Dual-phase dual-energy computed tomography (DECT) in assessing recurrence-associated histopathological features of parotid pleomorphic adenoma

Introduction

Pleomorphic adenoma (PA), the most common benign salivary gland neoplasm, exhibits heterogeneous histology with variable epithelial-stromal proportions encapsulated in fibrous tissue. While surgical resection (e.g., extracapsular dissection or parotidectomy) is the primary treatment, recurrence remains a major postoperative concern due to the rate of 2.3–6.7% of cases over 5–20 years (1), often necessitating revision surgeries with heightened facial nerve injury risks (2). Critically, the risk of malignant transformation in PAs with multiple recurrences is significantly increased (1,3).

Risk factors of recurrence include certain histopathological characteristics and inappropriate surgical operation (4,5). Capsular features and stroma-rich subtype are primary histopathological determinants of recurrence (5-7). Stroma-rich subtypes [>60% stromal component, by Handa’s classification (8)] demonstrate the highest recurrence risk among all subtypes due to propensity for capsular invasion including incomplete capsules, pseudopodia, satellite nodules (6,9). This aligns with Schapher et al.’s findings where the stroma-rich subtype of PA accounted for three-quarters of the recurrent cases (10).

Preoperative risk stratification based on histopathological features could optimize resection extent, such as widening margins for high-risk lesions (9,11). Fine-needle aspiration biopsy is not routinely recommended for apparently benign parotid tumors due to concerns over tumor spillage and potential facial nerve injury, despite its role in evaluating suspected malignancies (12,13).

Preoperative imaging non-invasively evaluates parotid tumor morphology and textural features. While conventional computed tomography (CT) provides anatomical images with densitometric information, and magnetic resonance imaging (MRI) excels in soft-tissue contrast and functional imaging, dual-energy computed tomography (DECT) offers a unique paradigm for quantitative tissue characterization. This technique acquires simultaneous high- and low-energy data to enable material decomposition based on differential X-ray absorption. This provides functional insights that are not available from other techniques, such as iodine concentration (IC), effective atomic number (Z_eff), and the spectral attenuation slope (λ_HU). These parameters have shown high diagnostic performance in distinguishing parotid tumor types (14,15). However, their ability to identify recurrence-associated histopathological features in PAs, specifically stroma-rich composition and capsular invasion remains unexplored.

Therefore, this study aims to evaluate dual-phase contrast-enhanced DECT for predicting stroma-rich composition and capsular invasion in PAs, and establish predictive models integrating DECT parameters and clinical covariates. We present this article in accordance with the TRIPOD reporting checklist (available at https://qims.amegroups.com/article/view/10.21037/qims-2025-aw-2383/rc).

Methods

Study subjects

The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Ethics Committee of West China Hospital of Stomatology, Sichuan University (No. WCHSIRB-CT-2025-399), and individual consent for this retrospective analysis was waived.

We retrospectively enrolled consecutive patients diagnosed with parotid PA at West China Hospital of Stomatology, Sichuan University from June 2024 to August 2025.

Inclusion criteria included:

• Parotid PA diagnosed by postoperative pathology;
• Preoperative dual-phase contrast-enhanced dual-energy CT scans of the head and neck;

Exclusion criteria included:

• Significant artifacts affecting image quality;
• Lesion diameter <1 cm or poor visualization on imaging;
• Other parotid-related medical histories.

Patient demographics and clinical data were collected, including age, sex, and symptom duration, defined as the interval from patient-noticed mass to surgery.

CT scanning criteria

Dual-phase enhanced CT was performed on a fourth-generation dual-source CT scanner (SOMATOM Drive, Siemens Healthineers, Germany). Automatic exposure control was applied to all scans to optimize radiation dose and image quality. All patients were scanned in a head-first, supine position using a standardized protocol (160 mAs for the 80-kVp tube; 80 mAs for the Sn140-kVp tube; rotation time, 0.5 s; pitch, 0.7). Images were reconstructed with slice thickness of 1.0 mm and section increment of 0.7 mm. All scanning slices included complete parotid gland and the lesion levels, with non-contrast images acquired first, followed by dual-phase enhanced scanning. During contrast-enhanced scanning, non-ionic iodinated contrast agent (Ioversol, 350 mg I/mL, Jiangsu, China) was administered intravenously via the right or left median cubital vein at a dose of 1 mL/kg body weight with a flow rate of 1.8–2.0 mL/s. Contrast images were acquired at 40 seconds (early phase) and 120 seconds (late phase) after injection.

Imaging characteristics

All images were independently evaluated by two radiologists blinded to the histopathological results (R.H., radiologist with 6 years of experience; Y.L., radiologist with 16 years of experience). Standard blended images were reconstructed at a fusion ratio of 0.4 (40% low-energy: 60% high-energy) for assessing lesion location, morphology, margins, and Hounsfield unit (HU) values. Enhancement amplitude was defined as the difference in Hounsfield units (∆HU) between contrast-enhanced and non-contrast CT scans. ΔHU = post-enhanced HU – pre-enhanced HU.

Quantitative DECT parameters

Dual-energy parametric analyses were subsequently processed using the Syngo.via workstation (Siemens Healthineers).

Quantitative parameters, including Z_eff, λ_HU, and IC, were derived from dedicated dual-energy CT image sets using a dual-energy workflow module (Figure 1).

λHU=(CT40keV−CT100keV)÷(100−40)[1]

Figure 1 Post-processing of DECT images in PA. (A-D) A patient with stroma-rich PA. (A) Early-phase enhanced standard blended image; (B) measurement of Z_eff in the Rho-Z module; (C) measurement of IC in the virtual non-contrast module; (D) measurement of λ_Hu in the Mono + module. (E-H) A patient with non-stroma-rich PA. (E) Early-phase enhanced standard blended image; (F) measurement of Z_eff in the Rho-Z module; (G) measurement of IC in the virtual non-contrast module; (H) measurement of λ_Hu in the Mono+ module. The white arrows in panels (B), (C), (F), and (G) indicate the boundaries of the manually placed region of interest used for quantitative measurements. CM, contrast medium; CT, computed tomography; DECT, dual-energy computed tomography; HU, Hounsfield unit; IC, iodine concentration; PA, pleomorphic adenomas; ROI, region of interest; VNC, virtual non-contrast; Z_eff, effective atomic number; λ_Hu, spectral attenuation slope.

For each lesion, measurements were obtained from the region of interest (ROI) and a reference artery at the same anatomical level. The ipsilateral external carotid artery was the preferred reference due to its role as the primary feeding artery of the parotid gland; when suboptimally visualized, the internal carotid artery at the same level was used instead. Lesion parameters were normalized by division with corresponding values from the reference artery to minimize confounding effects from individual circulatory variations.

NormalizedZeff= LesionZeff÷ReferencearteryZeff[2]

NormalizedλHU=Lesion λHU÷ReferencearteryλHU[3]

NormalizedIC=LesionIC÷ReferencearteryIC[4]

Segmentation of tumor ROI

Quantitative measurements were obtained from three ROIs per lesion and averaged: ROI1: Largest cross-sectional area at the lesion center; ROI2: Midway between the superior margin and ROI1 slice; ROI3: Midway between the inferior margin and ROI1 slice. All ROIs were manually delineated along the tumor border (Figure 1). This aims to make the measurements representative of the tumor as a whole.

Histopathological features

All histopathological specimens were fixed in 10% neutral buffered formalin, sectioned at 2–3 mm thickness, dehydrated through graded ethanol and xylene, embedded in paraffin, and stained with hematoxylin and eosin (H&E). Depending on tumor size, 2–6 histopathological sections were prepared per tumor. All H&E-stained slides were reviewed twice in an independent and blinded manner (with respect to CT findings) by a single experienced pathologist. The quantitative measurements from the two readings were averaged to derive a final result for analysis.

According to the Handa et al. (8) subtyping criteria, tumors were classified as stroma-rich when the stroma component exceeded >60%. Cell-rich and mixed subtypes with ≤60% stromal content were categorized as non-stroma-rich. Capsular invasion was defined as tumor cell invasion into or breach of the fibrous capsule, including pseudopodia and satellite nodules (Figure 2). A pseudopodium was identified as a tumor protrusion or nodule extending beyond the main tumor mass but localized within the main capsule or in contact with it. Satellite nodules were defined as discrete tumor foci separated from the main tumor mass by salivary or fat tissue (6).

Figure 2 Pathological characteristics of PA. (A) Non-stroma-rich subtype; (B) stroma-rich subtype; (C) thin capsule (white arrows); (D) focal capsular invasion (black arrow); (E) pseudopodia; (F) satellite nodule. All images are HE staining. Scale bar =400 µm. HE, hematoxylin and eosin; PA, pleomorphic adenomas.

Statistical analysis

Statistical analyses were performed using IBM SPSS Statistics (version 26.0; IBM Corp., Armonk, NY, USA) and R software (version 4.2.2; R Foundation for Statistical Computing, Vienna, Austria). Inter-observer agreement was assessed using the intraclass correlation coefficient (ICC) for continuous variables. Bivariate analyses comparing intergroup differences were conducted with independent samples t-tests (continuous variables) and chi-square tests (categorical variables). All potential predictors were screened by univariable logistic regression, followed by an assessment of multicollinearity using the variance inflation factor (VIF). When significant predictors consisted solely of highly collinear quantitative parameters, we prioritized evaluating the individual diagnostic performance of each parameter using the receiver operating characteristic (ROC) curve analysis to avoid model instability. In contrast, for cases involving a mix of clinical and imaging predictors, a multivariable model was constructed via the least absolute shrinkage and selection operator (LASSO) regression and subsequently underwent calibration assessment using the Hosmer-Lemeshow test and calibration plots. The performance of all classifiers was evaluated by ROC analysis, with the optimal cut-off determined by Youden’s index and the scale inverted for protective factors (by multiplying by −1).

Results

Patient characteristics

A flow chart illustrating the participant inclusion process is displayed in Figure 3. A total of 108 patients were enrolled, comprising 39 males and 69 females. Histopathological analysis revealed 38 patients in stroma-rich group and 70 patients in non-stroma-rich group. Capsular invasion was present in 65 patients, while 43 patients had intact capsules. Analysis of histopathological characteristics demonstrated a significant correlation between stromal richness and capsular invasion (r**=0.305, P=0.003). Patients in the capsular invasion group were significantly younger (P=0.031) and had a shorter symptom duration (P=0.021) compared to the intact capsule group. No significant differences in age or symptom duration were observed between the stroma-rich and non-stroma-rich groups (P>0.5) (Table 1).

Figure 3 Flowchart of patient enrollment. DECT, dual-energy computed tomography; PA, pleomorphic adenoma.

Imaging characteristics

Lesions in the stroma-rich group were more frequently located in the deep lobe, although this difference did not reach statistical significance (P=0.073). There were no significant differences in lesion diameter, shape, or margin between the stroma-rich and non-stroma-rich groups. Lesions in the capsular invasion group were more likely to exhibit a lobulated contour compared to the intact capsule group (P=0.045). However, no significant differences were found in lesion diameter, location, or margin between the two capsular status groups (Table 1).

Quantitative DECT parameters

The inter-observer agreement for DECT parameters was excellent, with ICCs ranging from 0.91 to 0.96 (Table S1). Among the early-phase quantitative parameters, the stroma-rich group exhibited significantly lower values for the DECT parameters (nZ_eff, NIC, nλ_Hu) and ∆HU compared to the non-stroma-rich group (P=0.033–0.001). In the late phase, nλ_Hu was significantly lower in the stroma-rich group (P=0.041). The capsular invasion group showed significantly lower early-phase nλ_Hu (P=0.038) and late-phase nZ_eff (P=0.038) compared to the intact capsule group (Table 2).

Predictor selection and performance evaluation

Variable selection and multicollinearity assessment

Univariate analysis revealed distinct predictor profiles for the two outcomes (Table 3). Stromal abundance was significantly associated (P≤0.003) solely with quantitative DECT parameters from both early and late phases. In contrast, capsular invasion was associated (P<0.05) with a combination of clinical characteristics (age, symptom duration), an imaging feature (tumor shape), and quantitative DECT parameters (late-phase nZ_eff, nλ_HU, and ∆HU from both phases). Multicollinearity assessment indicated severe collinearity (VIF >10) among all the DECT parameters. Given their intrinsic homology, constructing a stable multivariable model for stromal richness was precluded. Consequently, the analysis for this outcome focused on comparing the individual discriminatory performance of these parameters using ROC analysis. For the capsular invasion outcome, the heterogeneity of the significant predictors enabled the application of LASSO regression for variable selection and the development of a parsimonious multivariable model. The model subsequently underwent calibration assessment (Figure 4).

Figure 4 Variable selection and performance assessment of the multivariable model for predicting capsular invasion. (A) Coefficient profile plot from the LASSO regression. (B) Ten-fold cross-validation curve for selecting the optimal lambda (λ) value in LASSO. (C)​ Calibration plot of the final multivariable model. The diagonal dashed line represents perfect calibration. GLM, generalized linear model; LASSO, least absolute shrinkage and selection operator.

Diagnostic performance of DECT parameters for stromal richness

Diagnostic performance metrics are presented in Table 4 and Figure 5. The overall highest area under the curve (AUC) was observed for the traditional quantitative parameter, early-phase ∆HU (0.804), which also demonstrated a balanced sensitivity (84.2%) and specificity (71.4%). Its late-phase counterpart also showed high performance (AUC =0.801). Among the DECT-drived quantitative parameters, the early-phase nλ_Hu achieved the highest AUC (0.762) and was characterized by high specificity (84.3%). The early-phase NIC showed the highest sensitivity (94.7%) among all parameters.

Figure 5 ROC curves of the quantitative DECT parameters for predicting stromal richness. (A) ROC curves of ∆HU in early and late phases; (B) ROC curves of DECT-derived parameters in early and late phases. ∆HU, difference in Hounsfield units; AUC, area under the curve; DECT, dual-energy computed tomography; NIC, normalized iodine concentration; nZ_eff, normalized effective atomic number; nλ_Hu, normalized spectral attenuation slope; ROC, receiver operating characteristic curve.

Predictive model for capsular invasion

LASSO regression selected age, symptom duration, tumor shape, ∆HU of early phase for the final multivariable capsular invasion model (Table 5). The model achieved an AUC of 0.774 [95% confidence interval (CI): 0.698–0.850] and demonstrated good calibration (Hosmer-Lemeshow test, P=0.168). At the optimal probability cut-off determined by Youden’s index, the model yielded a sensitivity of 56.9% and a specificity of 88.4%. The corresponding ROC curve of the model is presented in Figure 6 and Table 6.

Figure 6 ROC curve of the model for predicting capsular invasion. AUC, area under the curve; ROC, receiver operating characteristic.

Discussion

Our findings provide strong evidence supporting the utility of quantitative dual-phase contrast-enhanced DECT parameters in predicting recurrence-associated histopathological features of PA. DECT demonstrated superior capability in identifying the stroma-rich component, as evidenced by significantly lower early-phase parameter values in stroma-rich tumors compared to non-stroma-rich counterparts (P<0.033).

DECT parameters collectively exhibited favorable diagnostic accuracy. The overall highest AUC was observed for the traditional quantitative parameter, early-phase ∆HU (0.804). Clinically, a lower value of this readily obtainable parameter indicates a higher probability of the stroma-rich subtype, reflecting the tumor’s lower perfusion. The late-phase ∆HU also demonstrated a comparably high AUC (0.801). This finding indicates that the exact scan timing has a relatively minor impact on the performance of enhanced ∆HU, suggesting its potential for maintaining diagnostic robustness across institutions despite variations in imaging protocols. Among the DECT-derived parameters, the early-phase nλ_Hu achieved a high AUC (0.762) and was characterized by high specificity (0.843), while the early-phase NIC showed the highest sensitivity (0.947) among all parameters. These distinct performance profiles highlight the significant complementary value of DECT-drived parameters to conventional metrics.

The basis for this performance lies in the fundamental physics of DECT. DECT parameters exhibit pronounced sensitivity to high-atomic-number elements (16,17). Iodine, a key contrast agent, generates fundamentally distinct photon interactions compared to soft tissues due to its high atomic number (Z=53). This differential response enables the precise profiling of iodine distribution heterogeneity. While the conventional ∆HU parameter primarily reflects iodine concentration dynamics, parameters such as nZ_eff, NIC, and nλ_Hu provide distinct physical dimensionality based on atomic number sensitivity, thereby enhancing tissue characterization (18). The intrinsic collinearity among these parameters, arising from their shared dependence on the photoelectric effect and material decomposition, precluded the construction of a stable multivariable model for stromal richness. For the capsular invasion model, this collinearity necessitated the use of LASSO regression to ensure stable variable selection from the set of clinical and imaging predictors.

Iodine distribution in parotid gland tumors is associated with microvascular formation. Kim et al. documented significant correlations between microvascular density and CT enhancement patterns in PAs (19). Cell-rich regions typically exhibit richer angiogenesis owing to the release by neoplastic cells of growth factors specific for endothelial cells (20). Conversely, stroma-rich regions demonstrate uneven vascular distribution and lower vascular maturity, frequently accompanied by increased vascular permeability. The pronounced early-phase enhancement in cell-rich regions, contrasted with the slower wash-in and wash-out kinetics in stroma-rich regions, provides a potential mechanistic explanation for the superior performance of early-phase parameters over late-phase ones in distinguishing tumor component.

The capsular invasion model, achieving an AUC of 0.774, reflects the multifactorial nature of this histopathological outcome. It incorporates clinical features (age and symptom duration) alongside a key morphological imaging characteristic (tumor shape). Younger age is a well-documented risk factor for PA recurrence often linked to heightened proliferative activity (1,2,10). Similarly, a lobulated tumor contour may reflect locally aggressive growth patterns and has been associated with positive surgical margins (19). While symptom duration is a subjective measure, a shorter interval often triggers surgical intervention for rapidly growing or symptomatic masses, serving as a clinical indicator of potentially aggressive tumor behavior. The novel contribution of this model lies in augmenting these factors with a quantitative imaging biomarker, early-phase ∆HU, thereby creating an integrated clinical-radiologic predictive tool. This integration achieves a clinical profile of high specificity (88.4%), which could aid in informing surgical planning decisions.

Current surgical consensus recommends extracapsular dissection or partial superficial parotidectomy for superficial lobe tumors, with variable parotidectomy ranges for deep lobe involvement (4). While total parotidectomy with facial nerve dissection can reduce recurrence rates, it carries significant risks of complications including facial paralysis, Frey’s syndrome, and cosmetic deformities (10). This presents a clinical dilemma: balancing recurrence prevention against facial nerve preservation. Our study provides predictive indicators for objective, preoperative risk assessment based on histopathology. The stroma-rich parameters and capsular invasion model together facilitate the identification of high-recurrence-risk patients for whom definitive initial resection may be beneficial. The high specificity of both the early-phase nλ_Hu for stromal richness and the capsular invasion model gives them particular potential as reliable “rule-in” tools to support surgical planning.

The concept that imaging parameters can reflect parotid PA histopathology is well-supported. The foundational work of Kim et al. (19) using single-phase CT demonstrated a correlation between strong enhancement and high epithelial component proportion, providing an early link between imaging features and PA composition. Subsequently, the capabilities of MRI, particularly diffusion-weighted imaging (DWI), were further illustrated by Kilictas et al. (21), who confirmed that the apparent diffusion coefficient (ADC) could differentiate histopathological subtypes of parotid tumors. Monestier et al. (22) employed these DWI-derived ADC values as the most valuable biomarker of PA cellularity and was a highly specific predictor for the hypocellular subtype, with a specificity of 94%. This study advanced the field by developing quantitative predictive models for a high-risk histological subtype.

Our work builds on this foundation by leveraging DECT to identify complementary quantitative imaging parameters that reflect stromal richness in PA. Furthermore, we extend the predictive paradigm to include the critical feature of capsular invasion, thereby providing a more comprehensive preoperative risk assessment tool.

Several limitations of this study should be acknowledged. First, the single-center design and the use of a specific scanner platform may limit the generalizability of our predictive models. Although consecutive enrollment minimized selection bias, the applicability of our models to other centers with different patient populations and scanning parameters requires external validation. Second, and specific to model validation, internal validation (e.g., bootstrapping) was not performed. Given the sample size, this raises the possibility of over-optimism in the reported performance metrics, underscoring the need for validation in larger, independent cohorts before clinical application. Third, the histopathological assessment lacked validation by a second pathologist; the absence of an inter-observer agreement assessment may introduce potential bias. Furthermore, while correlations between DECT parameters and stromal abundance were established, histopathological quantification of microvascular density was not systematically evaluated, which represents an important direction for future mechanistic research. Finally, the use of patient-reported symptom duration, while a practical clinical metric, is inherently subject to recall bias and may influence model estimates. Future prospective studies with objectively documented timelines are warranted to refine this association.

Conclusions

This study establishes that quantitative DECT parameters enable non-invasive prediction of the stroma-rich subtype in PAs. The early-phase ∆HU shows potential as a single-parameter predictor for stromal richness. A multivariable model incorporating clinical and DECT features shows good performance for predicting capsular invasion. This DECT-based approach provides a valuable tool for recurrence risk assessment based on histopathological risk factors, which could inform personalized surgical planning to mitigate recurrence.

Acknowledgments

We would like to thank Fan Yang and Zhengrong Nie (Siemens Healthcare Ltd., Chengdu Branch) for their assistance with post-processing CT images.

Footnote

Reporting Checklist: The authors have completed the TRIPOD reporting checklist. Available at https://qims.amegroups.com/article/view/10.21037/qims-2025-aw-2383/rc

Data Sharing Statement: Available at https://qims.amegroups.com/article/view/10.21037/qims-2025-aw-2383/dss

Funding: This study was supported by the Hospital-Level Research Project of West China Hospital of Stomatology, Sichuan University (No. LCYJ-QN-202512).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://qims.amegroups.com/article/view/10.21037/qims-2025-aw-2383/coif). All authors declared that this study was supported by the Hospital-Level Research Project of West China Hospital of Stomatology, Sichuan University (No. LCYJ-QN-202512). All authors reports assistance with post-processing CT images provided by Fan Yang and Zhengrong Nie from Siemens Healthcare Ltd., Chengdu Branch. The authors have no other conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Ethics Committee of West China Hospital of Stomatology, Sichuan University (No. WCHSIRB-CT-2025-399), and individual consent for this retrospective analysis was waived.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Valstar MH, de Ridder M, van den Broek EC, Stuiver MM, van Dijk BAC, van Velthuysen MLF, Balm AJM, Smeele LE. Salivary gland pleomorphic adenoma in the Netherlands: A nationwide observational study of primary tumor incidence, malignant transformation, recurrence, and risk factors for recurrence. Oral Oncol 2017;66:93-9. [Crossref] [PubMed]
2. Wittekindt C, Streubel K, Arnold G, Stennert E, Guntinas-Lichius O. Recurrent pleomorphic adenoma of the parotid gland: analysis of 108 consecutive patients. Head Neck 2007;29:822-8. [Crossref] [PubMed]
3. Pei Y, Li W. Clinical parameters predictors of malignant transformation of recurrent parotid pleomorphic adenoma. Sci Rep 2023;13:4543. [Crossref] [PubMed]
4. Quer M, Vander Poorten V, Takes RP, Silver CE, Boedeker CC, de Bree R, Rinaldo A, Sanabria A, Shaha AR, Pujol A, Zbären P, Ferlito A. Surgical options in benign parotid tumors: a proposal for classification. Eur Arch Otorhinolaryngol 2017;274:3825-36. [Crossref] [PubMed]
5. Malard O, Thariat J, Cartier C, Chevalier D, Courtade-Saidi M, Uro-Coste E, Garrel R, Kennel T, Mogultay P, Tronche S, Varoquaux A, Righini CA, Vergez S, Fakhry N. Guidelines of the French Society of Otorhinolaryngology-Head and Neck Surgery (SFORL), part II: Management of recurrent pleomorphic adenoma of the parotid gland. Eur Ann Otorhinolaryngol Head Neck Dis 2021;138:45-9. [Crossref] [PubMed]
6. Zbären P, Stauffer E. Pleomorphic adenoma of the parotid gland: histopathologic analysis of the capsular characteristics of 218 tumors. Head Neck 2007;29:751-7. [Crossref] [PubMed]
7. Dulguerov P, Todic J, Pusztaszeri M, Alotaibi NH. Why Do Parotid Pleomorphic Adenomas Recur? A Systematic Review of Pathological and Surgical Variables. Front Surg 2017;4:26. [Crossref] [PubMed]
8. Handa U, Dhingra N, Chopra R, Mohan H. Pleomorphic adenoma: Cytologic variations and potential diagnostic pitfalls. Diagn Cytopathol 2009;37:11-5. [Crossref] [PubMed]
9. Webb AJ, Eveson JW. Pleomorphic adenomas of the major salivary glands: a study of the capsular form in relation to surgical management. Clin Otolaryngol Allied Sci 2001;26:134-42. [Crossref] [PubMed]
10. Schapher M, Koch M, Agaimy A, Goncalves M, Mantsopoulos K, Iro H. Parotid pleomorphic adenomas: Factors influencing surgical techniques, morbidity, and long-term outcome relative to the new ESGS classification in a retrospective study. J Craniomaxillofac Surg 2019;47:1356-62. [Crossref] [PubMed]
11. Wong DS. Frozen section during parotid surgery revisited: efficacy of its applications and changing trend of indications. Head Neck 2002;24:191-7. [Crossref] [PubMed]
12. Riad MA, Abdel-Rahman H, Ezzat WF, Adly A, Dessouky O, Shehata M. Variables related to recurrence of pleomorphic adenomas: outcome of parotid surgery in 182 cases. Laryngoscope 2011;121:1467-72. [Crossref] [PubMed]
13. Patel KR, Scognamiglio T, Kutler DI, Kuhel WI, Gromis J, Phillips CD, Cohen MA. Retrospective Assessment of the Utility of Imaging, Fine-Needle Aspiration Biopsy, and Intraoperative Frozen Section in the Management of Parotid Neoplasms: The Weill Cornell Medical College Experience. ORL J Otorhinolaryngol Relat Spec 2015;77:171-9. [Crossref] [PubMed]
14. Li L, Zhao Y, Luo D, Yang L, Hu L, Zhao X, Wang Y, Liu W. Diagnostic value of single-source dual-energy spectral computed tomography in differentiating parotid gland tumors: initial results. Quant Imaging Med Surg 2018;8:588-96. [Crossref] [PubMed]
15. Wang Y, Hu H, Ban X, Jiang Y, Su Y, Yang L, Shi G, Yang L, Han R, Duan X. Evaluation of Quantitative Dual-Energy Computed Tomography Parameters for Differentiation of Parotid Gland Tumors. Acad Radiol 2024;31:2027-38. [Crossref] [PubMed]
16. McCollough CH, Boedeker K, Cody D, Duan X, Flohr T, Halliburton SS, Hsieh J, Layman RR, Pelc NJ. Principles and applications of multienergy CT: Report of AAPM Task Group 291. Med Phys 2020;47:e881-912. [Crossref] [PubMed]
17. Rajiah P, Parakh A, Kay F, Baruah D, Kambadakone AR, Leng S. Update on Multienergy CT: Physics, Principles, and Applications. Radiographics 2020;40:1284-308. [Crossref] [PubMed]
18. Liu T, Hong G, Cai W. A comparative study of effective atomic number calculations for dual-energy CT. Med Phys 2021;48:5908-23. [Crossref] [PubMed]
19. Kim H, Kim SY, Kim YJ, Ko JM, Park MJ, Kim JH, Hah JH, Kwon TK, Kim KH, Sung MW. Correlation between computed tomography imaging and histopathology in pleomorphic adenoma of parotid gland. Auris Nasus Larynx 2018;45:783-90. [Crossref] [PubMed]
20. Ribatti D, Vacca A, Dammacco F. The role of the vascular phase in solid tumor growth: a historical review. Neoplasia 1999;1:293-302. [Crossref] [PubMed]
21. Kilictas AU, Erdivanlı OC, Birinci M, Beyazal Celiker F, Terzi S, Celiker M, Coskun ZO, Dursun E. Evaluation of parotid tumors using systemic inflammation markers and diffusion-weighted magnetic resonance images. Acta Otolaryngol 2025;145:957-65. [Crossref] [PubMed]
22. Monestier L, Del Grande J, Haddad R, Santini L, Michel J, Varoquaux A, Fakhry N. Correlation between MRI (DWI and DCE) and cellularity of parotid gland pleomorphic adenomas. Eur Arch Otorhinolaryngol 2024;281:2655-65. [Crossref] [PubMed]