High b-Value Diffusion MRI to Differentiate Recurrent Tumors from Posttreatment Changes in Head and Neck Squamous Cell Carcinoma: A Single Center Prospective Study

Recently DW-MR Imaging has shown promising results in distinguishing between recurrent tumors and posttreatment changes in Head and Neck Squamous Cell Carcinoma (HNSSC). Aim of this study was to evaluate the diagnostic performances of DWI at high b-value (b = 2000 s/mm2) compared to standard b-value (b = 1000 s/mm2) and ADCratio values (ADCratio = ADC2000/ADC1000 × 100) to differentiate recurrent tumors from posttreatment changes after treatment of HSNCC. 20 patients (16 M, 4 F) underwent MR Imaging between 2 and 16 months (mean 7) after treatment. Besides morphological sequences, we performed single-shot echo-planar DWI at b = 1000 s/mm2 and b = 2000 s/mm2, and corresponding ADC maps were generated (ADC1000 and ADC2000, resp.). By considering contrast-enhanced T1-weighted images as references, ROIs were drawn in order to evaluate mean ADC1000, ADC2000, and ADCratio. The mean ADC1000 and ADC2000 in recurrent tumors were significantly lower than those in posttreatment changes (P = 0.001 and P = 0.016, resp.). Moreover, the mean ADCratio between the two groups showed a statistically significant difference (P = 0.002). Sensitivity, specificity, and accuracy of ADCratio were 82.0%, 100%, and 90%, respectively, by considering an optimal cutoff value of 65.5%. ADCratio is a promising value to differentiate between recurrent tumors and posttreatment changes in HNSCC and may be more useful than ADC1000 and ADC2000.


Introduction
Squamous cell carcinoma represents almost 90% of the head and neck tumors (HNSCC) and it shows different biological behaviors according to location [1].
Imaging techniques are commonly required in order to define tumor's locoregional extension [2]. Moreover MRI is increasingly becoming the preferred examination method as it provides additional information on tumor extension, muscles and lymph nodes involvement, and skull base and intracranial invasion. The role of MRI can therefore modify the clinical staging and consequently therapeutical approach, which is multidisciplinary with surgery, radiation therapy, and/or chemotherapy.
In these patients, however, even if treatment improves survival and quality of life, on the other hand, it can delay detection of residual or recurrent tumor in posttreatment follow-ups as surgery can modify anatomy and radiochemotherapy can result in edema, inflammation, fibrosis, and, sometimes, necrosis [3][4][5].
Conventional MR Imaging gives sometimes wrong information as postcontrast enhancement of a benign lesion may mimic a residual or a recurrent tumor, leading to an increased number of false positives and unnecessary surgical approaches. On the contrary, a recurrent tumor may not show enhancement after the administration of contrast medium.
Recently, DW (diffusion-weighted) MR Imaging has shown promising results. DWI is based on the Brownian motion of water protons, which is affected by the microstructure of tissue. Several promising studies have been reported on the usefulness of DWI in the discrimination between recurrent or residual tumor and posttreatment changes [6][7][8].
The aim of our present study was to evaluate the diagnostic performance of high b-value (b = 2000 s/mm 2 ) DW sequences compared with standard b-value (b = 1000 s/mm 2 ) DWI, and the ratio of ADC values from high and standard b-values for the differentiation between recurrent tumor and posttreatment changes after treatment (chemoradiotherapy and/or surgery) of HSNCC.
As regarding patients with tongue cancer, one patient (Pt. number 20) was treated with surgery and then chemotherapy (cisplatin) and another one (Pt. number 6) with concomitant chemoradiotherapy while the third one underwent both a conservative surgical approach and right laterocervical lymph nodes dissection (II-III levels) followed by concomitant chemoradiation.
Patients affected by oropharynx cancer (Pt. numbers 12 and 15) were treated with chemotherapy and radiotherapy.
Those with ductal epidermoid cancer of the parotid gland (Pt. numbers 10, 14, and 19) were treated with parotidectomy followed by chemotherapy and radiotherapy.
The patient with a transglottic laryngeal cancer (Pt. number 7) was treated with chemotherapy and concomitant radiation therapy.
The one with a spinocellular cancer of the right cheek (Pt. number 8) had the resection of the lesion followed by locoregional radiotherapy.
Radiotherapy was performed with a linear multileaf accelerator and radiation field dimensions were decided by taking into account tumor extension, staging, and risk of lymph nodes involvement.   ADC ratio maps (ADC ratio = ADC 2000 /ADC 1000 × 100, where ADC 1000 and ADC 2000 are the ADC values of DWI obtained with b = 0 and 1000 s/mm 2 , and b = 0 and 2000 s/mm 2 , resp.) were generated by a specific software by use of pixel-by-pixel computation of ADC maps.

MR
For each patient, two radiologists analysed both conventional MRI and diffusion-weighted (DW) sequences at the same time and recurrence versus posttreatment changes diagnosis was obtained by radiologists' concordance.
MRI examination was followed by biopsy in all patients in order to confirm the diagnosis.

Qualitative Analysis: Conventional MRI and DW
Sequences. The analysis included tumor site alterations such as high signal intensity on T2w and STIR sequences, tissue volume increase (focal or diffuse), mass effect, and enhancement pattern after contrast medium injection.
The increasing volume of focal areas of enhancement, at subsequent controls, was considered highly suspicious of tumor recurrence while volume reduction of the lesion or loss of growth at follow-ups was considered as posttreatment changes.
Regarding DW sequences, high signal intensity on high bvalue images but low signal intensity on corresponding ADC maps were considered suggestive of recurrent tumor whereas high signal intensity on high-b-value images but generally high signal intensity on the corresponding ADC maps were considered as a result of T2 shine-through effect and therefore predictive of posttreatment changes.

Quantitative Analysis: DW Sequences.
In order to analyse ADC values in suspected areas, radiologists placed ROIs (regions of interest) on the axial ADC 1000 maps with references of contrast-enhanced T1-weighted images obtained in 3 orthogonal planes. ROIs were drawn on the most representative section of the ADC map, in which the size of the tumor was the largest or the conspicuity of the lesion was the highest.
ROI was drawn as large as the visible tumor was on that section of the ADC map corresponding to the contrastenhanced T1-weighted images, trying to avoid necrotic portions and normal osseous structures.
Subsequently, ROIs were copied onto the corresponding ADC 2000 and ADC ratio maps, respectively.
The size of each ROI was also recorded, varying according to tumor extension (average size 189 ± 65 mm 2 ).

Statistical Analysis.
Statistical analysis was performed with SPSS Statistics 19.0 for Windows. For all statistical analyses, a 2-tailed value of 0.05 was considered to indicate a statistically significant difference. The 2-tailed independent Student's t-test was used to compare mean ADC 1000 , ADC 2000 , and ADC ratio values between the group with recurrent tumor and the group with post-treatment changes.
Student's t-test was also performed in order to compare mean ADC 2000 values decreasing percentages versus ADC 1000 ones between recurrent tumor group and posttreatment changes one.
A ROC (receiver operating characteristic) curve was drawn to investigate the optimal cutoff values for ADC 1000 , ADC 2000 , and ADC ratio in order to obtain the best sensitivity, specificity, and accuracy in distinguishing between recurrent tumor and posttreatment changes.

Results
In our study, 11 patients were found to have a recurrent tumor and 9 posttreatment changes. Conventional MRI sequences showed different degrees of anatomic changes after therapy, with a tissue asymmetric increased thickness as the most common. Suspected areas volume increase, together with a high signal intensity on T2w and STIR images and enhancement after contrast injection were found in 10 patients out of 11 with pathologically confirmed tumor and in 5 patients out of 9 with posttreatment changes (sensitivity 90%, specificity 44%, and accuracy 70%).
Recurrent tumors appeared as high signal intensity areas on DWI sequences and low signal ones in the corresponding ADC maps (Figure 1), except for one lesion that had a high signal intensity in the ADC maps; in fact this area showed a relatively high ADC 1000 value (1.50 × 10 −3 mm 2 /s) and intermediate ADC 2000 and ADC ratio values (0.85 × 10 −3 mm 2 /s and 56%, resp.) and it was considered as a posttreatment change (Pt. number 19).
With information given by DWI sequences, MRI accuracy was 85%, with sensitivity and specificity of 90% and 78%, respectively.
In our study the smallest lesion found to be a recurrent tumor had a maximum diameter of 8 mm. The mean ADC 1000 between the two groups showed a statistically significant difference ( = 0.001).
ADC ratio values varied from 54% to 80% with a mean value of 69.5 ± 9.3% in the recurrent tumor group, while they varied from 46% to 65% with a mean value of 56.8 ± 6.3% in the posttreatment changes group.
The difference between the mean ADC ratio of the recurrent tumor group and that of the posttreatment changes one was statistically significant ( = 0.002) (Tables 2 and 3).
ADC 2000 values were significantly lower than ADC 1000 ones ( < 0.001) with a decreasing percentage of 31 ± 9.7% in the recurrent tumor group and 42.6 ± 6.5% in the posttreatment changes group (Table 4).
A statistically significant difference was also found in the decreasing percentage of the mean ADC 2000 versus ADC 1000    between recurrent tumor group and posttreatment changes (31 ± 9.7% versus 42.6 ± 6.5%; = 0.007).

Discussion
DW imaging is a functional MR Imaging technique that enables depiction and quantification of the Brownian motion   [16]. According to these encouraging results, several studies have been conducted to assess the role of DW sequences in distinguishing between recurrent tumor and posttreatment changes in head and neck squamous cell tumors. In 2010 King et al. showed how DW imaging enables detection of early pathologic changes after chemoradiotherapy in 50 patients with HNSCC and particularly found that ADC 1000 value measured in a suspected area at follow-up is an important marker of recurrent tumor. Moreover, with a cutoff value of 1.40 × 10 −3 mm 2 /s, the accuracy was 95% [6].
In a retrospective study on 33 patients (Hwang et al., 2013) with high suspicion of recurrent tumor after chemoradiotherapy and/or surgery, with DW imaging at standard b-values (0 and 1000 s/mm 2 ) and high b-values (0 and 2000 s/mm 2 ), the comparison of ADC 1000 , ADC 2000 , and ADC ratio values between the two groups of patients (recurrent lesions versus posttreatment changes) depicted a statistically significant difference only for ADC 1000 and ADC ratio mean values in the two groups. By choosing a cutoff value of 1.46 × 10 −3 mm 2 /s for ADC 1000 and of 62.6% for ADC ratio the diagnostic accuracy was 85% and 84.8%, respectively [11].
Our results demonstrate that ADC values derived from standard b-value DWI (1000 s/mm 2 ) had a similar diagnostic performance compared with previous studies and that the mean ADC 1000 in the posttreatment changes group was significantly higher than the one of the recurrent tumor group (1.90 ± 0.44 × 10 −3 mm 2 /s versus 1.27 ± 0.22 × 10 −3 mm 2 /s, resp.; = 0.001).
The optimal cutoff value for the ADC 1000 (1.53 × 10 −3 mm 2 /s) to distinguish between recurrent tumor and posttreatment changes and the diagnostic accuracy (90%) is also related to previous studies (Table 6). These findings support the use of standard b-value DWI on posttreatment follow-up imaging of HNSCC as the mean values of the ADC 1000 in the recurrent tumor group are significantly lower than those of the group with posttreatment changes.
According to Hwang et al. 's study, we observed a significant difference in the mean ADC ratio between the recurrent tumor group and the posttreatment changes one (69.5 ± 9.3% versus 56.8 ± 6.3%, resp.; = 0.002); with a cutoff value of 65.5% the diagnostic accuracy was 90% (Table 7).
The two groups ADC values reflect the differences in histopathology and in water molecules distribution between tumors and posttreatment changes. Indeed, ADC is inversely related to tumor cellularity as the increase in cells number and in nucleus dimensions in malignant lesions leads to limitation of diffusion of water molecules and, consequently, to lower ADC values.
On the other hand, decreased cellularity and the presence of edema and inflammatory changes in posttreatment modification are related to higher ADC values.
According to our results, ADC values have also a substantial decrease by increasing the b-value over 1000 s/mm 2 .
DeLano et al. subjected 50 patients to brain diffusionweighted imaging with b-values of 0, 1000, 2000, 2500, 3000, and 3500 s/mm 2 . For 6 patients, apparent diffusion coefficient maps were generated at b-values of 0 and at b-values of 1000, 2000, and 3000 s/mm 2 . Quantitative assessments were made in multiple regions of interest in gray and white matter and, for some ROI, the ADC decreased approximately of 30% to 35% for b-values between 1000 and 3000 s/mm 2 . These data  indicate that if the relationship between MR signal and bvalues would have been monoexponential, the ADC value should remain constant as the b-value increases. A likely explanation is that apparent diffusion is more accurately represented by a biexponential relation [17]. Both in animal and human models, rapidly decaying and slowly decaying diffusion components have been described. At low b-values, signal intensity is dominated by a fast diffusion component, fast , whereas, at higher b-values, signal intensity is contributed largely by the slower component of diffusion, slow . Intracellular and extracellular water are considered to correspond to slow and fast diffusion components, respectively, even if not exactly equal to them.
The increasing of intracellular water due to hypercellularity of a recurrent tumor and the reduction of the extracellular space can thus explain the more significant influence of the slow diffusion component on DWI sequences; on the contrary, posttreatment changes edema is related to the predominance of the extracellular water and, consequently, of the fast diffusion component [18].
In our study ADC 1000 and ADC 2000 maps were obtained from DWI sequences at two b-values (0 and 1000 s/mm 2 ; 0 and 2000 s/mm 2 , resp.) with a total scan time of 3 min and 20 sec.
We hypothesized that, by increasing the b-value from 1000 s/mm 2 to 2000 s/mm 2 , the decrease in ADC value would reflect the proportion between slow and fast diffusion components in suspected areas.
The analysis of the two groups (recurrent tumor and posttreatment changes) showed a lower decreasing percentage of the mean ADC 2000 compared to the mean ADC 1000 in recurrent tumor group (31 ± 9.7% versus 42.6 ± 6.5%; = 0.007) and a higher ADC ratio (69.5 ± 9.3% versus 56.8 ± 6.3%; = 0.002). We suggest that the lower decrease of ADC values of recurrent lesions by increasing the b-value from 1000 to 2000 s/mm 2 can be easily explained by the higher cellularity thus reflecting the greater influence of the slow diffusion component. Furthermore, in tumors, the greater the proportion of slow diffusion component becomes, the higher the ADC ratio increases, and, for this reason, the latter might become an alternative biomarker for higher cellularity and it might also simplify the detection of recurrent tumor by visual inspection in the ADC ratio maps.
Our results are in agreement with those of earlier studies.
In the present study, qualitative and quantitative analysis of DWI sequences and the corresponding ADC maps revealed one false negative and two false positives.
Signal intensity on diffusion-weighted images is influenced by two factors: water diffusibility and the intrinsic T2 properties of the tissue being examined. In tissues with very long T2 relaxation times, the strong T2 signal may be mistaken for restricted diffusion, a phenomenon known as "T2 shine-through effect." The easiest way to distinguish between restricted diffusion and T2 shine-through is to generate an ADC map, on which the former appears as an area of low signal intensity (low ADC values) and the latter as a high-signal-intensity area [19].
Therefore, high signal intensity on high-b-value images and low signal intensity on corresponding ADC maps were considered suggestive of recurrent tumor whereas high signal intensity on high-b-value images and generally high signal intensity on the corresponding ADC maps were considered as a result of T2 shine-through effect and therefore predictive of posttreatment changes.
The relatively high ADC values of the recurrent tumor wrongly classified as posttreatment changes may be correlated with the presence of edema and small areas of postradiation colliquative necrosis as well reported by Matzek et al. in a previous study on oropharynx squamous cell carcinoma [20]. As concerns the two false positives, they underwent radiotherapy in the previous year. Nomayr et al. reported that the first visible effect in the first 6 months after irradiation is edema together with fibroinflammatory reaction and increasing permeability due to damage of vessels endothelium. Later on (6-24 months after irradiation) there is the regression of inflammatory edema and the increasing in fibrous reaction, which shows a restricted diffusion [4].
The choice of regions of interest (ROIs) places an important role in order to reduce pitfalls in the evaluation of mean ADC values. Firstly, ROIs should be drawn manually and no necrosis areas should be included; secondly, the more heterogeneous the lesion, the more the ROIs that should be considered from which a mean value should be extrapolated.
However, partial volume artifacts and areas of micronecrosis cannot be avoided during measurement of ADC values therefore still representing a sampling source of errors [21]. The overlapping of ADC values could be associated with partial volume artifacts due to structures sizes, wrongly low ADC values can be obtained with fibrosis, while wrongly high ADC values can be depicted in necrotic areas of recurrent tumors.
DWI single-shot echoplanar sequence (EPI) was performed for qualitative and quantitative analysis. It has a time resolution of less than 200 ms, thus guaranteeing data acquisition at different b-values in a relatively short total scan time. However its weakness is its extreme sensitivity to susceptibility artifacts consequently producing more geometrical distortion artifacts with the increase of b-values.
The head and neck region is particularly difficult for performing DW imaging acquisitions because it is very heterogeneous, containing a variety of tissues that include fat, muscle, air, glandular tissue, and bone. It also has very low acquirable signal (from muscle and suppressed fat) and a strongly changing geometric shape. This can yield images with a very low signal-to-noise ratio and strong susceptibility artifacts from many air tissue boundaries, as well as from metallic surgical implants and dental fillings. Moreover, head and neck area is subject to a number of movement-related problems [22][23][24].
One limitation of our study concerns patients' enrollment as the HNSCC population was quite heterogeneous (various locations and pathologic condition, different treatment modalities, and timing of imaging) and this heterogeneity has made the homogenization of our results difficult.
In our opinion, however, lesion location should not have a significant impact on ADC values whereas the standardization of imaging timing, which could theoretically have an influence on results, is difficult to apply.
Future studies on a less heterogeneous population are required to better analyze the role of these variables on ADC values.

Conclusions
The present prospective study on HNSCC patients treated with surgery and/or chemoradiotherapy has shown high accuracy of DW-MR Imaging in distinguishing between recurrent tumor and posttreatment changes. We also suggest that ADC measurement both at standard and high b-values is required to assess slow diffusion component, which is directly proportional to intracellular water. We suggest that the ADC ratio calculated from ADC 1000 and ADC 2000 is a promising value to differentiate between recurrent tumor and posttreatment changes in HNSCC and may be more useful than ADC 1000 and ADC 2000 alone.
High b-value DWI of head and neck region is technically feasible and requires a relatively short additional scan time but should always be analysed by taking into account morphological sequences in order to correctly localize lesions. However, larger studies are required for the standardization of this imaging to improve single patient application.