tumor partial pressure of oxygen. In other words, the DCE curve correlates with angiogenesis and oxygenation.^7–9

Figure 3.

Post-IV contrast dynamic enhancement curve: The height between yellow dashed lines along the vertical axis represents the contrast SI peak (C-peak), the distance between red dashed lines along horizontal axis represents TTP, the steepness of the contrast enhancement in blue represents IS and the wash-out of CM along the curve down slope in green represents the outflow.

IS: initial contrast uptake slope; IV: intravenous; SI: signal intensity; TTP: time to peak.

Although the flow (F), permeability (P), and vascular density (S) components cannot be separated by mathematical modeling, their combined effect is reflected in the transfer constant (K^trans) characterizing the influx of CM from the plasma into the tissue extracellular space. Additional tissue-specific parameters may be estimated by DCE including: the volume fraction of the extravascular extracellular space (v_e), the volume fraction of plasma in tissue (v_p), and the rate constant for efflux of CM back into plasma from the tissue extracellular space (k_ep).¹⁰

In our experience, the majority of the malignant lesions had a post-contrast enhancement peak near to or >50% of the arterial input function peak.

DCE-MRI quantitative analysis

Unlike computed tomography (CT) perfusion, Gd concentration does not have a linear relation with the SI. This is because the effect of the contrast agent is based on its T1 relaxation, decreasing the effect on surrounding water protons. In contrast with semiquantitative analysis, quantitative metrics allow true physiological parameters, such as K^trans and the initial area under the Gaussian curve at 60 or 90 s (IAUGC 60 or 90) to be assessed. However, with DCE-MRI, blood flow (BF) and blood volume (BV) are very rarely used, as the only way for BV not to be overestimated is if the CM does not leave the intravascular compartment. DCE-MRI quantitative analysis allows a common platform for DCE-MRI results from different centers. Moreover, quantitative parametric mapping can be potentially applied for different sub-volumes within the lesion.

The simplest quantitative reproducible measurement is the IAUGC 60 or 90 which is the initial area under a c-t curve calculated up to a given time of 60 or 90 s. This has been recommended by the United States National Cancer Institute as one of the valid physiological endpoint measurements, which combines information on the rate and extent of enhancement⁶ (Figure 4).

Figure 4.

Different ROIs (shown as circles) from different parts of tumor in top row. The bottom row shows the dynamic contrast enhancement curve with SI against time. Quantitative analysis (applying 4D software calculation, Siemens) showing linear relation of SI time curve, with different contrast–time curves from different ROIs of tumor different areas/sub-volumes (arrows).

ROI: region of interest; SI: signal intensity.

To apply quantitative analysis (Tissue 4D, Siemens) calculations, additional T1 measurements should be obtained. Since the effect of the contrast enhancement signal change is a property of T1 relaxation, a voxel-wise tissue (T10) is calculated prior to DCE runs by adding a T1-weighted SPGR-type sequence with different flip angles to calculate the T1 map (we usually apply the same VIBE sequence as the DCE, with three different flip angles 5, 10 and 15 before the contrast acquisition, each taking about 7–10 s). Alternatively, a T1W-inversion-recovery prepared turbo fast low-angle shot 3D sequence with several different inversion times (0.06, 0.2, 0.4, 0.8, and 1.6 s) can be applied. The T1 value from this series are used to quantify the T1map voxel-wise tissue (T10).¹⁰

A pharmacokinetic model (Tofts model) is used to fit the enhancement curves that enable the calculation of physiological parameters. The Tofts model considers the extravascular extracellular space (EES) tissue (t) and plasma (p) as two compartments individually well mixed with contrast agent concentrations (C). Permeability from Ct to Cp and are described by transfer rate constants K^trans and k_ep, respectively (Figure 5).¹¹

Figure 5.

Diagram for illustration of the Toft's two compartment model: The vascular and tissue compartments, their related physiological parameters and post-IV contrast injection exchange controlled by the transfer and rate constants K^trans and k_ep respectively (arrows) are shown.

IV: intravenous.

Tissue C_t is obtained from a ROI (SI) placed on the tumor and plasma C_p (arterial input function) is obtained from the ROI (SI) on a major vessel; the transport between these two compartments is determined by vascular permeability constants K^trans and k_ep. K^trans is the volume transfer constant between plasma and EES and from EES to plasma respectively.

Another parameter is the EES fractional volume (v_e) related to the two other parameters by the equation$v_e=K^{\mathit{trans}}/k_{\mathit{ep}}$.

If the contrast uptake is flow-limited, K^trans indicates the tissue perfusion/unit volume. If the contrast uptake is permeability-limited, then K^trans indicates the permeability surface area per unit volume of tissue. In real life, K^trans reflects both tissue perfusion and permeability. The transfer constant k_ep correlates with micro-vessel density and both K^trans and k_ep correlate with vascular endothelial growth factor (VEGF).^6–9

DCE curve changes during successful therapy

Early on during successful chemoradiotherapy (CRT; i.e. initial response), there is a decrease in the number of fast leaky components of the vascular wall resulting in a slower rise of the CM enhancement curve (a decrease in the rate of enhancement). With continued successful response to therapy (i.e. continued response), there is a regression of tumor angiogenic properties resulting in a decreased contrast peak. Simultaneously, fibrosis and other therapeutic effects also occur, resulting in re-diffusion with slower and gradual trapping (wash-in or accumulation pattern of enhancement; Figure 6). Functional changes in the vascular fenestration due to treatment effects leads to changes in the time–intensity curve from a rapid rise and wash-out in the pretreatment state to a slow rise and wash-in post-treatment state.^11–14

Figure 6.

Comparison of dynamic enhancement curve: (a) A tumor before therapy shows characteristic early wash-in and wash-out. (b) During successful therapy early response, the tumor shows a decrease in the number of fast leaky components of the vascular wall with a resultant slower rise of CM enhancement. (c) During continued successful therapy with continued response there is regression of tumor angiogenic properties with resultant decrease in C-peak; and fibrosis and other therapeutic effects resulting in slower re-diffusion with trapping (wash-in).

CM: contrast medium.

Current and potential applications of functional MRI in head and neck cancer imaging

Functional MRI in the evaluation of head and neck cancer can be discussed under the subheadings of seven applications given in Table 1.

Table 1.

Applications of functional magnetic resonance imaging in the evaluation of head and neck cancer.

1.	Resolving controversial scenarios in the diagnosis of primary cancer.
2.	More precise gross tumor volume delineation: Differentiation of primary tumor from associated inflammatory changes.
3.	Staging issues: Bone/cartilage invasion, vascular and prevertebral muscle involvement.
4.	Pretreatment prediction of response.
5.	Intra-treatment response assessment.
6.	Early (2 weeks) post-Rx: Response assessment for salvage management; specifically of locally advanced disease T3/4 tumor.
7.	Differentiation of early post-therapy changes from Rx failure: Residual versus post-therapy non-tumorous inflammatory changes.

1. Resolving controversial scenarios in the diagnosis of primary cancer

DWI and PWI can discriminate malignant from non-malignant lesions because of the different tissue composition and neoangiogenic properties. By applying dynamic post-contrast CT perfusion (CTP), head and neck cancer showed an increased BF, BV, and decreased mean transit time (MTT) compared with benign lesions and normal tissue.¹⁵ In CTP, MTT was found to best discriminate malignant from non-malignant tissue, with an average MTT of 3.5 s for malignant versus an MTT of 5.5 s for non-malignant tissue.^16,17 Similarly, DWI showed more restricted diffusion in tumors as compared with normal tissue and reactive inflammatory tissue.¹⁸

Upper aerodigestive tract inflammatory lesions are not infrequently encountered, and in some instances can pose a diagnostic dilemma to differentiate from neoplasia, with the final decision being made at pathological examination. However, functional MRI of the head and neck in our experience was able to confidently discriminate such cases (Figure 7).

Figure 7.

A 63-year-old female patient with known breast cancer under treatment, presented with a recent complaint of increasing dysphagia, dysphonia and no fever. On endoscopy, bulge and edema suggested a posterior hypopharyngeal mass lesion. Axial CT scan (a) shows a thickening and enhancement of the right pyriform sinus and right sided posterior hypopharyngeal wall. Axial T2W-MRI (b) shows corresponding bright signal intensity, DWI at b0 (c) shows bright signal intensity, b1000 (d) shows a faint increased residual signal abnormality and ADC map (e) shows high ADC signifying facilitated rather than restricted diffusion. DCE-MRI curve (f) from the lesion (dashed line) shows early peak with steep slope. However, the contrast enhancement curve peak remains well below 50% of AIF curve peak (AIF curve denoted by solid line), with gradual pooling rather than wash-out, signifying an inflammatory lesion. In our experience, the majority of the malignant lesions have post-contrast enhancement peak near to or >50% of the arterial input function peak.

The lesion subsequently resolved on steroids and antibiotics.

ADC: apparent diffusion coefficient; AIF: arterial input function; CT: computed tomography; DCE: dynamic contrast-enhanced; DWI: diffusion-weighted imaging; MRI: magnetic resonance imaging.

Changes associated with malignant lesions such as submucosal fibrosis of the buccal mucosa, often induced by chronic tobacco chewing, remains a big challenge to differentiate both clinically and with conventional imaging. In such case scenarios, the extent of the lesion and the need for surgical intervention is in question. In our experience, functional MRI of the head and neck has shown the potential in differentiating malignancy from adjacent submucosal fibrosis (Figure 8).

Figure 8.

A middle-aged Indian male with a history of chronic tobacco chewing, complaining of trismus. On examination, a palatal ulcerative mass lesion and an indurated buccal mucosa lesion extending to the retromolar region suspicious of squamous cell carcinoma were detected (not shown). Axial post-contrast fat-suppressed T1WI (a), and T2WI (b) shows thickening with enhancement and bright SI respectively. A DCE-MRI (c) curve shows steep enhancement, with contrast enhancement peak at about 50% of the AIF peak. However, there is subsequent continuous pooling as shown on sequential 9 s dynamic contrast scans (d), visual inspection of 5^th,10^th, 20^th, and 30^th runs (arrows) and/or ROI SI measurement (not shown) shows no change over time signifying trapping of contrast. DWI (e) from left to right hand side, at low b-value (b0) shows corresponding high signal, no residual signal on DWI high b-value (b1000), and high ADC value of 1.4 (×10⁻³ mm²/s) on the ADC map. The functional MRI features are consistent with inflammatory rather than neoplastic lesion. Pathology was consistent with submucosal fibrosis of the buccal mucosa associated with the palatal squamous cell carcinoma.

ADC: apparent diffusion coefficient; AIF: arterial input function; CT: computed tomography; DCE: dynamic contrast-enhanced; DWI: diffusion-weighted imaging; MRI: magnetic resonance imaging; ROI: region of interest; SI: signal intensity.

2. Extent of local disease: Gross tumor volume delineation

Gross tumor volume (GTV) denotes the tumor visible on clinical examination or delineated by radiological evaluation. Ill definition of tumor borders because of tumor-associated inflammatory changes and edema on morphological images is a well-recognized problem for GTV assessment. For example, the sole utilization of CT simulation scans in contouring of GTV is subject to a large degree of interobserver variability (53%), and co-registration of MRI and planning CTs substantially improved the interobserver delineation of target volume based on better delineation on MRI.^19,20 The addition of functional MRI for tumor assessment can aid in more accurate GTV delineation by excluding the surrounding reactive inflammatory changes (Figure 9).

Figure 9.

Added precision of functional MRI in GTV delineation in a case of carcinoma of right side of the tongue base, minimally extending to the posterior aspect of oral tongue. Routine morphologic spin echoes T1WI(a) and T2WI (b) show ill definition of the mass lesion (arrows), due to the surrounding peritumoral inflammatory changes and edema. DWI (c) from left to right hand side of image b0, b1000 and ADC map respectively, shows a smaller but better demarcated lesion (arrows), especially on the high b-value and the ADC map. Selected images from sequential post-contrast DCE-MRI (d) from left to right side of image, at early arterial, late arterial, venous and equilibrium phases show best delineation of lesion at the wash-in at late arterial phase. Note that the lesion appears to be mildly smaller on ADC than on DCE-MRI, probably due to increased perfusion related to tumor vascularity exceeding the restricted diffusion related to tumor cellularity.

ADC: apparent diffusion coefficient; CT: computed tomography; DCE: dynamic contrast-enhanced; DWI: diffusion-weighted imaging; GTV: gross tumor volume; MRI: magnetic resonance imaging.

Moreover, subtleness of signal abnormality on morphologic images is commonly encountered with squamous cell carcinoma in such anatomical sites such as the larynx and oral cavity, with poor delineation that poses difficulty in staging. Functional MRI (DWI, DCE-MRI) can better delineate the tumor volume at these sites (Figure 10).

Figure 10.

Carcinoma of the larynx (T1a). The abnormality was not delineated on morphologic images (not shown). An ADC image of a DWI series (a) delineates the lesion within the free-hand ROI of the left vocal cord, showing a mild diffusion restriction with ADC measuring 0.99 (×10⁻³ mm²/s). DCE-MRI (b) axial image (left) and dynamic curve from the lesion ROI (right), shows steep curve (red solid line) which initially parallels the AIF, however the contrast enhancement peak is <50% of the AIF peak (yellow dotted line) with a continuous build up, features which may be related to micronecrosis. Post-treatment ADC image (c) shows no diffusion restriction with interval increase in ADC value of 1.35 (×10⁻³ mm²/s). Post-treatment DCE-MRI (d) curve is now more typical of a successful post-therapy low peak and continuous pooling denoting good locoregional control with no residual DWI or DCE-MRI abnormality (note that the yellow dotted line curve corresponds to the lesion ROI and red solid line to AIF).

ADC: apparent diffusion coefficient; AIF: arterial input function; CT: computed tomography; DCE: dynamic contrast-enhanced; DWI: diffusion-weighted imaging; MRI: magnetic resonance imaging; ROI: region of interest; SI: signal intensity.

3. Staging issues

In the experience of the authors, decision making for certain staging issues can be further strengthened by evaluating the functional MRI images. Examples for this are laryngeal cartilage and bone (cortical and marrow) invasion assessment (Figure 11).

Figure 11.

T4 squamous cell carcinoma of the larynx 6 months post-CRT and tracheostomy. Endoscopy was suspicious for recurrent/residual mass at the anterior 1/3 of the vocal cord and a recent CT scan (not shown) was suspicious for recurrence with laryngeal cartilage involvement. MRI morphologic imaging axial T1WI (a), T2-fat-suppressed (b) and post-IV Gd fat-suppressed T1WI (c) shows soft tissue fullness extending to the anterior commissure. There is an associated abnormal intermediate signal intensity in the anterior 1/3 of left laryngeal cartilage on T1, bright signal intensity on T2 and enhancement on post-contrast images, concerning for probable laryngeal cartilage invasion (arrows in a, b, and c). DWI images (d) from left to right hand side at b0, b1000 and ADC map shows that the lesion has a bright signal on b0 and b1000 with T2 shine-through on an ADC map image, consistent with PTNT inflammatory tissue next to the inner cortex of thyroid cartilage (arrows). DCE-MRI post CM perfusion (e) 10 s coverage (time point) sequential phases; 7^th , 12^th , 30^th phases during early arterial, venous, and tissue equilibrium phases respectively showing gradual low peak enhancement of PTNT next to the inner cortex of thyroid cartilage; not extending into the cartilage. The patient was disease-free on clinical and MRI follow up after 1 year. CRT: chemo-radiotherapy; ADC: apparent diffusion coefficient; PTNT: post therapy non-tumorous; DCE: dynamic contrast enhanced.

There is also a potential that DCE-MRI color maps, fused with morphologic T2/T1WI would be more specific in differentiating between stages T3/T4a/T4b. This co-registration could improve the diagnosis of bone cortex and marrow invasion, vascular involvement and extension into surrounding soft tissue structures such as the prevertebral muscles (Figure 12).

Figure 12.

4D-map of initial area under the curve at 90 s (IAUC 90); grey shade intensity (a) and color map (b), overlaid on T1 morphologic axial image shows recurrent hypopharyngeal and laryngeal CA (circled ROI in a and b), with laryngeal cartilage invasion (arrowhead in b), extension into extra-laryngeal soft tissue encircling more than 50% of the right common carotid artery circumference (arrow in b), and flattening of and inseparability from the medial aspect of the right prevertebral muscle (double arrowheads in b) suggestive of a possible invasion (proved at surgery, resulting in incomplete resection). CA: cancer.

4. Prediction of tumor response by pretreatment tissue probing

Pretreatment tumor oxygenation is prognostic of locoregional tumor control after radiation therapy (RT) in advanced head and neck squamous cell carcinomas.²¹ The 2-year control rate of head and neck cancer treated with radiotherapy correlates with the initial pimonidazole-binding level of the tumor, where high pimonidazole-binding (hypoxic) tumors had significantly lower control rates of 48% compared with 87% for tumors with low pimonidazole-binding.²² DCE metrics such as K^trans, k_ep, initial area under the curve at 90 s (IAUC90), BV, and BF showed a significant correlation with the uptake of pimonidazole in tumors. This denotes that tumor vasculature is a surrogate marker for oxygenation state, and accordingly can be used to predict post-treatment outcomes in head and neck cancer.^22–24 Agarwal et al.²⁵ demonstrated that BF and BV were significantly higher in patients with a high T-stage (T3, T4) head and neck cancer and showed a better response to CRT.

Using pretreatment CTP, Truong et al.¹¹ showed that pretreatment tumor BF was significantly higher in patients who achieved locoregional control (LRC), 118.0 ml/100 g/min, compared with those with locoregional failure (LRF), 53.4 ml/100 g/min.

Agarwal et al.,²⁵ Kim et al.,²⁶ and Chawla et al.²⁷ studied the role of DCE-MRI metrics as predictors of response in head and neck cancers to CRT and showed that head and neck cancer with higher pretreatment BV, and pretreatment K^trans correlated with better response to CRT and disease-free survival.

Similarly, DWI can be used as a pretreatment marker for tumor hypoxia. Pretreatment head and neck cancers with necrosis due to tumor hypoxia and decreased vascularity have higher ADC values, and are less responsive to chemotherapy (CT) and RT. On the other hand, pretreatment head and neck cancers with higher BF and vascular permeability have better oxygenation with better access to CT and better response to RT.²⁸

Kim et al.²⁹ studied the efficacy of pretreatment DWI and early intra-treatment DWI changes in predicting the response of locally advanced head and neck squamous cell carcinoma to concurrent CRT and showed that DWI of head and neck cancer before treatment can predict the response to CRT. In their study, head and neck cancers with complete response showed pretreatment ADC values (ADC = 1.04 ± 0.19 × 10⁻³ mm²/s) which was significantly lower than head and neck cancers with a partial response (ADC = 1.35 ± 0.30 × 10⁻³ mm²/s).

Kato et al.³⁰ in their cohort of patients with head and neck squamous cell carcinoma, evaluated DWI MRI for the prediction of response to neoadjuvant therapy. They found that the tumors that responded well to neoadjuvant therapy tended to have lower ADC values, which they attributed to higher cellularity and higher cellular turnover, and the quantitative and qualitative signal intensities of the lesions on DWI showed positive correlations (r = 0.367 and 0.412, respectively, p < .05) with the tumor regression rates.

5. Initial/intra-treatment response prediction

CRT leads to cell death and reduction of the restrictive barriers, resulting in increase in mean ADC values early in the course of treatment. Early change of ADC after CRT for treatment of squamous cell carcinoma of the head and neck was studied by different authors. The change in ADC at 1 week of CRT was studied by Kim et al. and showed 86% sensitivity and 83% specificity for prediction of treatment response.²⁹ The study at 2 weeks after induction CT by Berrak et al.³¹ showed that a 22% increase in ADC compared with pretreatment ADC was associated with better post-therapy response and long-term survival.

According to Vandecaveye et al., the change in DWI at 2 and 4 weeks during concomitant CRT (% increase in ADC (ΔADC)) showed substantially greater accuracy than volumetric assessment for differentiation between responding and non-responding lesions at 2 years post-therapy. The ΔADC at 2 weeks was significantly lower for lesions with post-CRT recurrence than with complete remission (in %; primary tumor: 1.6 ± 9.7 versus 40.5 ± 25.6; and adenopathies: 10.9 ± 26.0 versus 52.1 ± 36.5, both p < 0.001).³²

Early response to therapy also results in decrease in vascular permeability of the neoangiogenic vessels, leading to decrease in the rate of enhancement. Powel et al. evaluated DCE parameters at 1 week after induction chemotherapy. In their cohort of patients, they compared the change of K^trans and the initial area under the curve at 60 s (IAUC60) between non-responders and responders, with the responders group showing a significant decrease of K^trans and IAUC60 post-treatment compared with pretreatment, falling by approximately 50%.³³

6. Prediction of response after completion of treatment (continued successful therapy)

DCE-MRI metric changes after the completion of radiotherapy reflect a reduction of vascular density, with a decrease in the overall SI and repair of treated tissue with changes in the vascular fenestration resulting in CM trapping.¹⁸ Knopp et al. applied DC-MRI at around 8 weeks after completion of CT/RT to predict the outcome and was successful in identifying patients at risk of local recurrence as candidates for dose escalation or salvage surgery. In their cohort of patients, those with local recurrence showed a high initial peak CP value and a short TTP (steep slope on the t–SI curve), whereas patients with local control showed a slow rise and a continuous wash-in accumulation pattern (Figure 13).¹³

Figure 13.

Glottic cancer larynx (T2N0), after completion of CT/RT showing continued successful therapy pattern at 8 weeks. Images from series 8, 15, 30 of DCE- MRI post I.V. contrast (a) shows low initial peak and continuous pooling with gradual increased SI on visual inspection of sequential series (arrows), as well as by reviewing the dynamic curve (b) which shows a low peak and continuous accumulation pattern. DWI images (c) b0, b1000 and ADC map images from left to right hand side (b0 = 202 arbitrary unit (absolute value) a.u., b1000 = 29 a.u., ADC = 1.8 (×10⁻³ mm²/s), shows remarkable facilitated diffusion of a successfully continued post-therapy response. CT/RT: chemotherapy and radiotherapy; DCE: dynamic contrast enhanced; SI: signal intensity; DWI: diffusion weighted images; ADC: apparent diffusion coefficient.

Similarly, an early DWI 3–4 weeks after completion of CT/RT can predict long-term outcomes (Figure 13). Vandecaveye et al., examined the percentage change of ADC at 3 weeks post-treatment between responding and non-responding lymph nodes in correlation with the outcome after 2 years of follow up. It was shown that there was a significant increase in the percentage change of ADC % at 3 weeks in patients with continued successful therapy and subsequent remission as compared with patients with subsequent recurrence; and showed higher specificity and positive predictive value for early identification of risk of recurrence as compared with a CT scan at 3 months.³³

7. Differentiating a recurrent tumor from a post-therapy non-tumorous lesion

Vascular changes associated with a recurrent tumor (rT) represent neoangiogenesis, whereas post-therapy non-tumorous (PTNT) lesions shows vascular changes of continued successful therapy and fibrosis. These changes in PTNT lesions leads to a high ADC in DWI, whereas locoregional failure with tumor recurrence leads to restricted diffusion, with a subsequent decrease in ADC.³⁴ Using contrast-enhanced CT quantitative measurements of perfusion and permeability, Bisdas et al. showed that the mean BF of an rT was 69.71 ml/min/100 g whereas the mean of PTNT lesions was 45.31.³⁵ Vandecaveye et al.³⁶ used DWI to differentiate an rT from PTNT lesions in their cohort of patients with head and neck squamous cell carcinoma. The SI on b0 and b1000 of DWI was significantly lower and higher respectively for rT than for PTNT with resulting ADC values significantly lower for rT than for PTNT (p < 0.0001), with a sensitivity of 94.6%, specificity of 95.9%, and accuracy of 95.5% (Figure 14). In the same clinical scenario of cancer treatment with CRT, patients may present for follow up with a tumefactive lesion related to the treatment site. It is difficult in most instances to characterize the lesion by using conventional imaging and typically, biopsy is sought to solve the mystery. However, based on the above principles, both DWI, and post-contrast DCE-MRI have the potential of better detecting as well as differentiating rT from PTNT changes³⁷ (Figure 15).

Figure 14.

Squamous cell carcinoma of the right lateral oral tongue at 2 months after the end of treatment evaluation in which clinically local failure with possible recurrent tumor (rT) versus PTNT was suspected. Top row T2WI (a), and fat-suppressed T1WI post I.V. Gadolinium (b) respectively shows indeterminate signal abnormality. Bottom row (c, d, and e) DWI b0, b1000 and ADC map images, probing of the abnormal signal shows SI of 196 a.u., 42 a.u., and no diffusion restriction with ADC 1.43 (×10⁻³ mm²/s) respectively, excluding tumor. The patient remained free of disease on the 2-year follow up. PTNT: post therapy non-tumorous; SI: signal intensity; DWI: diffusion weighted imaging; ADC: apparent diffusion coefficient.

Figure 15.

Recurrent buccal CA, surgery and radiotherapy ended 2 years back; the patient was then complaining of swelling in the upper gum. The soft tissue abnormality seen in the right gingiva is subtle on morphologic T1W, T2W and T1W fat-suppressed post-IV gadolinium (a, b, and c respectively) images in the top row. In contrast, there is more precise delineation on DWI b0, b1000, and ADC map (d, e, and f respectively) images in the bottom row, measuring 440 a.u., 164 a.u., and 0.9 (×10⁻³ mm²/s) respectively (tumor demonstrated by arrow in a, b, c, d, e, and f). Post-contrast dynamic DCE-MRI curve (g) shows abnormal curve (dotted line denoted by short arrow) with early peak enhancement and a perfusion peak >50% AIF (solid line curve denoted by long arrow), and gradual wash-out, confirming the clinical suspicion of a recurrent tumor. CA: cancer; DWI: diffusion weighted imaging; ADC: apparent diffusion coefficient; DCE: dynamic contrast enhanced; AIF: arterial input function.

Limitations and future applications

The first limitation is that metastatic squamous cell carcinoma lymph nodes (LNs) are heterogeneous with solid and variable degree of necrotic or micronecrotic areas, which makes it hard to have reproducible measurements between different centers, as the heterogeneity leads to higher inter- and even intra-observer variability which is a source for conflicting results in their characterization using DCE-MRI or DWI. In a study by Zhang et al., the mean ADC value of solid and necrotic portions of metastatic squamous cell carcinoma LNs were 0.93 ± 0.16 × 10⁻³ mm²/s and 2.02 ± 0.36 × 10⁻³ mm²/s respectively, and the mean ADC of inflammatory LNs were 1.25 ± 0.15.³⁸ Moreover, nodal reactive changes with multiple germinal centers and stromal fibrosis result in microstructural barriers within the reactive LNs resulting in decreased ADC.³⁹ This has led to overlapping results in the literature between metastatic and reactive LNs.^40,41 Similarly, the limitation of DCE-CT parameters in differentiating benign and malignant LNs was demonstrated by Bisdas et al.³⁵

The second limitation is related to the heterogeneity of the tumor response to treatment. CRT can reduce the bulk of tumor burden, but small regions of the tumor might not respond to ongoing treatment. These areas of local failure can be the reason why the assessment of tumor using morphological criteria may be misguiding. A promising solution to this is the delineation of tumor sub-volumes using parametric mapping information, voxel-by-voxel mapping of histogram analysis distribution of ADC values as well as DCE-MRI quantitative parameters within different ROIs, which could result in tumor dose painting and will have a substantial impact on patient management.^42,43

Lastly, functional MRI correlation with tumor immunohistochemistry biomarkers has received little attention. However, it is very likely that, by applying parametric ADC mapping and quantitative physiological perfusion parametric mapping, it will be possible to correlate tumor parameters with immunohistochemical features, for example ADC with HER2 (human epidermal growth factor 2) expression for assessment of cell proliferation, and K^trans and k_ep with COX-2 (cyclo-oxygenase 2) expression, and VEGF for angiogenesis.⁴⁴

Conclusion

Functional MRI including DWI and DCE-MRI of head and neck cancers has become a feasible effective diagnostic tool with little additional time on the scanner. It can be advantageous to sort out diagnostic and staging issues like tumor delineation, tumor extension with differentiation of T3–T4a–T4b stages, and the prediction of tumor response by identifying intra- and early post-treatment biological changes in functional MRI before morphological alterations parameters can predict a prognostic outcome. However, the most important point is that these advanced imaging studies can provide clues into biological behavior, including post-therapy changes, failure of response or recurrence of a lesion which can direct clinicians on further management. For example, if the advanced imaging features indicate post-treatment changes, the lesion can be safely followed up. On the other hand, if the advanced MRI suggest a malignant lesion, it could be used to guide a biopsy.

Funding

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Conflict of interest

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.