Histogram analysis of dynamic contrast-enhanced magnetic resonance imaging for differentiating malignant from benign orbital lymphproliferative disorders

Abstract

Background

Dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) has been used for assessing orbital lymphoproliferative disorders (OLPDs). However, only the mean values of quantitative parameters were obtained in previous studies and tumor heterogeneity was ignored.

Purpose

To assess the value of DCE-MRI derived histogram parameters in differentiating malignant from benign OLPDs.

Material and Methods

Forty-eight OLPDs patients (25 malignant and 23 benign) who had undergone DCE-MRI for pre-treatment evaluation were retrospectively included. Histogram parameters of K^trans, k_ep, and v_e were calculated and compared between two groups using the independent sample’s t-test. Receiver operating characteristic (ROC) curve analyses were used to determine the diagnostic value of each significant parameter. Multivariate stepwise logistic regression analysis was used to identify the independent predictors of malignant OLPDs.

Results

Tenth k_ep, mean k_ep, median k_ep, and 90th k_ep were significantly higher in the malignant OLPD group than in the benign OLPD group. Tenth v_e was significantly lower in the malignant OLPD group than in the benign OLPD group. Ninetieth k_ep was the only independent predictor of malignant OLPDs (P = 0.019), with an area under ROC curve of 0.828, a sensitivity of 92.00%, and a specificity of 78.26% at a cut-off value of 1.057 min⁻¹.

Conclusion

Histogram analysis of DCE-MRI derived parameters may help to differentiate malignant from benign OLPDs. The 90th k_ep hold the potential as an independent predictor for malignant OLPDs.

Keywords

Orbit lymphoproliferative disorders differential diagnosis magnetic resonance imaging perfusion

Introduction

Orbital lymphoproliferative disorders (OLPDs) represent a broad spectrum of benign and malignant diseases, including lymphoid hyperplasia, atypical lymphoid hyperplasia, ocular adnexal lymphoma, idiopathic inflammatory pseudotumor, and IgG4-related ophthalmic disease (1). Low-dose radiation therapy is considered the first-line treatment for orbital lymphoma, while benign OLPDs require corticosteroid therapy (2,3). Therefore, accurate prospective differentiation of benign and malignant OLPDs is highly clinically important. The value of the clinical information is limited because of the notable overlap between two entities.

Conventional magnetic resonance imaging (MRI) plays a very important role on the delineation of orbital tumors in clinical practice (4). Despite Haradome et al. reporting that “flow void sign” was more associated with benign OLPDs, the imaging features on conventional MRI were usually non-specific for the differentiation (1). Besides that, dependence on the experience of the reader restricts the wide application of conventional MRI features. Therefore, quantitative MRI techniques which allow for less dependence on the expertise of the reader may offer greater predictive value in the differential diagnosis of OLPDs.

Recently, dynamic contrast-enhanced MRI (DCE-MRI) which provides non-invasive assessment of tumor perfusion and vascularity, has been widely applied in the field of orbital imaging (2 –8). Previously, Hu et al. used DCE-MRI to differentiate malignant from benign OLPDs; they found that v_e (volume fraction of the extracellular extravascular space) and k_ep (reflux rate constant from extracellular extravascular space to blood plasma) might be potential imaging biomarkers for differentiating malignant from benign OLPDs (6). However, only the mean value of the pharmacokinetic parameters was used as the differentiating metric, which would underestimate the heterogeneity of the tumors. Heterogeneity is an intrinsic characteristic of tumors, which makes tumor tissue different from normal tissue (9). Previous studies have indicated that assessment of tumor heterogeneity was very crucial for tumor diagnosis, prognosis prediction, and treatment monitor (9 –11).

Histogram analysis is a widely used tool for heterogeneity quantification, particularly for MRI features, which are not always well visualized with naked eye (11,12). Histogram analysis of DCE-MRI has been proven to be more effective in differentiating primary central nervous system lymphoma and atypical glioblastoma (GBM), predicting the progression of enhancing lesions in GBM patients after standard treatment and detecting local tumor recurrence after definitive treatment of head and neck squamous cell carcinoma (13 –16). In the field of orbital imaging, histogram analysis of apparent diffusion coefficient (ADC) maps has been used to demonstrate the tumor heterogeneity and has been proven to be promising for improving the performance of diffusion-weighted imaging (DWI) for differentiating malignant from benign orbital tumor (17). However, to the best of our knowledge, no study has applied histogram analysis of DCE-MRI in the differentiation of benign and malignant OPLDs till now.

Therefore, the purpose of this study was to assess the performance of DCE-MRI in differentiating malignant from benign OPLDs by using histogram analysis approach.

Material and Methods

Study participants

This retrospective study was approved by our institutional review board and informed consent was waived. Between March 2013 and May 2017, a total of 71 consecutive OLPDs patients underwent orbital MRI examination for pre-treatment evaluation in our hospital. Twenty-three patients were excluded for the following reasons: (i) DCE-MRI was not scanned (n = 15); (ii) the image quality was not adequate for further analysis (n = 1); (iii) the diameter of the lesion was < 1 cm (n = 2); (iv) secondary lymphoma (n = 2); (v) the patient had a prior history of corticosteroid or radiation therapy before MRI scan (n = 3). Finally, 48 consecutive patients (32 men, 16 women; mean age = 56.4 ± 15.9 years) were included in our study.

The patients and tumor characteristics for the two groups are summarized in Table 1. Malignant group (n = 25; 17 men, eight women; mean age = 61.7 ± 12.8 years) included 21 patients with mucosa-associated lymphoid tissue (MALT) lymphoma, two patients with diffuse large B-cell lymphoma, and two patients with follicular lymphoma. All these 25 patients in the malignant group suffered from localized lymphoma lesion within the orbit, without involvement of another organs. Therefore, they should be staged as clinical stage I according to the AJCC staging system. Benign group (n = 23; 15 men, eight women; mean age = 50.6 ± 17.2 years) included 11 patients with idiopathic inflammatory pseudotumor, eight patients with reactive lymphoid hyperplasia, and four patients with IgG4-related ophthalmic disease. The final diagnosis was made based on the surgical pathological results in 42 patients and at the follow-up after steroid treatment in six patients with inflammatory pseudotumor.

Table 1.

Clinical characteristics of the patients.

Parameters	Benign OLPDs group (n = 23)	Malignant OLPDs group (n = 25)	P value
Mean age (years)	50.6 ± 17.2	61.7 ± 12.8	0.014
Gender (M/F)	15/8	17/8	0.838
Histologic subtypes	RLH (8)	MALT lymphoma (21)
	IIP (11)	DLBCL (2)
	IgG4-related disease (4)	Follicular lymphoma (2)

Data in parentheses indicates the number of the corresponding patients in our study.

OLPD, orbital lymphoproliferative disorder; M, male; F, female; MALT, mucosa-associated lymphoid tissue; DLBCL, diffuse large B-cell lymphoma; IIP, idiopathic inflammatory pseudotumor; RLH, reactive lymphoid hyperplasia.

MRI protocol

All patients were examined using a 3.0-T MRI system (Verio; Siemens, Germany) with a 12-channel head coil. Conventional MR sequences included unenhanced axial T1-weighted (T1W) imaging (TR/TE = 600/10 ms, slice thickness = 3 mm, matrix = 384 × 292, field of view [FOV] = 19 × 19 cm), axial T2-weighted (T2W) imaging (TR/TE = 4700/79 ms, slice thickness = 3 mm, matrix =384 × 292, FOV = 19 × 19 cm) with fat saturation, and coronal T2W imaging (TR/TE = 3500/79 ms, slice thickness = 3.5 mm, matrix = 384 × 307, FOV = 20 × 20 cm) with fat saturation.

DCE-MRI was performed using a two-dimensional (2D) turbo fast low angle shot (FLASH) sequence with integrated parallel acquisition technique (iPAT) in the axial plane. Before the injection of contrast material, pre-contrast T1W sequences (flip angle [FA] = 5°) was completed, which allows conversion of the changes of MR signal intensity into those of the gadolinium concentration during passage of the contrast agent. Gadolinium-diethylene triamine pentaacetic acid (Magnevist; Bayer Schering Pharma AG, Berlin, Germany) was then intravenously bolus injected via a power injector at the rate of 4 mL/s and a dose of 0.1 mmol/kg, followed by a flush of 20 mL of normal saline. DCE-MRI consisted of five baseline sets and 90 contrast-enhanced sets of images (total = 95 dynamics) with a temporal resolution of 3.3 s, and completed over a period of 5 min 15 s. Other detailed imaging parameters for the DCE-MRI were as follows: TR/TE = 474.66/1.43 ms; FA = 12°, average = 1, FOV = 230 mm, matrix = 128 ×128, section thickness = 4.5 mm, number of sections = 7. Axial, coronal, and sagittal T1W images were collected after DCE-MRI.

Image data analysis

All DCE-MRI data were analyzed by using a dedicated image processing software (OmniKinetics; GE Healthcare) (6). Pharmacokinetic parameters (the volume transfer constant between the plasma and the extracellular extravascular space [EES], K^trans; the volume fraction of the EES, v_e; and the reflux rate constant from EES to blood plasma, k_ep) were calculated by using the extended Tofts model (18). In terms of the arterial input function (AIF), it was extracted by manually drawing a small circle region of interest (ROI) on the ipsilateral internal carotid artery that located proximal to the tumor.

Two radiologists (with six and four years of experience in head and neck radiology, respectively) independently drew the ROIs manually. They were blinded to the clinical information, pathological results and study design. ROIs were drawn on all the slices those encompassed as much as tumor area. With conventional T2W and contrast-enhanced T1W images used as reference, visual necrotic, hemorrhagic areas, and surrounding blood vessels were excluded. In case of bilateral lesions, the lesion with the larger diameter was selected for analysis.

After the ROIs were drawn on each slice of the tumor foci, they were summated to derive the voxel-by-voxel values for histogram analysis. The histograms were plotted with the parametric values on the x-axis. The relative frequency was plotted on the y-axis of the histograms as a percentage of the total tumor volume. Histogram parameters consisting of mean, median, 10th percentiles, 90th percentiles, skewness, and kurtosis of K^trans, v_e, and k_ep were calculated based on DCE–MRI parameter maps. The nth percentile point is the point at which n% of the voxel values that form the histogram were found to the left of the histogram (14). The average of the two measurement results of two readers was used for further statistical analysis.

Statistical analyses

Each parameter was assessed for normality using the Kolmogorov–Smirnov test. The distribution of patient gender between benign and malignant OLPDs groups was compared using the chi-square test. The patient age and DCE-MRI histogram parameters between the two groups were compared using the independent samples t-test. Receiver operating characteristic (ROC) curves were constructed to determine the diagnostic performance of each significant pharmacokinetic parameter for differentiating malignant from benign OLPDs at its optimum threshold. Then, clinical characteristics and quantitative parameters which were shown to have univariate association with malignant OLPDs were included into a multivariable stepwise logistic regression analysis to identify the independent predictors of malignant OLPDs.

Inter-observer agreement for pharmacokinetic parameters was assessed using the two-way intraclass correlation coefficient (ICC). ICCs were interpreted as follows: 0.00-0.20 = poor agreement; 0.21–0.40 = fair agreement; 0.41–0.60 = moderate agreement; 0.61–0.80 = strong agreement; and 0.81–1.00 = almost perfect agreement (3). All statistical analyses were performed by using the statistical software MedCalc (version 11.0, MedCalc, Mariakerke, Belgium). A two-sided P value < 0.05 was considered statistically significant.

Results

A significant difference was observed in patient age between two groups (P = 0.014), while no difference on the distribution of patient gender (P = 0.838) (Table 1).

DCE-MRI derived histogram parameters in both groups were summarized in Table 2. Inter-observer agreement in the analysis of DCE-MRI was perfect for all pharmacokinetic parameters (ICCs, range = 0.977–0.988) (Table 2). For k_ep related histogram parameters, 10th k_ep (P = 0.001), mean k_ep (P = 0.004), median k_ep (P = 0.005) and 90th k_ep (P < 0.001) were significantly higher in the malignant OLPDs group than in the benign OLPDs group, while no significant differences were found on kurtosis (P = 0.061) and skewness (P = 0.676). For v_e-related histogram parameters, 10th v_e was significantly lower in the malignant OLPDs group than in the benign OLPDs group (P = 0.002), while no significant differences were found on mean v_e (P = 0.063), median v_e (P = 0.063), 90th v_e (P = 0.140), kurtosis (P = 0.353), and skewness (P = 0.934). For K^trans, no significant differences were found on any K^trans-related histogram parameters between two groups (all P values > 0.05) (Fig. 1).

Table 2.

Histogram parameters of the benign and malignant OLPD groups.

Parameters	Benign OLPD group (n = 23)	Malignant OLPD group (n = 25)	P value	ICC
Histogram of K^trans
10th K^trans	0.171 ± 0.071	0.220 ± 0.222	0.318	0.977
Mean K^trans	0.272 ± 0.116	0.343 ± 0.324	0.329	0.981
Median K^trans	0.258 ± 0.107	0.333 ± 0.311	0.276	0.980
90th K^trans	0.404 ± 0.204	0.474 ± 0.459	0.504	0.978
Kurtosis	5.472 ± 6.220	4.450 ± 4.554	0.517	0.984
Skewness	0.683 ± 1.245	0.579 ± 1.141	0.752	0.982
Histogram of v_e
10th v_e	0.268 ± 0.113	0.173 ± 0.090	0.002	0.982
Mean v_e	0.329 ± 0.164	0.246 ± 0.138	0.063	0.983
Median v_e	0.325 ± 0.177	0.238 ± 0.136	0.063	0.984
90th v_e	0.417 ± 0.187	0.331 ± 0.209	0.140	0.981
Kurtosis	3.987 ± 3.222	4.779 ± 2.619	0.353	0.980
Skewness	0.587 ± 0.692	0.623 ± 1.954	0.934	0.985
Histogram of k_ep
10th k_ep	0.532 ± 0.249	0.821 ± 0.320	0.001	0.985
Mean k_ep	0.873 ± 0.419	1.376 ± 0.682	0.004	0.982
Median k_ep	0.815 ± 0.396	1.327 ± 0.745	0.005	0.988
90th k_ep	1.141 ± 0.571	2.015 ± 0.982	<0.001	0.981
Kurtosis	3.587 ± 1.471	4.627 ± 2.177	0.061	0.979
Skewness	0.398 ± 0.881	0.543 ± 1.427	0.676	0.978

Except for P vales, data are reported as mean ± standard deviation. Unit for K^trans and k_ep is × min⁻¹.

OLPD, orbital lymphoproliferative disorder.

Fig. 1.

Boxplots for (a) K^trans, (b) k_ep, and (c) v_e histogram parameters between malignant and benign OLPDs groups. * indicates the difference between the two groups is significant.

Table 3 summarized the diagnostic performance of the 10th k_ep, mean k_ep, median k_ep, 90th k_ep, and 10th v_e for discriminating malignant from benign OLPDs. With a cut-off value of 1.057 min⁻¹, 90th k_ep showed the optimal diagnostic performance (area under the curve = 0.828, sensitivity = 92.00%, specificity = 78.26%) (Fig. 2). The maps for the representative cases of malignant and benign OLPDs are shown in Fig. 3.

Table 3.

ROC curve analyses of the significant histogram parameters for differentiating malignant from benign OLPDs.

Parameters	Cut-off value	Area under curve	Sensitivity (%)	Specificity (%)
Mean k_ep	≥ 0.842	0.767 (0.623–0.911)	88.00 (68.8–97.5)	73.91 (51.6–89.8)
Median k_ep	≥ 0.851	0.786 (0.647–0.926)	84.00 (63.9–95.5)	78.26 (56.3–92.5)
10th k_ep	≥ 0.604	0.790 (0.654–0.925)	84.00 (59.3–95.5)	78.26 (56.3–92.5)
90th k_ep	≥ 1.057	0.828 (0.701–0.955)	92.00 (74.0–99.9)	78.26 (56.3–92.5)
10th v_e	≥ 0.244	0.779 (0.639–0.919)	84.00 (63.9–95.5)	73.91 (51.6–89.8)

Data in parentheses are 95% confidence intervals. Unit for K^trans and k_ep is × min⁻¹.

OLPD, orbital lymphoproliferative disorder.

Fig. 2.

ROC curves of five significant histogram parameters for differentiating malignant from benign OLPDs.

Fig. 3.

Representative MR images and histogram maps in patients with MALT lymphoma (a–d) and reactive lymphoid hyperplasia (e–h). Compared with the adjacent muscles, both masses showed relatively high signal intensity on T2W images (a, e) and markedly enhancement on contrast-enhanced T1W images (b, f). As the colored k_ep maps (third column) showed, lymphoma (c) demonstrated significantly higher k_ep values than reactive lymphoid hyperplasia (g). Meanwhile, histogram distribution of k_ep maps showed higher k_ep values and frequency in the patient with lymphoma (d) than that in the patient with reactive lymphoid hyperplasia (h).

We included age, 10th k_ep, mean k_ep, median k_ep, 90th k_ep, and 10th v_e into the multivariate stepwise logistic regression analysis. The results indicated that 90th k_ep was the only independent differentiating variable for malignant OLPDs (P = 0.019) (Table 4).

Table 4.

Multivariable stepwise logistic regression analysis results for identifying the independent predictors of malignant OLPDs.

Parameters	B	SE	Wald	P value
Mean k_ep.	–8.847	5.845	2.291	0.130
Median k_ep.	–0.402	3.929	0.010	0.919
10th k_ep	5.510	3.774	2.132	0.144
90th k_ep	5.301	2.253	5.534	0.019
10th v_e	–8.618	5.486	2.467	0.116
Age	0.050	0.030	2.711	0.100

OLPD, orbital lymphoproliferative disorder; SE, standard error.

Discussion

In this study, we investigated the predicting value of DCE-MRI derived histogram parameters in differentiating malignant from benign OLPDs. Among various pharmacokinetic parameters, 10th k_ep, mean k_ep, median k_ep, and 90th k_ep were significantly higher in the malignant OLPDs group than in the benign OLPDs group, while 10th v_e were significantly lower in the malignant OLPDs group. In the further analysis using multivariate stepwise logistic regression analysis, 90th k_ep was found to be the only independent differentiating variable for malignant OLPDs.

K^trans is a pharmacokinetic parameter well-known for reflecting vascular perfusion and permeability (19). Previously, Ro et al. applied DCE-MRI for differentiating malignant from benign orbital tumors (7). They reported that K^trans was significantly higher in malignant masses than in benign masses. For satisfying tumor growth, malignant tumors were usually associated with a higher proportion of immature and hyper-permeable vessels, and therefore they would show higher K^trans than benign mimics (20). However, similar results were not seen in our study; we speculated that this conflicting result was because we only studied OLPDs. Previously, Haradome et al. reported that benign OLPDs were also characterized by high vascularization and its contrast enhancement ratio was even higher than that of lymphoma (1). Therefore, it was not surprising that the difference of K^trans did not reach significant between malignant and benign OLPDs groups.

v_e is a quantitative metric reflecting the volume of EES on DCE-MRI (21). Previously, Hu et al. reported that malignant OLPDs showed significantly lower v_e than benign mimics (6). They explained that interstitial edema within benign OLPDs would broaden the extracellular space, while high cellularity within lymphoma would limit the extracellular space; therefore, v_e would differ significantly between two groups. In our study, malignant OLPDs also showed relatively lower v_e-related histogram parameters than benign mimics. However, only the difference of the 10th v_e reached significance, while that of the 90th, mean, and median v_e did not. ADC was commonly viewed as an imaging metric that inversely associated with tumor cellularity (22). However, previous two studies reported that v_e was not related to ADC in GBM and breast cancer (23,24). Therefore, based on these findings, we thought that tumor cellularity was indeed a crucial factor affecting v_e, otherwise v_e would not demonstrate such-like change trend. However, cellularity might not be the sole factor affecting v_e, future radio-pathological correlation study is needed to reveal the more accurate physiological meaning of v_e.

k_ep is the reflux rate constant from EES to blood plasma (19). Previously, Ro et al reported that k_ep was significantly higher in malignant orbital masses (7). Similarly, Hu et al reported that malignant OLPDs showed higher k_ep than benign OLPDs, as seen in our study (6). Both K^trans and k_ep were quantitative parameters associated with vascular permeability, why did k_ep differ between the two groups while K^trans did not? In our opinion, it could be explained using prior findings. As we explained above, due to the high vascularization, both malignant and benign OLPDs showed high K^traⁿ, and no difference was found. However, the EES of malignant OLPDs was smaller than that of benign OLPDs (6). Therefore, within the same examination time, the contrast in the malignant tumors would begin to wash out from the EES to the draining vein earlier than the benign mimics, thus k_ep would differ significantly between two groups.

Heterogeneity is an intrinsic characteristic of tumors; accurate assessment of tumor heterogeneity was of high clinical importance for tumor diagnosis, prognosis prediction, and treatment monitor (9 –11). In our study, compared with mean k_ep, 90th k_ep showed better diagnostic efficiency and might be the independent differentiating variable for malignant OLPDs. Improved differentiating performance achieved by histogram analysis of DCE-MRI might lead to a more targeted diagnostic and therapeutic approach. Therefore, our results suggested that histogram analysis of DCE-MRI could effectively delineate the tumor heterogeneity and serve as a useful adjunctive diagnostic method for assessing OLPDs.

Our study had several limitations. First, it was a retrospective study with a relatively small number of patients. Further study with more patients is warranted to strength the statistical power and confirm our results. Second, ROIs placements were performed in a manual pattern, which was prone to measurement variability. A semi-automatic lesion segmentation can provide a reduction in variability and improvement in reproducibility (25). Third, the malignant group was dominated by the relative indolent MALT lymphoma. Future studies should enroll more lymphoma types and subgroup analysis among different lymphoma types would be more valuable. Fourth, DCE-MRI protocol and data analysis have not been generalized and standardized. Therefore, the specific cut-off value derived in our study could not be generally used in other study groups. Finally, some other advanced MRI techniques had also been proven to be useful for diagnosing orbital lesions, such as DWI (26,27). A multi-parametric model of advanced MRI techniques would be more promising for solving this clinical problem.

In conclusion, the present study demonstrated that histogram parameters derived from DCE-MRI, including 10th k_ep, mean k_ep, median k_ep, 90th k_ep, and 10th v_e, may help to differentiate malignant from benign OLPDs. In particular, the 90th k_ep holds potential as a candidate imaging biomarker for differentiation in future prospective studies.

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by Jiangsu Province’s Young Medical Talents Program (QNRC2016560 to Xu XQ).

References

Haradome

Usui

, et al. Orbital lymphoproliferative disorders (OLPDs): value of MR imaging for differentiating orbital lymphoma from benign OPLDs. Am J Neuroradiol 2014; 35: 1976–1982.

Sun

Song

Wang

, et al. Lymphoma and inflammation in the orbit: Diagnostic performance with diffusion-weighted imaging and dynamic contrast-enhanced MRI. J Magn Reson Imaging 2017; 45: 1438–1445.

Liu

, et al. Benign and malignant orbital lymphoproliferative disorders: Differentiating using multiparametric MRI at 3.0T. J Magn Reson Imaging 2017; 45: 167–176.

Xian

Zhang

Wang

, et al. Value of MR imaging in the differentiation of benign and malignant orbital tumors in adults. Eur Radiol 2010; 20: 1692–1702.

Yuan

Kuai

Chen

, et al. Assessment of dynamic contrast-enhanced magnetic resonance imaging in the differentiation of malignant from benign orbital masses. Eur J Radiol 2013; 82: 1506–1511.

Liu

, et al. Orbital benign and malignant lymphoproliferative disorders: Differentiation using semi-quantitative and quantitative analysis of dynamic contrast-enhanced magnetic resonance imaging. Eur J Radiol 2017; 88: 88–94.

Asbach

Siebert

, et al. Characterization of orbital masses by multiparametric MRI. Eur J Radiol 2016; 85: 324–336.

Qian

, et al. Combined diffusion-weighted imaging and dynamic contrast-enhanced MRI for differentiating radiologically indeterminate malignant from benign orbital masses. Clin Radiol 2017; 72: 903.e9–903.e15.

Nguyen

Shah

Mortazavi

, et al. Non-invasive quantification of tumour heterogeneity in water diffusivity to differentiate malignant from benign tissues of urinary bladder: a phase I study. Eur Radiol 2017; 27: 2146–2152.

10.

Hectors

Wagner

Bane

, et al. Quantification of hepatocellular carcinoma heterogeneity with multiparametric magnetic resonance imaging. Sci Rep 2017; 7: 2452–2452.

11.

Just

. Improving tumour heterogeneity MRI assessment with histograms. Br J Cancer 2014; 111: 2205–2213.

12.

Yoon

Park

, et al. Tumor heterogeneity in lung cancer: assessment with dynamic contrast-enhanced MR imaging. Radiology 2016; 280: 940–948.

13.

Choi

Lee

Ahn

, et al. Primary central nervous system lymphoma and atypical glioblastoma: differentiation using the initial area under the curve derived from dynamic contrast-enhanced MR and the apparent diffusion coefficient. Eur Radiol 2017; 27: 1344–1351.

14.

Kim

Choi

Yun

, et al. Prognosis prediction of non-enhancing T2 high signal intensity lesions in glioblastoma patients after standard treatment: application of dynamic contrast-enhanced MR imaging. Eur Radiol 2017; 27: 1176–1185.

15.

Yoo

Choi

Kim

, et al. Dynamic contrast-enhanced MR imaging in predicting progression of enhancing lesions persisting after standard treatment in glioblastoma patients: a prospective study. Eur Radiol 2017; 27: 3156–3166.

16.

Choi

Lee

Choi

, et al. Detection of local tumor recurrence after definitive treatment of head and neck squamous cell carcinoma: histogram analysis of dynamic contrast-enhanced T1-weighted perfusion MRI. Am J Roentgenol 2017; 208: 42–47.

17.

, et al. Utility of histogram analysis of ADC maps for differentiating orbital tumors. Diagn Interv Radiol. 2016; 22: 161–167.

18.

Tofts

Brix

Buckley

, et al. Estimating kinetic parameters from dynamic contrast-enhanced T1-weighted MRI of a diffusable tracer: standardized quantities and symbols. J Magn Reson Imaging 1999; 10: 223–232.

19.

Gaddikeri

Tailor

, et al. Dynamic contrast-enhanced MR imaging in head and neck cancer: techniques and clinical applications. Am J Neuroradiol 2016; 37: 588–595.

20.

Jung

Yeom

Kim

, et al. Glioma: Application of histogram analysis of pharmacokinetic parameters from T1-weighted dynamic contrast-enhanced MR imaging to tumor grading. Am J Neuroradiol 2014; 35: 1103–1110.

21.

Choi

Kim

Lee

, et al. The added prognostic value of preoperative dynamic contrast-enhanced MRI histogram analysis in patients with glioblastoma: analysis of overall and progression-free survival. Am J Neuroradiol 2015; 36: 2235–2241.

22.

Gao

, et al. Utility of dynamic contrast-enhanced magnetic resonance imaging for differentiating glioblastoma, primary central nervous system lymphoma and brain metastatic tumor. Eur J Radiol 2016; 85: 1722–1727.

23.

Mills

Soh

Rose

, et al. Candidate biomarkers of extravascular extracellular space: a direct comparison of apparent diffusion coefficient and dynamic contrast-enhanced MR imaging–derived measurement of the volume of the extravascular extracellular space in glioblastoma multiforme. Am J Neuroradiol 2010; 31: 549–553.

24.

Arlinghaus

Rahman

, et al. On the relationship between the apparent diffusion coefficient and extravascular extracellular volume fraction in human breast cancer. Magn Reson Imaging 2011; 29: 630–638.

25.

Heye

Merkle

Reiner

, et al. Reproducibility of dynamic contrast-enhanced MR imaging. Part II. Comparison of intra- and interobserver variability with manual region of interest placement versus semiautomatic lesion segmentation and histogram analysis. Radiology 2013; 266: 812–821.

26.

Razek

Elkhamary

Mousa

. Differentiation between benign and malignant orbital tumors at 3-T diffusion MR-imaging. Neuroradiology 2011; 53: 517–522.

27.

Sepahdari

Politi

Aakalu

, et al. Diffusion-weighted imaging of orbital masses: multi-institutional data support a 2-ADC threshold model to categorize lesions as benign, malignant, or indeterminate. Am J Neuroradiol. 2014; 35: 170–175.