The Diagnostic Value of Cervical Lymph Node Metastasis in Head and Neck Squamous Carcinoma by Using Diffusion-Weighted Magnetic Resonance Imaging and Computed Tomography Perfusion

Purpose. The aim of this study was to compare diffusion-weighted magnetic resonance imaging (DWI) with computed tomography perfusion (CTP) for preoperative detection of metastases to lymph nodes (LNs) in head and neck squamous cell carcinoma (SCC). Methods. Between May 2010 and April 2012, 30 patients with head and neck SCC underwent preoperative DWI and CTP. Two radiologists measured apparent diffusion coefficient (ADC) values and CTP parameters independently. Surgery and histopathologic examinations were performed on all patients. Results. On DWI, 65 LNs were detected in 30 patients. The mean ADC value of metastatic nodes was lower than benign nodes and the difference was statistically significant (P < 0.05). On CTP images, the mean value in metastatic nodes of blood flow (BF) and blood volume (BV) was higher than that in benign nodes, and mean transit time (MTT) in metastatic nodes was lower than that in benign nodes. There were significant differences in BF and MTT values between metastatic and benign LNs (P < 0.05). There were significant differences between the AUCs of DWI and CTP (Z=4.612, P < 0.001). Conclusion. DWI with ADC value measurements may be more accurate than CTP for the preoperative diagnosis of cervical LN metastases.


Introduction
Squamous cell carcinoma (SCC) is the most common malignancy of the head and neck region. It accounts for 5% of all malignant tumours worldwide [1]. A meta-analysis by Dünne et al. [2] showed a 5-year survival rate between 17% and 55.8% for SCC with cervical node metastases and 44.6-76% for SCC patients without cervical node metastases. The presence of multiple metastatic lymph nodes (LNs) is presumed to be a worse prognostic sign [3]. The detection of cervical node metastases provides very important prognostic information and often helps decide the treatment of head and neck SCC. Node sampling is the definitive method of differentiating benign LNs from metastatic LNs, but biopsy methods are invasive and operator-dependent, with a high incidence of false-negative results [4,5]. With advancements in imaging methods, several noninvasive imaging techniques have arisen, with the potential for identifying benign and metastatic LNs in head and neck SCC, thus avoiding the complications due to biopsy sampling [6].
Although ultrasound, routine contrast-enhanced computed tomography (CT) and magnetic resonance imaging (MRI) allow the detection of enlarged cervical LNs, none of these methods can distinguish between benign and malignant causes of enlargement with any accuracy [7], as they use standard parameters (shape, size, internal architecture, extranodal invasion, and vascular features) that are nonspecific for malignancy [8,9].
Single photon emission computed tomography (SPECT) and positron-emission tomography (PET) are imaging techniques that supply functional information, but they involve radiation exposure and are expensive with low availability and are hampered by relatively low spatial resolution [10][11][12][13].
Recently, other functional imaging techniques such as MRI diffusion-weighted imaging (DWI) and CT perfusion imaging (CTP) have shown promise in detecting metastatic cervical LNs, and there is increasing experience in head and neck SCC. DWI is an MR technique that depicts molecular diffusion, which is the Brownian motion of water protons in biologic tissues. Examination of molecular diffusion using DWI performs with an EPI (echo planar imaging) sequence and a linear regression after a logarithmic transformation of the signal intensity was used to calculate the ADC values [14]. To date, the diagnosis of LN metastases has been based mainly on size criteria; however, nonenlarged nodes may harbour malignancy, whereas benign reactive nodes may be enlarged. Promising results with DWI to help detect cervical LN metastases (especially in normal-sized nodes) and to differentiate between benign and malignant enlarged nodes have been reported. The general consensus is that ADCs of malignant nodes are significantly lower than those of benign nodes [15].
Unlike conventional contrast-enhanced CT, which is normally assessed visually, perfusion imaging requires quantification of the enhancement in tissue and blood at certain time points following intravenous injection. These enhancement data are used to calculate blood flow (BF), blood volume (BV), and mean transit time (MTT) for each voxel, depicted in a color-coded display. Data processing methods are based on the robust physiological principles of compartmental analysis or linear systems theory. In the compartmental modeling technique, analysis can be done by one compartment method which assumes the intravascular and extravascular compartment as a single compartment and allows measurement of tissue perfusion during the first pass of contrast [15,16].
Although DWI and CTP are increasingly used to detect LN metastases in head and neck SCC, comparison of the diagnostic value of the two imaging modalities has been rarely performed. The aim of this study was to compare the value of DWI and CTP for detecting metastatic LNs which were confirmed by pathologic diagnosis in patients with head and neck SCC.

Ethics Statement.
The study was approved by the local ethics committee and all patients signed review boardapproved consent before participation.

Imaging Protocols
2.3.1. MRI/DWI. All MRI examinations were performed using a 1.5 T MRI unit (Philips Intera Achieva, Philips Medical Systems, Best, The Netherlands) with a head and neck coil. Thirty patients underwent conventional MRI and DWI to include nodes from the base of the skull to the suprasternal notch. Before scanning, all patients were trained to avoid swallowing during the MRI examination.
The mean BF, BV, and MTT values for metastatic nodes were 114.62 ± 14.26 mL/100 g/min, 32.15 ± 13.21 mL/100 g, and 5.56 ± 0.39 s, respectively. The mean values for BF, BV, and MTT in benign nodes were 67.82 ± 13.84 mL/100 g/min, 19.36 ± 7.34 mL/100 g, and 9.46 ± 3.23 s, respectively. There were significant differences in BF and MTT values between metastatic and benign LNs ( < 0.05) ( Table 1). The optimum threshold BF value for differentiating malignant from benign nodes was 100.36 mL/100 g/min, yielding a sensitivity of 68.18%, specificity of 52.94%, accuracy of 64.46%, PPV of 80.48%, and NPV of 37.50%. Figure 2(b) shows the ROC curves of the ADC and BF values used for differentiating benign from metastatic LNs. The areas under the curve (AUC) were 0.830 and 0.605. There were significant differences between the AUCs of DWI and CTP ( = 4.612, < 0.001) ( Table 2) for diagnosing metastatic LNs in head and neck SCC.

DWI and ADC Values.
DWI is an MR imaging-based technique whereby diffusion properties of water can be quantified using the ADC. Hypercellular tissue, such as solid tumour, is characterized by a low ADC, while more hypocellular tissue, such as normal tissue, is typically characterized by a higher ADC. As several studies reported, metastatic nodes showed a reduction of diffusivity, which can be attributed to hypercellularity, to an increased nuclear-to-cytoplasmic ratio, and to perfusion [17,18]. King et al. [19] reported that DWI could improve the accuracy in the distinction between benign and malignant nodes.
Pathological states, such as malignant tumor cell, strongly affect water diffusion more than benign tumor cell and so influence DWI, and hence the technique has attracted considerable research attention for both benign and malignant disease processes.
One of the DWI investigations performed in the head and neck [20] showed the potential of this technique for differentiating benign from malignant lesions. Srinivasan et al. [21] and Abdel Razek et al. [22] reported that there was a significant difference in ADCs between benign and malignant lesions. In one of these investigations [21], the authors found a lack of overlap between the higher ADCs of benign and the lower ADCs of malignant head and neck lesions. Using an ADC obtained with two values (0 and 800 s/mm 2 ) and a 3 T MR unit, they established an optimal ADC threshold of 1.3 × 10 −3 mm 2 /s for diagnosis. The authors of the other DWI investigation [22] confirmed the difference in ADCs between benign and malignant lesions. In that study, values of 0, 500, and 1000 s/mm 2 at 1.5 T yielded a significant difference between benign and malignant lesions ( < 0.001). Also, an optimal ADC threshold of 1.25 × 10 −3 mm 2 /s to help differentiate benign from malignant lesions was established, consistent with that of the previous study [21], and yielded an accuracy of 92.8%, sensitivity of 94.4%, and specificity of 91.2%. Although these results showed the advantage of DWI for diagnosis of head and neck lesions, there will be exceptions and overlap in ADC results. Therefore, a single ADC threshold cannot be used in all conditions, and combining it with specific sites and morphological findings will be necessary.
In our series, the evaluation with DWI showed that metastatic nodes appeared hyperintense ( = 600 mm 2 /s); conversely, benign nodes were hypointense ( = 600 mm 2 /s). We chose 0.960 × 10 −3 mm 2 /s as the optimal ADC threshold value for distinguishing benign from metastatic nodes, with a sensitivity of 89.58%, a specificity of 76.47%, an accuracy of 86.15%, a PPV of 91.48%, and an NPV of 72.22%. Our data are not in agreement with those of Kato et al. [23] who found, for SCC LNs, the mean ADC value (1.45 ± 0.48 × 10 −3 mm 2 /s) to be higher than that in benign lymphadenopathies (0.89 ± 0.21 × 10 −3 mm 2 /s).
Differences among these studies can be attributed to several causes. One is the choice of the values: a lower value increases signal-to-noise ratio but lowers the sensitivity to diffusion. Other factors are the size of the ROI (solid portion, necrotic portion, or whole area) and the use of sequences that reduce the artefacts in order to make the measurement of the ROI more precise [24].

CT Perfusion.
CTP is a technique that allows quick qualitative and quantitative evaluation of tissue blood perfusion by generating maps of BF, BV, and MTT. Gandhi et al. [25] reported that CT perfusion parameters may provide valid information on angiogenic activity induced by neoplastic cells invading LNs. Tumour vessels in malignant LNs have certain characteristics, including short artery and vein circuits and a lack of smooth muscle around the vessel walls, which can result in increased blood perfusion [26]. In our study, 13/17 metastatic LNs showed high BF and BV and short MTT. There were significant differences between metastatic LNs and benign LNs in BF and MTT. In metastatic nodes, BF is increased by newly developed vessels, while MTT is usually decreased by the presence of pathological arteriovenous shunts. Changes in vascular endothelium and in the function of vessels are induced by neoangiogenesis [27,28].
In our study, 15/48 metastatic LNs also had lower BF, reflecting the necrotic tumour elements (although any macroscopic nonviable areas were excluded from the analysis), a quite common phenomenon in malignant tumours [29,30]. This type of inter-or intralesional heterogeneity sometimes resulted in high standard deviations of the perfusion values. This heterogeneity could be a pitfall of measuring perfusion values in ROIs (variable in size and location) scattered through the LNs [26]. It may have been the factor behind the significant differences in BF values, and not in BV values, between benign and metastatic nodes. The best threshold BF value for differentiating malignant from benign nodes was 100.36 mL/100 g/min, yielding a sensitivity of 68.18%, specificity of 52.94%, accuracy of 64.46%, PPV of 80.48%, and NPV of 37.50%.

Comparison between DWI and CT
Perfusion. DWI has been reported to be able to distinguish between malignant and benign LNs with sensitivities ranging from 52% to 98% and specificities ranging from 88% to 97% [31,32]. ADC values played an important role in differentiating between benign and metastatic LNs. Some authors reported similar threshold ADC values such as 0.94 × 10 −3 mm 2 /s. Our data was in agreement with the results mentioned above and obtained a sensitivity of 89.58%, a specificity of 76.47%, and an accuracy of 86.15%.
In CTP, there were general differences between benign and metastatic LNs, but with some overlap [33]. We obtained a sensitivity of 68.18%, specificity of 52.94%, and accuracy of 64.46% with BF 100.36 mL/100 g/min as the threshold value. The ROC curve was used to evaluate the two imaging techniques, with an AUC of 0.830 (DWI) and 0.605 for CTP ( < 0.001). In addition, the patients undergoing CTP were injected with an iodinated contrast agent that has been confirmed to have some nephrotoxicity; meanwhile, patients inevitably suffer from radiation injury.

Conclusion
In conclusion, DWI exhibited better sensitivity and specificity than CTP. DWI may be the preferred technique for the preoperative assessment of LNs in head and neck SCC.

Abbreviations
DWI: Diffusion-weighted magnetic resonance imaging CTP: Computed tomography perfusion LN: Lymph node SCC: Squamous cell carcinoma ADC: Apparent diffusion coefficient PPV: Positive predictive value NPV: Negative predictive value.