**4. Discussion**

In this study, the head-to-head comparison of PET/MR with PET/CT in DTC patients after comprehensive treatment revealed that PET/MR was more accurate in determining recurrent/residual and metastatic lesions, both from a patient-based and a lesion-based perspective. PET/MR provided superior conspicuity, as shown in clarity score evaluation, making it more useful for the identification of lesions. Both malignant and benign lesions had a significantly higher SUVmax, making it essential to incorporate the morphological content provided by MR to strengthen diagnosis efficacy. Therefore, additional neck PET/MR may be recommended for the detection of recurrence/lymph nodes metastases in DTC patients after comprehensive treatment, as it provides clearer images, more accurate identification, and more precise resection scope to avoid over-resection.

PET/MR can identify recurrent lesions and metastatic lymph nodes more accurately. According to patient analysis, PET/MR detected 26 of 27 patients with tumor burden, while PET/CT only detected 21. Based on lesion analysis, PET/MR combined with the advantages of MR had higher detection rates (91.5% vs. 80.8%), image conspicuity (especially in level II), and diagnostic efficacy than PET/CT. It was found that PET/MR showed higher diagnostic efficacy, and the MR part of PET/MR showed different positive results compared to 18F-FDG PET/CT, which suggested that the combination of the two modalities had a synergistic effect. This is consistent with published research [16,35]. However, there are still false positives for both imaging modalities, which are mainly caused by local postoperative hyperplasia. Based on the available pathological diagnosis in this study, suspicious areas depicted on PET/MR or PET/CT were fibrofatty tissue closely associated with postoperative local hyperplasia. In addition, high physiological 18F-FDG uptake can be observed in Waldeyer's ring, active muscles (vocal cord movement, swallowing), salivary glands, and brown fat [36].

MR plays an important role in the detection of complex thyroid lesions and metastatic lymph nodes [18–20] not only because MR has superior soft tissue contrast and additional imaging techniques, but also because multi-position, multi-parameter, and multi-sequence images can provide several benefits. These benefits include clear lesion contour and location, signal alteration and enhancement, diffusion restriction, and accurate assessment of lesion invasion of surrounding tissue structures (e.g., envelope, cartilage) [37]. On conventional MR, malignant thyroid nodules exhibit irregular margins and blurred boundaries with cystic changes, diffuse enlargement of the thyroid gland, and peripheral and distant invasion. They usually reveal low to intermediate T1- and high T2-weighted signal intensities, or occasionally high T1- and T2-weighted signal intensities with heterogeneous enhancement on enhanced T1WI [38]. On DWI, recurrence or metastases present strong contrast with the dark background signal from fat deposited around, based on metabolic or physiological changes [39,40]. Additionally, one should note that the clear volumetric information provided by MRI may be meaningful for determining the treatment dose [41]. In regard to image conspicuity, PET/MR had obvious advantages over PET/CT in detecting and displaying thyroid nodules as well as suspicious lymph nodes regardless of the region of interest. Compared with PET/CT, high resolution and multiple sequences such as DWI and ADC of MR have made contributions to the imaging of lesions, especially in level II with complex anatomy and high incidence of nonspecific lymph nodes [42]. Of note, the updated NCCN Guidelines of Thyroid Carcinoma (Version 3.2020) recommend that if iodine-131 imaging is negative and stimulated Tg > 2–5 ng/mL, additional non-radioiodine imaging modalities should be considered, including central and lateral neck compartments ultrasound, neck CT with contrast, chest CT with contrast, and PET/CT. However, enhanced CT is not routinely performed in patients with highly suspected recurrence or metastases in our department, in order to avoid the influence of the iodine contrast agent on the subsequent 131I treatment.

18F-FDG PET/CT is mainly used for postoperative surveillance of poorly differentiated thyroid cancer [43]. The uptake ability of 18F-FDG by thyroid cancer depends on the tumor differentiation degree [44,45]. The lower the differentiation of the lesion, the less it takes up 131I, but the more it takes up 18F-FDG, so positive Tg and negative 131I-WBS lesions are prone to be identified. This study also verified that SUVmax associated with 18F-FDG uptake was meaningful for the characterization of cervical lymph nodes. For PET alone, modern PET/MR can provide a superior SUVs than PET/CT, with time-of-flight (TOF) technology based on a new generation of crystals and the latest photovoltaic conversion technology.

There are contradictory results on the value of PET/CT and additional MR according to previous research. On the one hand, Hempel et al. concluded that the combination of PET/CT with MR was not suitable for routine clinical application [16,46]. Loeffelbein et al. reported that there was no significant difference between PET, MRI, and PET + MR in identifying recurrences in the neck [47]. On the other hand, earlier research argued that PET/CT and additional MR can provide additional diagnostic information [48,49]. PET/MR allows accurate, temporally and spatially unified multiparametric imaging of PET and MR. It can provide additional information and further improve the diagnostic efficacy than non-simultaneous PET and MR [50]. Previous studies have reported that PET/MR did not outperform PET/CT or even slightly underperform PET/CT in the detection of lung metastases and bone metastases [26,41]. Our results reveal that PET/MR is comparable to PET/CT in terms of diagnostic specificity. In addition, for the same lesion, SUVmax, SUVmean, and diameters measured on PET/MR and PET/CT were consistent and had significant correlation. Although there were significant differences between the measured parameters of PET/MR and PET/CT, this may be related to different tracer elimination, attenuation correction, and associated scatter correction.

There are several limitations in this study. First, the number of patients and lesions involved was limited. Most lesions were determined by a follow-up combination of Tg levels and imaging (ultrasound, CT, and/or MR) instead of surgical biopsy or FNAB; therefore, the absence of histopathological confirmation of most lesions may potentially affect the final diagnostic accuracy. Second, quantitative comparisons of parameters were limited by different imaging techniques, different attenuation correction and associated scatter correction methods, as well as different tracer elimination caused by the singleinjection double-examination protocol. In addition, the study was retrospective and only included patients with DTC who underwent both PET/CT and PET/MR. Multicenter clinical trials that prospectively include patients are necessary to assess the validity of our results. Furthermore, cervical lymph nodes are often inflammatory, which may lead to nonspecific 18F-FDG uptake. Therefore, the exploitation of specific imaging agents should be the orientation of future research. In this study, although 18F-FDG uptake differed between malignant and benign disease, there was also overlap in both SUV values, which may have an impact on the accuracy of the results. Third and lastly, PET/MR examinations are expensive, and the clinical application of this study should also consider the overall

benefit to patients in relation to financial expenditures. Further analysis of the economic benefits is expected.
