High resolution diffusion weighted imaging for neoadjuvant chemotherapy assessment of nasopharyngeal carcinoma

Background To evaluate the feasibility of high resolution diffusion weighted imaging (DWI) for the prediction about the response of neoadjuvant chemotherapy (NAC) in patients. Methods 119 NPC patients who underwent magnetic resonance imaging (MRI) including readout-segmented diffusion weighted imaging (RESOLVE-DWI) before treatment enrolled in this study from September 2017 to December 2019, and all patients underwent follow-up MRI at 3 month after two cycles of NAC. Patients were classied into stable disease (SD), partial response (PR) or completed response (CR) based on Response Evaluation Criteria in Solid Tumors (RECIST). Apparent diffusion coecient (ADC) values were measured from RESOLVE-DWI in the rst MRI examination. The ADC values were compared between responders (CR and PR) and non-responders (SD) using Independent samples t-test, and the receiver operating characteristic curve (ROC) was calculated to assess the diagnostic ecacy.


Abstract Background
To evaluate the feasibility of high resolution diffusion weighted imaging (DWI) for the prediction about the response of neoadjuvant chemotherapy (NAC) in patients.
Methods 119 NPC patients who underwent magnetic resonance imaging (MRI) including readout-segmented diffusion weighted imaging (RESOLVE-DWI) before treatment enrolled in this study from September 2017 to December 2019, and all patients underwent follow-up MRI at 3 month after two cycles of NAC. Patients were classi ed into stable disease (SD), partial response (PR) or completed response (CR) based on Response Evaluation Criteria in Solid Tumors (RECIST). Apparent diffusion coe cient (ADC) values were measured from RESOLVE-DWI in the rst MRI examination. The ADC values were compared between responders (CR and PR) and non-responders (SD) using Independent samples t-test, and the receiver operating characteristic curve (ROC) was calculated to assess the diagnostic e cacy.

Conclusions
The ADC value obtained by RESOLVE-DWI, with a high diagnostic accuracy for NAC assessment, could be used as a pretreatment imaging biomarker to predict the early response to NAC in NPC patients.

Background
Nasopharyngeal carcinoma (NPC) is a common head and neck malignancy among Asians, especially in the southern provinces of China and Southeast Asia [1]. Neoadjuvant chemotherapy (NAC) followed by intensity modulated radiotherapy (IMRT) has gradually showed superiority in the treatment of patients with NPC [2][3][4][5], due to not only reduce the tumor size to optimize the IMRT plan [6] but also potentially protect the critical tissue and improve the quality of life of patients with NPC [7]. More importantly, the main reason of the failure to the treatment of NPC is distant metastasis [8] and the NAC could reduce the micro metastasis.
Magnetic resonance imaging (MRI) examination plays an important role in patients with NPC due to its excellent resolution [9], and diffusion-weighted imaging (DWI) has been widely applied to the NPC detection, staging, characterization, and treatment response prediction [10][11][12][13]. However, image quality of single-shot echo-planer imaging (SS-EPI) DWI is unsatisfactory due to the complex anatomic structure with a lot of cavity around the nasopharynx, which might cause motion and magnetic sensitive artifacts [14,15], leading to geometric distortion, signal intensity dropouts [16,17]. Based on segmenting kspace along the readout direction to shorten the echo spacing, the technique of readout-segmented echoplanar imaging diffusion weighted imaging (RESOLVE-DWI) exhibits reduced spatial distortion and improved image quality in the head and neck areas [16,18]. The apparent diffusion coe cient (ADC) maps from RESOLVE-DWI have a higher signal-to-noise ratio and resolution compared with conventional DWI, thus potentially improving the diagnostic and reliability of ADC value [19]. Therefore, this study assessed the ADC value obtained by RESOLVE-DWI for NAC's response in patients with NPC and explored valuable imaging biomarker to predict the local outcome of primary tumor.

Study Population
The ethics committee of Sun Yat-sen University Cancer Center approved this prospective study and all patients signed written informed consent before the rst treatment. We enrolled 122 newly diagnosed NPC patients with no prior treatment and referred to NAC from September 2017 to December 2019. However, we excluded three patients from the study for the following reasons: two patients were due to the small thickness, and one had inadequate image quality due to serious motion on MRI examination. Ultimately, the nal study population consisted 119 patients. All enrolled patients were candidates to receive cisplatin-based NAC for 2 cycles every 21 days, and then underwent follow-up MRI. In each cycle,

Images Analysis
The T2-weighted and DWI images were evaluated independently by three radiologists with more than 10 years experiences, and any differences were resolved by consensus. The morphological regression of tumors was measured according to T2-weighted images from the both MRI examinations. Meanwhile, at the rst MRI examination, three region of interests (ROIs) were drawn manually at the largest section of the tumor and its adjacent up/down sections on ADC maps referring to T2-weighted images, avoiding the hemorrhagic and necrotic region, and the ADC values obtained from the above three ROIs were averaged to obtained the nal results (Fig. 1).
The patients were classi ed into responders and non-responders based on Response Evaluation Criteria in Solid Tumors (RECIST) [20]. Responders were de ned as completed response (CR) and partially response (PR) patients, while non-responders were referred to as stable response (SD) or progressed on the follow-up MRI.
All patients' TNM stage were determined by radiologists with reference to the 8th edition of the International Union against Cancer/ American Joint Committee on Cancer (UICC/AJCC) manual.

Statistical Analysis
The consistency between observers was evaluated by intra-class correlation coe cient (ICC), and an ICC of 1.0 was considered to indicate perfect agreement; 0.81-0.99 indicated almost perfect agreement; 0.61-0.80, substantial agreement; 0.41-0.60, moderate agreement; 0.21-0.40, fair agreement; and ≤ 0.20, slight agreement [21]. The data of ADC values was assessed for normality with the Shapiro-Wilk test. The Independent sample t-test was used to compare the ADC values of responders and non-responders. Categorical or continuous variables (age, gender, T classi cation, N classi cation and clinical stage) were compared using the Chi-square test. Receiver operating characteristic (ROC) curves were plotted, and the area under the curve (AUC) and 95% con dence intervals (CI) were calculated to evaluate the diagnostic e ciency of the ADC values in distinguishing responders and non-responders. All of the analyses were performed using the SPSS 25.0 statistical software package (https://www.ibm.com/support/pages/node/589145), with a value of two-tailed p < 0.05 considered to be a signi cant difference.

Results
A total of 119 patients were enrolled in this study nally, including 87 (73.2%) males and 32 (26.8%) females with the median age 46 years (19-63 years). The primary tumor treatment responses were distributed as follows: responders, 93 (78.2%) patients; non-responders, 26 (21.8%) patients. Baseline characteristics of patients were shown in Table 1.  Fig. 2). The primary tumor ADC for differentiating responders and non-responders using ROC curve analysis was 624.45 × 10 − 3 mm 2 /s, and the AUC was 0.879 with a sensitivity of 82.8% and speci city of 84.6% (Fig. 3).

Discussion
Our study demonstrated that ADC values obtained from RESOVLE-DWI was able to predict tumor response to NAC treatment in patients with NPC, and there was a greater diagnostic accuracy to distinguish responders from non-responders after two NAC cycle treatments. The patients with higher pretreatment ADCs had a better response to NAC compared to those with lower pretreatment ADCs. This could promote the individualized treatment for patients with NPC and protect some patients from ineffective and unnecessary treatment toxicities.
As a useful MRI functional sequence, DWI provides functional and structural information on the microanatomy in tissues, which plays a signi cant role in routine MRI examination and serves as a good supplement to improve the accuracy of assessing the properties and biological behaviors of nasopharyngeal carcinoma. Meanwhile, ADC value derived from DWI has been shown to have potential as imaging bio-markers for predicting the histopathological grade and indicating the aggressiveness of nasopharyngeal carcinoma [12,22]. In this study, the mean ADC value of responders was signi cantly higher than that of non-responders, which were consistent with those of previous research [23,24]. However, Zhang et al [25] reported that the mean ADC value of responders was signi cantly lower than that of non-responders, the reason might be that the patients in their study were obtained three cycles NAC.
In addition, they used SS-EPI DWI sequence to obtain ADC values, which is challenging to perform ADC measurement in the head and neck due to susceptibility and motion artifacts. In the present study, the ADC values were obtained by RESOLVE-DWI. Based on a two-dimensional navigator reacquisition, "Readout Segmentation Of Long Variable Echo-trains" divides the k-space into multiple segments to shorten the echo spacing and accelerates the k-space traversal along the phase-encoding direction to diminish geometric distortion, signal dropout and image blurring [26][27][28]. RESOLVE technology can effectively reduce these artifacts and provide high resolution images,which can better display the tumor to delineate the target area, thus greatly improving the accuracy of ADC value. Zhao et al [29] reported that RESOLVE-DWI signi cantly improved the image quality for evaluations of Sino nasal lesions and offered more accurate ADC value than SS-EPI.
Based on ROC analysis, The ADC value had a better AUC and relatively high sensitivity of differentiation between responders and non-responders for NAC treatment of NPC, suggesting that ADC value obtained by RESOLVE-DWI may serve as a valuable imaging biomarker for NAC.
The present study has several limitations. First, a small sample size in this study, and most of the patients with patients responded to NAC, which could result in statistical bias. Second, this study is a single-center study, and the ndings should be validated in other institutions with diverse population.

Conclusions
In conclusion, the ADC value obtained by RESOLVE-DWI could serve as a valuable imaging biomarker to predict the response to NAC, and the higher pretreatment ADCs tended to response better for NAC.

Declarations
Ethics approval and consent to participate The ethics committee of Sun Yat-sen University Cancer Center approved this prospective study and all patients signed written informed consent before the rst treatment.

Consent for publication
Not applicable.

Availability of data and materials
The data are not available for public access because of patient privacy concerns but are available from the corresponding author on reasonable request approved by the institutional review boards of Sun Yatsen University Cancer Center.
MTB, KLR and XCM contributed to study design. MTB, LJ, WFX, DW, ZLL, MXG, HW and LH collected the study data. MTB, LJ and WFX contributed to data analysis. MTB, LJ and WFX contributed to manuscript writing. MTB, KLR and XCM contributed to quality control. All authors reviewed the manuscript and approved the nal version.  Box plot of ADC value for responders and non-responders. Abbreviation: ADC, apparent diffusion coe cient. The diagnostic accuracy of ADC value in predicting the induction chemotherapy response of nasopharyngeal carcinoma. Abbreviation: AUC, the area under the curve; CI, con dence intervals.