Predictive Value of Diffusion, Glucose Metabolism Parameters of PET/MR in Patients With Head and Neck Squamous Cell Carcinoma Treated With Chemoradiotherapy

Purpose: This study aims to evaluate the predictive value of the pretreatment, metabolic, and diffusion parameters of a primary tumor assessed with PET/MR on patient clinical outcomes. Methods: Retrospective evaluation was performed using PET/MR image data sets acquired using the single tracer injection dual imaging of 68 histologically proven head and neck cancer patients 4 weeks before receiving definitive chemoradiotherapy (CRT). PET/MR was performed before the CRT and 12 weeks after the CRT for response evaluation. Image data (PET and MRI diffusion-weighted imaging [DWI]) was used to specify the maximum standard uptake value, the peak lean body mass corrected, SUVmax, the metabolic tumor volume, the total lesion glycolysis (SUVmax, SULpeak, MTV, and TLG), and the mean apparent diffusion coefficient (ADCmean) of the primary tumor. Based on the results of the therapeutic response evaluation, two patient subgroups were created: one with a viable tumor and another without. Metabolic and diffusion data, from the pretreatment PET/MR and the therapeutic response, were correlated using Spearman's correlation coefficient and Wilcoxon's test. Results: After completing the CRT, a viable residual tumor was detected in 36/68 (53%) cases, and 32/68 (47%) patients showed complete remission. However, no significant correlation was found between the pretreatment parameter, ADCmean (p = 0.88), and the therapeutic success. The PET parameters, SUVmax and SULpeak, MTV, and TLG (p = 0.032, p = 0.01, p < 0.0001, p = 0.0004) were statistically significantly different between the two patient subgroups. Conclusion: This study found that MRI-based (ADCmean) data from FDG PET/MR pretreatment could not be used to predict therapeutic response although the PET parameters SUVmax, SULpeak, MTV, and TLG proved to be more useful; thus, their inclusion in risk stratification may also be of additional value.


INTRODUCTION
Head and neck carcinomas are the sixth most common cancers nowadays. These carcinomas make up 6% of all new cancer cases recorded yearly (1,2). The majority of head and neck carcinomas belong to the histopathological group of squamous cell carcinoma of the head and neck (HNSCC) (3).
The main clinical staging component for diagnosing HNSCC is the endoscopy, but conventional radiological staging methods, such as computed tomography (CT) and magnetic resonance imaging (MRI) have proven more accurate and informative in setting up a diagnosis (4).
Beyond these conventional imaging methods, hybrid imaging has also shown an outstanding staging ability, particularly in detecting or characterizing head and neck cancers (5). Hybrid imaging, such as positron emission tomography/computed tomography (PET/CT) or PET/MR, is an imaging solution that could be used simultaneously with anatomical information to provide metabolic data (with a 18 F-FDG: 2-Deoxy-2-[18F]fluoro-D-glucose [ 18 F-FDG] tracer). With the information obtained using hybrid imaging, oncological practice could conduct many diagnostic or therapeutic procedures, for example, whole-body staging/restaging, irradiation planning, or even the evaluation of the disease prognosis (6,7).
To characterize HNSCC, it is essential to use PET imaging with an 18 F-FDG tracer. Multiparametric data obtained from the 18 F-FDG PET evaluation was not only linked with histopathologically confirmed tumor properties, but also connected with PET/MR parameters (such as the apparent diffusion coefficient, ADC, derived from diffusion-weighted imaging [DWI] examinations, the maximum standardized uptake value [SUV max ], and the peak lean body mass corrected, SUV max [SUL peak ]), and treatment-associated failure, locoregional recurrence, and death (8). In many malignancies, 18 F-FDG accumulation (in the most common form, known as SUV) appears to be a good indicator of disease aggressiveness (9). There are numerous studies aimed at the utility of FDG PET parameters in predicting response to CRT in head and neck cancer specifically (10,11).
The PET/MR method, combined with conventional contrastenhanced (CE) MR sequences, is excellent for getting data on anatomical and metabolic conditions although data on the cellular diffusion of the scanned area could also be derived via DWI methods during the MR acquisition (12). Besides the clinical and histopathological factors, imaging parameters may provide important prognostic biomarkers in different malignancies (8). DWI can measure water molecules' movement and the tumor cell density in tissue in vivo (13). DWI methods could be used for staging HNSCC. In some cases, DWI allows for a more accurate staging method than PET/CT (e.g., evaluating cN0) (14). DWI-derived variables, such as ADC, may have a prognostic and predictive value that relate to the post-therapeutic status of the disease and the outcome of chemoradiotherapy (15). To investigate the predictive value of ADC, scans must be performed before and after treatment. According to the results using this method, ADC could be an indicator of locoregional failure, which is a component of the treatment response. The ADC mean values (ADC mean ) are, therefore, possible parameters for prediction, per the suggestion of Martens et al. (16). Leifels et al. found that tumor metabolism, cellularity, and perfusion show complex relationships in HNSCC. Furthermore, these associations depend on tumor grading (17).
Moreover, metabolic tumor volume (MTV) and total lesion glycolysis (TLG) seem to be predictors of the postoperative survival of patients that have been diagnosed via PET/CT with MTV seeming to be a better predictor than TLG (18). Overall, there are numerous studies aimed at the utility of FDG PET parameters in predicting response to CRT in head and neck cancer specifically (10,11).
This study aimed to determine the best predictors for the treatment outcome of patients diagnosed using a single tracer injection dual imaging acquisition protocol in PET/MR from a set of previously described parameters (SUV max , SUL peak , ADC mean , MTV, and TLG). This study also aimed to evaluate the connection between the possible above parameters that prove to be the most predictive of the HNSCC outcome.
Primary tumor localizations were: pharyngeal (n = 32), sublocalized into seven patients nasopharyngeal, 13 patients oropharyngeal, and 12 patients with hypopharyngeal. Laryngeal (n = 36), sublocalized into 26 with supraglottic, four glottic, and six subglottic. The epidemiology data specific to the tumor and lymph node and the response to therapy are summarized in Table 1.

PET/MR Acquisition
Examinations were performed using a hybrid PET/MR scanner (Biograph mMR, Siemens Healthcare GmbH, Erlangen, Germany). Blood glucose level was checked before tracer injection to ensure the patients were euglycemic. The patients received intravenous administration of 4 MBq/kg activity of 18 F-FDG. Then, PET/CT (Truepoint 64, Siemens Healthcare GmbH, Erlangen, Germany) was performed, using FDG initially injected for PET/CT (60 ± 10 min of the uptake period) before PET/MR (15 ± 5 min after PET/CT). Further tracer injection was not applied for PET/MR (single tracer injection dual imaging acquisition protocol). After proper patient preparation (removal of metal implants, hearing aids, metal objects in the region), images were obtained of the head and neck position using dedicated coils. Thus, only PET/MR parameters were included in the research.
Diffusion-weighted (DW) measurement was done as part of a routine examination. In this case, a 2-D spin-echo DWI echo-planar (EP) sequence (FOV: 315 mm, TR: 9,900 ms, TE minimum: 70 ms, TI 200 ms, slice thickness: 5 mm) was used. An ADC map was automatically generated from the DWI pictures via the implemented software. The restricted diffusion rate was quantified by calculating the apparent diffusion coefficient. To reduce the perfusion effect (on the ADC calculation, a 50 s/mm 2 "b" value was used as the first measurement (the other b values were 800 and 1,000 s/mm 2 ). Furthermore, an axial Dixon FS T1-weighted TSE sequence and a coronal TSE Dixon FS sequence were conducted after 0.1 mmol per kg of bodyweight contrast material (Gadovist © Bayer Healthcare, Leverkusen, Germany) was injected into the patient (Figure 2). The imaging was repeated after the completion of the CRT for therapeutic response assessment.
For PET data collection, a magnetic resonance-based attenuation correction ([MRAC], using a CAIPIRINHAaccelerated T1-weighted Dixon 3D-VIBE sequence) was used for PET attenuation correction, and the wide range bed position PET Emission scan was acquired for 900 s with a fixed FOV range (20 cm) and a (172 × 172) matrix without bed movement. An iterative ordered subset expectation maximization (3-D OP-OSEM) PET image reconstruction algorithm was used with Frontiers in Oncology | www.frontiersin.org three iterations and eight subsets as well as 4-mm Gaussian filtering settings. PET data was corrected for scatter, random coincidences, and attenuation using the MR data.

Image Analysis
Metabolic parameters were calculated using a dedicated Syngo.via (Siemens Medical Solutions, VB20, Siemens Healthcare GmbH, Erlangen, Germany) multimodality image evaluation and postprocessing application based on fusioned PET/MR imaging. The SUV max , SUL peak , MTV, and TLG data of the primary head and neck cancers were collected using the volume of interest (VOI) technique. This study was built only on one observer assessment. VOIs were assessed by a nuclear medicine physician with 15 years of experience. The SUV max represents single voxel activity concentration in a particular lesion with the highest uptake. The SUL peak is defined as a lean body mass normalized-average SUV value measured in a 1 cm 3 volume spheric region of interest (ROI) centered around the hottest point in the tumor foci. For the MTV and TLG definition, the relative threshold at 50% of tumor SUV max was used as proposed by Deron et al. (19) where MTV represents the volume of the above given VOI while TLG is the product of the VOI average SUL (SUL mean ) multiplied by the corresponding MTV.
The localization of lesions was assessed on the ADC map using eRAD PACS Desktop Viewer 8.0 software. This study applied the single slice measurement method; we have chosen the largest and the most homogeneous part of the tumor as a standard for all objects (20). ROI was placed manually on the most solid part of the tumor, which shows the highest signal intensity on DWI images (hyperintense) and hypointense on ADC map (21,22). ROIs were measured by a radiologist with 10 years of experience in DWI measurement. Thus, during the ADC measurement, the researchers took precautions, such as excluding areas of gross necrosis from the sample (ROI) while plotting an elliptic ROI (15). In all lesions, ADC mean was used as a standard measurement unit to minimize the effect of tumor heterogeneity, it also was the standard unit to be used as a reliable parameter because it reflects the heterogeneity of the tumor in the specified slice to enable the researcher to distinguish the different entities in the same image (5,6,22) (Figure 2).

Clinical Evaluation
To evaluate the therapeutic tumor responses based on pre-and post-treatment PET/MR and PET/CT information, the European Organization for Research and Treatment of Cancer (EORTC) (23) system was used. Two patient groups were established according to the results of the PET/MR therapeutic response evaluation and the clinical follow-up. Furthermore, patient subgroups were also set up, namely a complete remission (CR) group defined as patients with an absence of a viable primary tumor tissue and a non-complete remission (NCR) group defined as patients with any pernicious proliferations including partial response, stable disease, and progressive disease groups (24,25).

Statistical Analysis
For all the statistical analyses conducted, R-scripts developed inhouse based on the R-software environment for statistical   The Shapiro-Wilks test (26) was used to check the normality of the measured SUV max , SUL peak , TLG, MTV, and ADC mean data. Because these tests showed non-normality distributions of the SUV max (p < 0.0001), SUL peak (p = 0.0001), TLG (p < 0.0001), and MTV (p < 0.0001) in the population, Spearman's correlation coefficient was used to describe the strength of the correlation between the data pairs, and Wilcoxon's rank-sum test was used for group comparison. The estimated parameters were correlated in different tumor subgroups (grade 1-2 and 3) per suggestion of Leifels et al. (17).

RESULTS
A total of 68 patients were enrolled in this study. The patients' characteristics are summarized in Table 1
The results of the correlation analysis are summarized in Figure 4. No significant correlation between SUV max , SUL peak , TLG, MTV, and the ADC mean for the patients diagnosed using the single tracer injection dual imaging acquisition protocol was noted.
On the next step, in two separated tumor subgroups, the estimated parameters were correlated. In G1/2 tumors, all PET parameters correlated well ( Table 3). In G3 tumors, PET parameters also have shown significant correlations ( Table 4). FIGURE 4 | Correlation plots (lower triangle), histograms (diagonal), and Spearman correlation coefficients in the upper triangle are shown in the plot-matrix of pretreatment maximum of standardized uptake value (SUV max ), peak lean body mass corrected SUV max (SUL peak ), total lesion glycolysis (TLG), metabolic tumor volume (MTV), mean apparent diffusion coefficient (ADC mean ). The significance levels are denoted by stars (**p < 0.01, ***p < 0.001, no star: no significant).
Finally, PET imaging-based parameter values did not correlate with ADC mean in both groups.

Measured Parameters and Response
According to Wilcoxon's rank-sum test, no statistically significant difference was found for the ADC mean (p = 0.88) of patients that achieved a complete response and subjects with a viable tumor tissue after CRT. Nevertheless, SUV max , SUL peak , TLG, and MTV (p = 0.032, p = 0.01, p < 0.0001, p = 0.0004) proved to be significantly different between the two different outcome groups (Figure 5).

DISCUSSION
The radiotherapy of HNSCC patients based on modern complex oncological treatment is usually combined with chemotherapy and/or surgical resection (27). HNCs still have a bad overlook in the overall prognosis of the combined treatment modalities.    An overall loco-regional recurrence may occur in up to 40% of locally advanced head and neck patients after the first 2 years (28). Due to the anatomical features of the head and neck region, organ preservation is important to maintain functions and to minimize aesthetic changes (29). Hoffman et al. raised some attention regarding neoadjuvant treatment strategies for tumor reduction before surgery. They also pointed out the efficacy of CRT and neoadjuvant chemotherapy followed with definitive radiotherapy for advanced HNSCC patients (30). The study also highlighted the need to accurately predict the outcome of possible treatment options in daily clinical practice. The high mortality rate of advanced HNSCC patients and the precise cancer staging of radical resections are, therefore, essential, as both allow clinicians to select the relevant treatment strategies that could predict the prognosis of the patients (31). Hence, it is essential to identify the potential predictive indicators for these treatments. Pretreatment 18 F-FDG-PET/MR were evaluated for their predictive value for clinical outcomes. It is crucial to prognosticate the disease response of treatments in the pretreatment period to establish a more aggressive treatment for selected HNSCC patients (7,32).
Overall, in this research, we focused on the combined role of DWI and PET imaging parameters for predicting tumor response to therapy in the head and neck region.
In this examination, SUV max , SUL peak MTV, and TLG values of HNSCC patients were the predictive factors for determining response to therapy. After CRT, the risk of NCR was significantly higher in patients with high SUV max , SUL peak , MTV, and TLG values than in patients with low SUV max , SUL peak , MTV, and TLG values. Thus, the current results confirm that both TLG and MTV can add valuable information for prediction, further supporting Pak et al.'s finding, which argues that patients who have a higher risk of death and adverse events have high MTV or TLG (33). Additionally, in the current study, the noncomplex (SUV max , SUL peak ) parameters supported this finding as well.
The present study investigated numerous patients treated with CRT and diagnosed with histopathologically proven HNSCC. Furthermore, the study also investigated the correlations between PET and MRI-DWI parameters that were acquired simultaneously.
Via this approach, these parameters could be used to select a treatment strategy to address the higher SUV max , SUL peak , TLG, and MTV values that indicate a poorer treatment outcome. Therefore, it is worth taking the parameters suggested above into daily routine, especially the ones that significantly predict the patient outcome in daily routines to achieve more patienttailored therapy.
Several studies found negative associations between SUV max and ADC values (34,35). However, in our study, no significant linear correlations were found between the investigated parameters. Our results are similar to the results found by Rasmussen et al. Furthermore, when we classified the patients into two different groups based on the primary tumor degree of differentiation, no significant correlations were found. Contrary to a study by Wong et al., who reported that the ADC was a predictive factor to assess response to chemo-radiotherapy, we couldn't find a significant difference in the post-treatment ADC mean between the two groups, there was no noticeable difference in the ADC values (36).
The simultaneous imaging in PET/MR provides the same bed positions and acquisition at the same time, which leads to more accurate results compared to studies that have examined them separately on individual modalities. Compared to previous studies, this study finds that both parameters (SUV max , SUL peak , MTV, and TLG) had predictive values while using the single tracer injection double imaging acquisition protocol. In this research, the SUV max , SUL peak , MTV, and TLG values were measured; thus, their predictive value was discovered very first in homogeneously treated head and neck cancer patients.
In contrast, a few limitations must be acknowledged. The 1st weak point of this study was the retrospective analysis and the single-institute implementation. Moreover, a long-term follow-up might be more accurate to determine the therapeutic response. A multicenter and prospective study with more patients could be more representative of the population. Surov et al. found that more complex PET-and DWI-based parameters proved useful to reflect several histopathological parameters (37). However, our study included only the conventional parameters, which might be one of the limitations of this research.
Despite these limitations, this report provides important contributions to the field because it is the first study to show the predictive value of SUV max , SUL peak MTV, and TLG in patients with diagnostically confirmed HNSCC that were diagnosed with single tracer injection dual imaging acquisition. The usefulness of the 18 F-FDG PET/MR is important; nevertheless, it has questionable added value, because ADC mean has not shown significant differences between two patient groups (CR: n = 36 and non-CR: n = 32), probably due to the small number of patients (n = 68). In addition, this study also reported no correlations between PET and MRI-DWI based parameters. The findings suggest a need for further studies that involve more patients and more PET parameters as well as wider patient treatment modalities.

CONCLUSION
Pretreatment MRI-DWI values were unable to predict therapeutic response. However, 18 F-FDG PET parameters found to be more useful for this purpose in patients with HNSCC.
The strength of this study is the use of an MRI-DWI parameter, which includes diffusion evaluations that were collected simultaneously during PET/MR. SUV max , SUL peak MTV, and TLG values, significantly predicted the clinical outcome; thus their inclusion in risk stratification may be of additional value for predicting patient treatment outcomes.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/supplementary material, further inquiries can be directed to the corresponding author/s.

ETHICS STATEMENT
The studies involving human participants were reviewed and approved by Kaposi Mor Teaching Hospital ethics committee. The participants provided written informed consent to participate in this study.