Seventy percent of salivary gland tumors (SGTs) are benign, which are mostly pleomorphic adenomas (PAs) and Warthin's tumors (WTs). The remaining are malignant tumors (MTs), most of which are carcinomas [1,2]. Most SGTs (80%) arise in the parotid glands [3]. The diagnosis of SGT should be confirmed by histopathological analysis of specimens obtained by ultrasound-guided fine-needle aspiration or core biopsy [4], while the main role of computed tomography (CT) or magnetic resonance imaging (MRI) is to map the location and extent of the tumor prior to surgical resection. However, there are circumstances in which imaging features are required to characterize SGT. This is true for deep tumors that are inaccessible to biopsy and for heterogeneous tumors, in which the sample may not accurately represent the true nature of the tumor [5]. The imperfect accuracy of fine-needle aspiration cytology due to issues related to sampling and challenges in cytological interpretation is well documented and a recent large meta-analysis showed that malignant SGTs may be missed or over-diagnosed using parotid fine-needle aspiration cytology [6]. Therefore, imaging can be helpful in characterization in terms of supporting or questioning the provisional cytological diagnosis. Furthermore, benign SGTs in the parotid gland are a common incidental imaging finding, such as when using MRI to stage other head and neck tumors. In these scenarios, it is helpful to indicate in the report whether the incidental SGT is likely to be benign or malignant.

MRI features that suggest malignancy include irregular tumor margins, local invasion into adjacent structures, and perineural tumor spread [7]. However, many malignant SGTs do not show these features or may have features that overlap with those of benign SGTs [8]. For example, irregular tumor margins are absent in 50% of malignant SGTs and present in 15% of benign ones [9,10]. Functional MRI has the benefit of providing information in addition to anatomical information, to aid the discrimination of benign and malignant tumor. In this regard, diffusion parameters obtained from two functional MRI techniques, conventional diffusion-weighted imaging (DWI) [10], [11], [12], [13], [14] and intravoxel incoherent motion (IVIM) [15], [16], [17], [18], [19], [20], have shown promise for the characterization of SGTs.

DWI provides estimates of the apparent diffusion coefficient (ADC) using a mono-exponential model [14]. IVIM uses a bi-exponential model that takes longer to acquire and involves more sophisticated analyses. It is less widely available, but it provides additional information (pure diffusion coefficient (D), pseudo-diffusion coefficient (D*) and perfusion volume fraction (f)) by separating pure diffusion from diffusion related to microcirculation (pseudo-diffusion) [20]. However, conflicting results have been reported when using these parameters to discriminate between different categories of SGTs. Specifically, although several studies have shown significant differences in ADC [12,21] or D values [16,17,20] between malignant and benign SGTs and significant differences in ADC [22], [23], [24] and D* [20] between WTs and malignant SGTs, other studies found some conflicting results [11,17,18,21,[24], [25], [26], [27], [28], [29], [30], [31]]. The reason behind the conflict in some results and differences in diagnostic performance is probably multifactorial but may include differences in tumor type prevalence between studies, inclusion of uncommon benign and malignant SGTs, inclusion or exclusion of cystic components and different methods to select the optimal threshold for clinical practice. Accordingly, the respective capabilities of DW-MRI and those of IVIM role for discriminating between PA, WT and carcinoma have not been fully elucidated.

The purpose of this study was to evaluate the diagnostic performances of DWI and those of IVIM for discriminating between benign and malignant SGT.