Multiparametric MRI for differential diagnosis of primary central nervous system lymphoma and atypical glioblastoma: an analysis incorporating DWI, DCE-MRI, and contrast agent preload DSC-PWI (original) (raw)

Abstract

Purpose

The differential diagnosis of primary central nervous system lymphoma (PCNSL) and atypical glioblastoma (aGBM) exhibiting homogeneous enhancement and negligible necrosis poses a significant challenge for conventional MRI. The study aims to investigate diffusion-weighted imaging (DWI), dynamic contrast-enhanced (DCE) MRI, and contrast agent (CA) preload dynamic susceptibility contrast perfusion-weighted imaging (DSC-PWI) to differentiate aGBM and PCNSL.

Materials and methods

This retrospective study analyzed 27 patients with aGBM (solid enhancement without visible necrosis) and 105 patients with PCNSL, all undergoing preoperative DWI, DCE-MRI, and CA preload DSC-PWI. The relative apparent diffusion coefficient (rADC) and relative cerebral blood volume (rCBV) were obtained from DWI and DSC-PWI. The pharmacokinetic parameters (Ktrans, Ve, Kep, and iAUC) were acquired using DCE-MRI. The independent-samples t-test and Mann-Whitney U test were utilized to compare parameters. A binary logistic regression analysis was performed to assess the combined effect of various parameters. Before regression analysis, collinearity analysis of parameters was performed. The diagnostic capability of each parameter and their combination were evaluated by receiver operating characteristic (ROC) with area under the curve (AUC) and compared with DeLong test.

Results

In comparison to aGBM, the Ktrans, Ve, and iAUC were significantly elevated in PCNSL, whereas the rCBV and rADC were significantly lower (p < 0.05 for all comparisons). Meanwhile, these parameters allowed excellent diagnostic performance (AUC = 0.817 [rCBV], 0.751 [rADC], 0.808 [Ktrans], 0.765 [Ve], and 0.801 [iAUC]; DeLong test, _p_ > 0.05 for all comparisons). Notably, the combination of all these parameters significantly increased the probability of distinguishing aGBM from PCNSL (AUC = 0.966).

Conclusions

DWI, DCE-MRI, and CA preload DSC-PWI can effectively differentiate aGBM from PCNSL, and the combination of all three techniques significantly enhances the discriminatory efficacy.

Key results

View this article's peer review reports

Similar content being viewed by others

Introduction

Primary central nervous system lymphoma (PCNSL), the majority of subtypes is diffuse large B-cell lymphoma (DLBCL), albeit a rare form of brain tumor [1]. Conversely, glioblastoma (GBM), IDH-wildtype, stands as the most commonly occurring malignant primary brain tumor [2]. Both diseases have increased in prevalence in many countries and can be attributed to several factors, including population aging and advances in diagnostic technology [3]. Given the vast differences in treatment modalities and prognostic outcomes associated with these conditions [1, 4], it is of utmost importance to ensure an accurate and prompt pretreatment diagnosis.

Numerous studies have consistently demonstrated the significance of multimodal MRI in distinguishing between PCNSL and GBM [5,6,7,8,9]. Morphological MRI techniques, however, encounter limitations in cases where atypical GBM (aGBM) exhibits minimal or no necrosis, and the typical imaging of immunocompetent PCNSL presenting as a homogeneously enhanced mass, thereby complicating the differential diagnosis [7, 10]. Consequently, multiparametric MRI has been employed to overcome these challenges, including diffusion-weighted imaging (DWI), dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI), and dynamic susceptibility contrast perfusion-weighted imaging (DSC-PWI).

DCE-MRI serves as a tool that reflects alterations in tumor blood flow, vascular permeability, and interstitial/intravascular volume, through the utilization of various parameters such as the volume transfer constant (Ktrans), extracellular/extravascular volume fraction (Ve), rate constant (Kep), and initial area under the curve (iAUC). These parameters enable the DCE-MRI to distinguish between the diagnoses of GBM and PCNSL [6, 11]. In general, the permeability of PCNSL is perceived to be superior to that of GBM. This may stem from the preserved vascular integrity of GBM, despite the presence of endothelial cell proliferation. Conversely, tumor cells in PCNSL tend to infiltrate the vicinity of blood vessels, resulting in an incomplete disruption of the vascular structure [12]. Therefore, identifying the precise role of DCE-MRI parameters in differentiating aGBM from PCNSL is crucial for guiding clinical decision-making.

DSC-PWI has exhibited significant diagnostic proficiency in distinguishing GBM from PCNSL through noninvasive in vivo evaluation of the microvascular environment, achieving a diagnostic accuracy of up to 93% [13, 14]. However, there exists a debate surrounding the perfusion status of PCNSL. Notably, most studies have concurred that the angiocentric growth pattern of PCNSL leads to suboptimal perfusion [15]. Conversely, some investigations have observed that a fraction of PCNSL cases exhibit hyperperfusion, comparable even to that seen in GBM [6, 16,17,18,19]. On a different note, the accurate measurement of relative cerebral blood volume (rCBV) in PCNSL has been hampered by the underestimation resulting from contrast agent (CA) spillover caused by blood-brain barrier disruption. To address this issue, the application of preload correction in DSC-PWI has proven effective in elevating rCBV values in lymphomas [20, 21]. Clarifying the role of CA preload DSC-PWI in differentiating aGBM from PCNSL is essential to improve the accuracy of clinical diagnosis.

Additionally, DWI, a frequently employed sequence in clinical diagnostic imaging, can assess tumor characteristics by reflecting cell density, thus possessing discriminative value in distinguishing between aGBM and PCNSL [22, 23].

The purpose of this study is to investigate the effectiveness of multiparameter MRI including, DWI, DCE-MRI, and CA preload DSC-PWI, in accurately differentiating between aGBM and PCNSL.

Materials and methods

Patient selection

This study received approval from the Institutional Review Board of the First Affiliated Hospital of Fujian Medical University, and the requirement for patient informed consent was waived. Spanning from January 2017 to December 2023, our clinical database encompassed 27 instances of aGBM (homogeneous enhancement without apparent necrosis), alongside 105 cases of PCNSL. The inclusion criteria were as follows: (1) histopathological diagnosis of IDH wild-type GBM or PCNSL according to the World Health Organization Classification of Tumors of the Central Nervous System 2021 [24]; (2) untreated, immunocompetent adult patients; and (3) availability of complete preoperative MR images including DWI, DCE-MRI, and DSC-PWI. The exclusion criteria were as follows: (1) absence of any required MR images or insufficient image quality for analysis; (2) typical cases with marked necrotic inhomogeneous enhancement in GBM patients, as evaluated by two independent readers with 13 and 2 years of experience in neuroradiological imaging, respectively. Immunohistochemical analysis of 5-µm-thick sections of paraffin-embedded tumor specimens from all patients evaluated was performed to clarify IDH gene mutation status. A research flow diagram is shown in Fig. 1.

Fig. 1

Flowchart of the included participants. GBM = Glioblastoma, PCNSL = Primary central nervous system lymphoma

MR imaging

All MR images were acquired in the routine clinical workup using 3.0-T MR imaging systems (Magnetom Skyra, Magnetom Prisma; Siemens Healthcare) equipped with identical 20-channel head/neck coils. The morphological MRI sequences encompassed axial T1-weighted imaging (TR/TE = 250/2.48 ms), axial T2-weighted imaging (TR/TE = 4000/125 ms), axial fluid-attenuated inversion recovery imaging (TR/TE = 9000/94 ms), and three-orthogonal-plane contrast-enhanced T1-weighted imaging (CE-T1WI, TR/TE = 250/2.48 ms). The remaining parameters were the same, including FOV (220 × 220 mm2), slice thickness (5 mm), matrix (256 × 256), and slice spacing (1 mm).

DWI was conducted in the axial plane utilizing an axial echo-planar sequence (TR/TE, 8200/102ms). A diffusion gradient was applied in three orthogonal directions (b = 0 and 1000 s/mm2). The remaining parameters included FOV, 220 × 220 mm; number of excitations, 2.0; slice thickness, 5 mm; matrix, 128 × 128; slice spacing, 1 mm. The MRI system automatically generated the corresponding apparent diffusion coefficient (ADC) maps using a Siemens workstation equipped with standard software (syngo. via; Siemens Healthcare).

DCE-MRI was performed as a preloaded method with parameters (TR/TE, 5.08/1.79 ms; matrix, 138 × 192; FOV, 220 × 220 mm; number of excitations, 1.0; slice thickness, 5 mm; slice spacing, 1 mm; a flip-angle of 15°). The bolus protocol involved the injection of a standard gadobenate dimeglumine was adopted, that a standard gadobenate dimeglumine (Gd-BOPTA, MultiHance; Bracco Imaging SpA) was injected (0.2 mL/kg), at a rate of 3 mL/s using a power injector (ACIST Empower MR Injector, Bracco or OptiStar Elite MR Contrast Delivery System, Mallinckrodt Pharmaceuticals), followed by a 20 mL continuous saline flush at the same injection rate. Voxel T1 values were measured using the T1-mapping technique with the following parameters: TR/TE, 4.95/1.81 ms; matrix, 138 × 192; FOV, 220 × 220 mm; slice thickness, 2 mm; and flip angle, 2° and 15°. The total acquisition time for T1 maps was 1 min 47 s, and DCE-MRI was 5 min. Immediately following the DCE-MRI acquisition, DSC-PWI scanning was conducted.

CA preload DSC-PWI was performed with the following parameters: gradient-echo echo planar imaging with TR/TE, 1600/30 ms; matrix, 128 × 128; FOV, 220 × 220 mm; number of excitations, 1.0; slice thickness, 5 mm; slice spacing, 1 mm; a flip-angle of 90°. Initially, non-enhanced images were captured during the first three to four phases to establish a pre-contrast baseline. Subsequently, the DSC-PWI scan commenced, during which the Gd-BOPTA was injected at a rate of 0.2 mL/kg, adhering to the same bolus protocol as utilized in the DCE-MRI. In the present study, DSC-PWI was acquired with a high flip angle (90°), and a partial suppression of T1 effect was achieved with a full-dose preloading (0.2 mL/kg).

Image postprocessing and analysis

All imaging assessments were performed on a Siemens workstation with standard software (Syngo. via; Siemens Healthcare). All morphological MRI data were assessed by two neuroradiologists who were blinded to tumor histology. When two observers disagreed, a senior neuroradiologist made the final decision. For each patient, the enhancing solid of the tumor was identified on T2WI, FLAIR, and CE-T1WI. All the regions of interest (ROIs) should be taken care of to exclude calcified, hemorrhagic, necrotic, and cystic lesions that may affect the assessment. In addition, five ROIs of 20–25 mm² size were outlined from the contralateral unaffected white matter to obtain the corresponding relative values.

To obtain the DWI data, the ADC map was aligned with the CE-T1WI, and the ROIs at each level were carefully outlined along the boundaries of tumor enhancement on the CE-T1WI after alignment, and the corresponding ROIs were generated on the ADC map to generate the ADC values. The rADC was calculated by dividing the mean whole-tumor ADC value by the mean ADC value of the contralateral unaffected white matter. ADC values were expressed as × 10− 3 mm2/s.

DCE-MRI processing was performed using commercial software tools (Tissue 4D, Syngo.via, Siemens Healthcare). Post-processing procedures included motion correction of dynamic sequences; alignment of precontrast and morphology to a dynamic reference; plotting ROIs on DCE-MRI images containing lesions and adjacent normal tissue; generation of concentration profiles based on the Tofts model [25]; and use of a population-based arterial input function (AIF) with settings of “fast” [26], “intermediate” [27], or “slow” [28] model with the smallest rank-square parameter; the rank-square parameter is a measure of the error of the fit. The smaller the parameter, the better the fit; parametric plots (Ktrans, Kep, Ve, and iAUC) of the ROI were calculated from the Tofts model. Since DCE-MRI is a repeated acquisition of T1-weighted images at intervals, it is able to outline the boundaries of the enhancing tumor directly on the map. Therefore, in this study, we chose to carefully outline the ROIs on the color images along the boundaries of the enhanced tumor at the largest level, generate the time-intensity curves, and record the corresponding Ktrans, Kep, Ve, and iAUC parameter values.

DSC-PWI processing was performed using commercial software tools (Neurology, Syngo.via, Siemens Healthcare). A local AIF using automatic arterial pixel selection was used to construct the local arterial input function. A global clustering algorithm, which analyses time series for all voxels and selects a viable AIF, was used to automatically construct the AIF for each dataset. The CBV colour map was automatically reconstructed based on the single-compartment model and the exogenous perfusion deconvolution algorithm. And CBV leakage correction was performed, leakage-corrected CBV with T1 correction was performed mainly by the Boxerman-Schmainda-Weisskopf (BSW) model [29]. To obtain the DSC-PWI data, the CBV map was aligned with the CE-T1WI. Following this alignment, the ROIs at each level were carefully outlined along the boundaries of the tumor enhancement on the CE-T1WI. The corresponding ROIs were then generated on the CBV map, thus generating the CBV values. The rCBV was calculated by dividing the mean whole-tumor CBV value by the mean CBV value of the contralateral unaffected white matter.

To assess the inter-observer agreement, a neuroradiologist with 3 years of experience in brain imaging, who was blinded to the tumor histology, consistently measured the parameters across all patients. To evaluate intra-observer consistency, a neuroradiologist with 13 years of experience, who was blinded to the tumor histology, measured the parameters repeatedly with a minimum interval of 1 month. The initial measurement results were subsequently employed for further statistical analysis.

Statistical analysis

Statistical analysis was performed using SPSS software (Version 22.0, SPSS Inc., Chicago, USA) and MedCalc (Version 12.1.0, MedCalc Inc., Mariakierke, Belgium). The Kolmogorov-Smirnov and the quantile-quantile plot (Q-Q plot) tests were utilized to ascertain whether the continuous variables were distributed normally. The Levene test was used to test the equality of variances of normally distributed continuous variables. An independent-sample t-test was performed on normally distributed variables of the continuous variables, and the Mann-Whitney U test on nonparametrically distributed variables. A binary logistic regression analysis was performed to assess the combined effect of various parameters on the prediction of pathology results. The dependent variable of this logistic regression model was the pathologic outcome and the independent variables were the parameters that were statistically significant according to the statistical methods described above. Before regression analysis, collinearity analysis of parameters was performed. The presence of covariance was indicated if the tolerance (Tol) was less than 0.1 or the variance inflation factor (VIF) was greater than 10, and parameters with significant covariance and low AUC values were excluded. For each parameter, the ability to discriminate between aGBM and PCNSL was quantified by the area under the curve (AUC) of the receiver operating characteristic (ROC) analysis. The statistical differences between the AUC values were evaluated using the DeLong test. Multiple comparisons were corrected for p values using the Holm-Bonferroni correction method. The reproducibility of rADC, DCE-MRI-derived parameters, and rCBV was assessed using intraclass correlation coefficients (ICCs). All p < 0.05 were considered to represent statistical significance.

Results

Basic characteristics

The clinical characteristics of these patients are described in Table 1. There were 105 patients with PCNSL (mean age, 60.82 ± 10.48 years; 46 [43.81%] were men) and 27 patients with aGBM (mean age, 56.85 ± 10.39 years; 20 [74.07%] were men). Examples of images of selected patients are shown in Figs. 2 and 3, respectively.

Table 1 Clinical characteristics of the patients

Full size table

Fig. 2

Pretreatment CE-T1WI, DWI, ADC map, DSC-PWI, and DCE-MRI (Ktrans and iAUC map), a patient with glioblastoma. (A) Axial contrast-enhanced image shows a homogeneous enhancing mass in the right frontoparietal lobe. (B) (C) Axial DWI and ADC map at b = 1000s/mm2 show restricted diffusion in the contrast-enhancing tumor (rADC = 1.471). (D) Axial rCBV map illustrated rCBV of 9.860 in contrast-enhancing tumor. (E) (F) Axial Ktrans and iAUC maps of DCE-MRI (Ktrans = 0.045, iAUC = 0.121). CE-T1WI = Contrast-enhanced T1WI, DCE-MRI = Dynamic contrast-enhanced magnetic resonance imaging, DSC-PWI = Dynamic susceptibility contrast perfusion-weighted imaging, DWI = Diffusion-weighted imaging, iAUC = Initial area under the gadolinium concentration-time curve, Ktrans = Volume transfer constant, rADC = Relative apparent diffusion coefficient, rCBV = Relative cerebral blood volume

Fig. 3

Pretreatment CE-T1WI, DWI, ADC map, DSC-PWI, and DCE-MRI (Ktrans and iAUC map), a patient with PCNSL. (A) Axial contrast-enhanced image shows a prominently homogeneous enhancing mass in the right basal ganglia region. (B) (C) Axial DWI and ADC map at b = 1000s/mm2 show restricted diffusion in the contrast-enhancing tumor (rADC = 1.204). (D) Axial rCBV map illustrated rCBV of 3.849 in contrast-enhancing tumor. (E) (F) Axial Ktrans and iAUC maps of DCE-MRI (Ktrans = 0.075, iAUC = 0.146). CE-T1WI = Contrast-enhanced T1WI, DCE-MRI = Dynamic contrast-enhanced magnetic resonance imaging, DSC-PWI = Dynamic susceptibility contrast perfusion-weighted imaging, DWI = Diffusion-weighted imaging, iAUC = Initial area under the gadolinium concentration-time curve, Ktrans = Volume transfer constant, PCNSL = Primary central nervous system lymphoma, rADC = Relative apparent diffusion coefficient, rCBV = Relative cerebral blood volume

Comparison of DWI, DCE-MRI, and DSC-PWI parameters between aGBM and PCNSL

The differences in rADC, DCE-MRI parameters, and rCBV of aGBM and PCNSL are shown in Table 2 and Fig. 4. The rADC and rCBV of aGBM were higher than those of PCNSL (rADC, 1.329 vs. 1.116, p < 0.001; rCBV, 8.253 vs. 4.667, _p_ < 0.001). And, DCE-MRI parameters including Ktrans, Ve, and iAUC, were significantly lower in aGBM compared with PCNSL (Ktrans, 0.041 vs. 0.075; Ve, 0.122 vs. 0.227; iAUC, 0.079 vs. 0.129; _p_ < 0.001 for all comparisons). Meanwhile, these parameters allowed excellent diagnostic performance (AUC = 0.751 [rADC], 0.817 [rCBV], 0.808 [Ktrans], 0.765 [Ve], and 0.801 [iAUC]; DeLong test, _p_ > 0.05 for all comparisons) (Fig. 5). There was no statistical difference between the Kep parameters of aGBM and PCNSL in this study (0.360 and 0.306, p > 0.05).

Table 2 Differences of parameters between aGBM and PCNSL

Full size table

Fig. 4

Violin plots of different parameters in patients with aGBM and PCNSL. The lines in the violinplot indicate the median and interquartile range. The parameters rADC, Ktrans, Ve, iAUC, and rCBV were statistically different between aGBM and PCNSL (p < 0.001) and the parameters Kep were not statistically different (_p_ > 0.05). aGBM = atypical glioblastoma, iAUC = Initial area under the gadolinium concentration-time curve, Ktrans = Volume transfer constant, PCNSL = Primary central nervous system lymphoma, rADC = Relative apparent diffusion coefficient, rCBV = Relative cerebral blood volume, Ve = Volume of extravascular extracellular space per unit of tissue, Kep = Ktrans/Ve

Fig. 5

ROC curves show the diagnostic ability of different parameters (A) and combinations of parameters (B) to distinguish album from PCNSL. aGBM = Atypical glioblastoma, iAUC = Initial area under the gadolinium concentration-time curve, Ktrans = Volume transfer constant, PCNSL = Primary central nervous system lymphoma, rADC = Relative apparent diffusion coefficient, rCBV = Relative cerebral blood volume, Ve = Volume of extravascular extracellular space per unit of tissue, Kep = Ktrans/Ve

Multi-parameter comparison of aGBM and PCNSL

Table 2 and Fig. 5 show the results of parameters for the differentiation of aGBM and PCNSL. To minimize the possibility of overfitting, the covariance analysis of statistically significant parameters showed significant covariance between Ktrans and iAUC parameters (Variance Inflation Factor, VIF > 10), therefore, we chose to keep the Ktrans parameter with the higher AUC value in the regression model. Comparisons between the different parametric models are detailed in Table S1 in the Supplementary Material, and the relative contribution of each parameter to the combined model is shown in Table S2. The combination of both imaging parameters mostly improved the discriminative efficacy of aGBM and PCNSL compared to a single parameter (p < 0.05), except for the model of rADC combined with DCE-MRI vs. the model of DCE-MRI in comparison (_p_ > 0.05). A model for the union of three technologies yielded the best diagnostic performance and was significantly superior to the two technologies models (AUC = 0.966 vs. 0.877, 0.936, and 0.918, DeLong test, p < 0.05 for all comparisons).

Reproducibility of the parameters

The ICC values for rADC, DCE-MRI parameters, and rCBV are shown in Table 3.

Table 3 The ICC analysis of different parameters in patients with aGBM and PCNSL

Full size table

Discussion

Our investigation evaluated the diagnostic performance of three advanced technological modalities: DWI, DCE-MRI, and CA preload DSC-PWI. The objective was to enhance the preoperative differentiation between aGBM and PCNSL, which are indistinguishable using morphological MRI. Our findings revealed that patients diagnosed with PCNSL exhibited notably lower rADC and rCBV compared to significantly higher Ktrans, Ve, and iAUC than those with aGBM. Additionally, the AUC results underscored the substantial increase in the likelihood of distinguishing between these two diseases when employing a multi-parametric assessment approach. Notably, the parameters demonstrated a high degree of measurement repeatability.

This study underscores the significant discriminatory potential of preoperative ADC values in distinguishing aGBM from PCNSL. Prior research has established the role of ADC values in differentiating these two tumor types, consistently demonstrating lower ADC values in PCNSL cases [8, 30]. Notably, the rADC values in our aGBM patients ranged from 1.138 to 1.471, which are comparatively lower than previous reports [8, 31, 32]. This discrepancy can be attributed to the atypical nature of the GBMs in our study, which lacked significant necrosis and exhibited increased cell density, leading to decreased ADC values. While aGBM exhibits uniform enhancement, morphologically resembling lymphoma, histopathologically, it is characterized by the presence of pseudopalisading necrosis [33]. This histopathological feature likely contributes to the relatively higher ADC values observed in atypical GBMs compared to lymphomas.

This study investigates the utility of DCE-MRI parameters in the preoperative differential diagnosis of aGBM and PCNSL. Statistically significant differences were observed in Ktrans, Ve, and iAUC, which are partially indicative of the distinct tumor microvasculature characteristics of aGBM and PCNSL. Lu et al. [34] reported significantly higher mean Ktrans and Ve values in PCNSL than in GBM. An elevated Ktrans reflects a greater degree of blood-brain barrier disruption and vascular permeability [35]. Our study concurs with these findings, demonstrating significantly higher Ktrans, Ve, and iAUC values in PCNSL. However, the results for iAUC were inconsistent with Choi et al. [36], who reported higher iAUC values in GBM than in PCNSL, suggesting that the differences in iAUC might be attributed to the higher perfusion in GBM rather than the higher permeability in PCNSL. Nevertheless, the underlying histopathological mechanisms responsible for the differences in iAUC between PCNSL and GBM remain unclear and warrant further exploration.

Multiple studies have established the ability of DSC-PWI to differentiate GBM from PCNSL, which typically exhibits hypoperfusion, based on rCBV values [6, 7, 10, 37]. Intriguingly, our analysis revealed that some PCNSL cases displayed high rCBV values, exhibiting both hypoperfusion and hyperperfusion patterns. This observation concurs with the perspectives of certain scholars [18, 19, 38]. Yasuo et al. [38] postulated that PCNSL may not require significant neo-angiogenesis in the early stages but might demand an accelerated neo-angiogenesis rate as it infiltrates and destroys brain parenchyma. Xing et al. [39] provided evidence suggesting that the rCBV value of PCNSL is location-dependent, with lower values observed in the white matter group and higher values in the cortical gray matter group. In our study, the CA preloading method was employed for rCBV correction, leading to increased rCBV in aGBM and PCNSL compared with the uncorrected method because both tumors have blood-brain-barrier damage. Nevertheless, DSC-PWI (especially the rCBV parameter) has shown excellent potential in differentiating aGBM from PCNSL. However, the histopathology of PCNSL angiogenesis is unknown, and perfusion is influenced by factors such as tumor location, highlighting the need for further studies to optimize diagnostic accuracy.

Multiparameter MRI techniques, comprising DWI, DCE-MRI, and CA preloaded DSC-PWI, have significantly augmented the potential to distinguish aGBM from PCNSL. In addition to the ADC, CBV, and DCE metrics evaluated in this study, other advanced MRI biomarkers have shown promise in differentiating PCNSL from GBM. Notably, the percentage of signal recovery (PSR) derived from DSC-PWI and intratumoral susceptibility signals (ITSS) from susceptibility-weighted imaging (SWI) is increasingly utilized in clinical research. Lower PSR values in PCNSL compared to GBM have been reported, reflecting differences in vascular permeability and contrast leakage patterns [20]. However, PSR calculation requires non-preload DSC sequences, which were not employed in our study due to institutional protocol standardization focused on preloaded DSC for CBV quantification. Similarly, ITSS, defined as hypointense foci within tumors on SWI, are more prevalent in GBM due to microhemorrhages and abnormal neovascularization [7], whereas aGBM lacks these features. Additionally, Suh et al. [40] found that the AUC of three neuroradiologists for recognizing aGBM and PCNSL images were 0.707, 0.759, and 0.695, respectively. In the present study, we found that the combined application of DWI, DCE-MRI, and DSC-PWI had an ACU of 0.966, which significantly improved the diagnosis of aGBM and PCNSL. The multiparametric approach improved the diagnostic accuracy in this cohort from 57.6% to 81.1% concurrently to 93.2% compared to relying on a single parameter. These three modalities offer distinct insights into tumor characteristics, encompassing tumor cell density, permeability, and perfusion patterns. Integrating these three techniques provides additional tumor information crucial for differential diagnosis and clinical treatment guidance. Based on our findings, the diagnostic process of aGBM and PCNSL can be further refined by suggesting that patients undergo concurrent DWI, DSC-PWI, and DCE-MRI scans, and that DCE-MRI, as a preloading modality, can provide information on vascular permeability at the same time.

The primary limitations of this study are outlined below. Firstly, due to its retrospective design, potential biases in patient selection and data processing may have been introduced. Second, these data came from only one institution with a small sample size, which may limit the representativeness of our findings despite the excellent results. To address these limitations and to test our hypothesis, a multicenter prospective study with long-term follow-up is necessary to include more atypical cases. Third, our DSC-PWI acquisition protocol used non-standard parameters (90° flip angle with full-dose preloading), which differ from the current consensus guidelines (60° flip angle with preloading or 30° flip angle without preloading). Although we applied BSW mathematical leakage correction during post-processing, the use of these parameters may result in differences in rCBV quantification compared to standardised methods. Future studies should follow established guidelines to improve cross-centre comparability. Additionally, it is difficult to fully elucidate the complex microstructure of biological tissues with limited techniques. Other methods, such as arterial spin labeling (ASL) and 18 F-fluorodeoxyglucose-positron emission tomography (FDG-PET), may be able to provide more tumor information.

In conclusion, our findings underscore the substantial discriminatory potential of DWI, DCE-MRI, and CA preload DSC-PWI in distinguishing aGBM from PCNSL. Furthermore, the combined quantitative analysis of these modalities significantly enhances diagnostic accuracy. The clinical significance of this integrated approach lies in its potential to improve diagnostic certainty and treatment outcomes for patients with these challenging tumor types.

Data availability

The data that support the findings of this study are not openly available due to reasons of sensitivity and are available from the corresponding author upon reasonable request.

Abbreviations

CA:

Contrast agent

DCE-MRI:

Dynamic contrast-enhanced magnetic resonance imaging

DSC-PWI:

Dynamic susceptibility contrast perfusion-weighted imaging

DWI:

Diffusion-weighted imaging

aGBM:

Atypical Glioblastoma

iAUC:

Initial area under the gadolinium concentration-time curve

Ktrans:

Volume transfer constant

PCNSL:

Primary central nervous system lymphoma

rADC:

Relative apparent diffusion coefficient

rCBV:

Relative cerebral blood volume

Ve :

Volume of extravascular extracellular space per unit of tissue

References

  1. de Koning ME, Hof JJ, Jansen C, et al. Primary central nervous system lymphoma. J Neurol. 2024;271(5):2906–13.
    Article PubMed Google Scholar
  2. Ostrom QT, Price M, Neff C, et al. CBTRUS statistical report: primary brain and other central nervous system tumors diagnosed in the united States in 2016–2020. Neuro Oncol. 2023;25(12 Suppl 2):iv1–99.
    Article PubMed PubMed Central Google Scholar
  3. Grech N, Dalli T, Mizzi S, et al. Rising incidence of glioblastoma multiforme in a Well-Defined population. Cureus. 2020;12(5):e8195.
    PubMed PubMed Central Google Scholar
  4. Schaff LR, Mellinghoff IK. Glioblastoma and other primary brain malignancies in adults: A review. JAMA. 2023;329(7):574–87.
    Article PubMed PubMed Central Google Scholar
  5. Toh CH, Wei KC, Chang CN, Ng SH, Wong HF. Differentiation of primary central nervous system lymphomas and glioblastomas: comparisons of diagnostic performance of dynamic susceptibility contrast-enhanced perfusion MR imaging without and with contrast-leakage correction. AJNR Am J Neuroradiol. 2013;34(6):1145–9.
    Article PubMed PubMed Central CAS Google Scholar
  6. Kang KM, Choi SH, Chul-Kee P, et al. Differentiation between glioblastoma and primary CNS lymphoma: application of DCE-MRI parameters based on arterial input function obtained from DSC-MRI. Eur Radiol. 2021;31(12):9098–109.
    Article PubMed CAS Google Scholar
  7. Kickingereder P, Wiestler B, Sahm F, et al. Primary central nervous system lymphoma and atypical glioblastoma: multiparametric differentiation by using diffusion-, perfusion-, and susceptibility-weighted MR imaging. Radiology. 2014;272(3):843–50.
    Article PubMed Google Scholar
  8. Yu X, Hong W, Ye M, et al. Atypical primary central nervous system lymphoma and glioblastoma: multiparametric differentiation based on non-enhancing volume, apparent diffusion coefficient, and arterial spin labeling. Eur Radiol. 2023;33(8):5357–67.
    Article PubMed PubMed Central Google Scholar
  9. Han Y, Wang ZJ, Li WH, et al. Differentiation between primary central nervous system lymphoma and atypical glioblastoma based on MRI morphological feature and signal intensity ratio: A retrospective multicenter study. Front Oncol. 2022;12:811197.
    Article PubMed PubMed Central Google Scholar
  10. Feng A, Li L, Huang T, et al. Differentiating glioblastoma from primary central nervous system lymphoma of atypical manifestation using multiparametric magnetic resonance imaging: A comparative study. Heliyon. 2023;9(4):e15150.
    Article PubMed PubMed Central CAS Google Scholar
  11. Saini J, Kumar Gupta P, Awasthi A, et al. Multiparametric imaging-based differentiation of lymphoma and glioblastoma: using T1-perfusion, diffusion, and susceptibility-weighted MRI. Clin Radiol. 2018;73(11):986.e987-986.e915.
  12. Kickingereder P, Sahm F, Wiestler B, et al. Evaluation of microvascular permeability with dynamic contrast-enhanced MRI for the differentiation of primary CNS lymphoma and glioblastoma: radiologic-pathologic correlation. AJNR Am J Neuroradiol. 2014;35(8):1503–8.
    Article PubMed PubMed Central CAS Google Scholar
  13. Pons-Escoda A, Garcia-Ruiz A, Naval-Baudin P, et al. Presurgical identification of primary central nervous system lymphoma with normalized Time-Intensity curve: A pilot study of a new method to analyze DSC-PWI. AJNR Am J Neuroradiol. 2020;41(10):1816–24.
    Article PubMed PubMed Central CAS Google Scholar
  14. Suh CH, Kim HS, Jung SC, et al. MRI as a diagnostic biomarker for differentiating primary central nervous system lymphoma from glioblastoma: A systematic review and meta-analysis. J Magn Reson Imaging. 2019;50(2):560–72.
    Article PubMed Google Scholar
  15. Koeller KK, Smirniotopoulos JG, Jones RV. Primary central nervous system lymphoma: radiologic-pathologic correlation. Radiographics. 1997;17(6):1497–526.
    Article PubMed CAS Google Scholar
  16. Dandois V, De Coene B, Laloux P, Godfraind C, Cosnard G. Increased relative cerebral blood volume (rCBV) in brain lymphoma. J Neuroradiol. 2011;38(3):191–3.
    Article PubMed CAS Google Scholar
  17. Blasel S, Jurcoane A, Bahr O, et al. MR perfusion in and around the contrast-enhancement of primary CNS lymphomas. J Neurooncol. 2013;114(1):127–34.
    Article PubMed Google Scholar
  18. Hua S, Liu Y, Chen Z. [Supratentorial primary central nervous system lymphoma with hyperperfusion:report of one case]. Zhongguo Yi Xue Ke Xue Yuan Xue Bao. 2020;42(3):421–6.
    PubMed Google Scholar
  19. Lee B, Park JE, Bjornerud A, et al. Clinical value of vascular permeability estimates using dynamic susceptibility contrast MRI: improved diagnostic performance in distinguishing hypervascular primary CNS lymphoma from glioblastoma. AJNR Am J Neuroradiol. 2018;39(8):1415–22.
    PubMed PubMed Central CAS Google Scholar
  20. Wang F, Zhou X, Chen R, et al. Improved performance of non-preloaded and high flip-angle dynamic susceptibility contrast perfusion-weighted imaging sequences in the presurgical differentiation of brain lymphoma and glioblastoma. Eur Radiol. 2023;33(12):8800–8.
    Article PubMed CAS Google Scholar
  21. Boxerman JL, Quarles CC, Hu LS, et al. Consensus recommendations for a dynamic susceptibility contrast MRI protocol for use in high-grade gliomas. Neuro Oncol. 2020;22(9):1262–75.
    Article PubMed PubMed Central Google Scholar
  22. Ozturk K, Soylu E, Cayci Z. Differentiation between primary CNS lymphoma and atypical glioblastoma according to major genomic alterations using diffusion and susceptibility-weighted MR imaging. Eur J Radiol. 2021;141:109784.
    Article PubMed Google Scholar
  23. Cheong EN, Park JE, Park SY, Jung SC, Kim HS. Achieving imaging and computational reproducibility on multiparametric MRI radiomics features in brain tumor diagnosis: Phantom and clinical validation. Eur Radiol. 2024;34(3):2008–23.
    Article PubMed Google Scholar
  24. Louis DN, Perry A, Wesseling P, et al. The 2021 WHO classification of tumors of the central nervous system: a summary. Neuro Oncol. 2021;23(8):1231–51.
    Article PubMed PubMed Central CAS Google Scholar
  25. Tofts PS. Modeling tracer kinetics in dynamic Gd-DTPA MR imaging. J Magn Reson Imaging. 1997;7(1):91–101.
    Article PubMed CAS Google Scholar
  26. Orton MR, d’Arcy JA, Walker-Samuel S, et al. Computationally efficient vascular input function models for quantitative kinetic modelling using DCE-MRI. Phys Med Biol. 2008;53(5):1225–39.
    Article PubMed Google Scholar
  27. Parker GJ, Roberts C, Macdonald A, et al. Experimentally-derived functional form for a population-averaged high-temporal-resolution arterial input function for dynamic contrast-enhanced MRI. Magn Reson Med. 2006;56(5):993–1000.
    Article PubMed Google Scholar
  28. Weinmann HJ, Laniado M, Mützel W. Pharmacokinetics of gddtpa/dimeglumine after intravenous injection into healthy volunteers. Physiol Chem Phys Med NMR. 1984;16(2):167–72.
    PubMed CAS Google Scholar
  29. Boxerman JL, Schmainda KM, Weisskoff RM. Relative cerebral blood volume maps corrected for contrast agent extravasation significantly correlate with glioma tumor grade, whereas uncorrected maps do not. AJNR Am J Neuroradiol. 2006;27(4):859–67.
    PubMed PubMed Central CAS Google Scholar
  30. Eisenhut F, Schmidt MA, Putz F, et al. Classification of primary cerebral lymphoma and glioblastoma featuring dynamic susceptibility contrast and apparent diffusion coefficient. Brain Sci. 2020;10(11):886.
    Article PubMed PubMed Central CAS Google Scholar
  31. Lin X, Lee M, Buck O, et al. Diagnostic accuracy of T1-Weighted dynamic Contrast-Enhanced-MRI and DWI-ADC for differentiation of glioblastoma and primary CNS lymphoma. AJNR Am J Neuroradiol. 2017;38(3):485–91.
    Article PubMed PubMed Central CAS Google Scholar
  32. Hagiwara A, Oughourlian TC, Cho NS, et al. Diffusion MRI is an early biomarker of overall survival benefit in IDH wild-type recurrent glioblastoma treated with immune checkpoint inhibitors. Neuro Oncol. 2022;24(6):1020–8.
    Article PubMed CAS Google Scholar
  33. Hambardzumyan D, Bergers G. Glioblastoma: defining tumor niches. Trends Cancer. 2015;1(4):252–65.
    Article PubMed PubMed Central Google Scholar
  34. Lu S, Gao Q, Yu J, 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(10):1722–7.
    Article PubMed Google Scholar
  35. Tofts PS, Brix G, Buckley DL, et al. Estimating kinetic parameters from dynamic contrast-enhanced T(1)-weighted MRI of a diffusable tracer: standardized quantities and symbols. J Magn Reson Imaging. 1999;10(3):223–32.
    Article PubMed CAS Google Scholar
  36. Choi YS, Lee HJ, Ahn SS, 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(4):1344–51.
    Article PubMed Google Scholar
  37. Lee MD, Baird GL, Bell LC, Quarles CC, Boxerman JL. Utility of percentage signal recovery and baseline signal in DSC-MRI optimized for relative CBV measurement for differentiating glioblastoma, lymphoma, metastasis, and meningioma. AJNR Am J Neuroradiol. 2019;40(9):1445–50.
  38. Sugita Y, Takase Y, Mori D, et al. Endoglin (CD 105) is expressed on endothelial cells in the primary central nervous system lymphomas and correlates with survival. J Neurooncol. 2007;82(3):249–56.
    Article PubMed CAS Google Scholar
  39. Xing Z, Kang N, Lin Y, et al. Performance of diffusion and perfusion MRI in evaluating primary central nervous system lymphomas of different locations. BMC Med Imaging. 2020;20(1):62.
    Article PubMed PubMed Central Google Scholar
  40. Suh HB, Choi YS, Bae S, et al. Primary central nervous system lymphoma and atypical glioblastoma: differentiation using radiomics approach. Eur Radiol. 2018;28(9):3832–9.
    Article PubMed Google Scholar

Download references

Acknowledgements

Not applicable.

Funding

This work was supported by the National Natural Science Foundation of China (No. 82371905), Joint Funds for the Innovation of Science and Technology, Fujian Province (No. 2021Y9154 and 2021Y9093).

Author information

Author notes

  1. Lan Yu and Shujie Yu: The first two authors contributed equally to this manuscript and should be the co-first authors.
  2. Dairong Cao and Zhen Xing contributed equally to this study.

Authors and Affiliations

  1. Department of Radiology, The First Affiliated Hospital, Fujian Medical University, 20 Cha-Zhong Road, Fuzhou, Fujian, 350005, P.R. China
    Lan Yu, Shujie Yu, Feng Wang, Xiaofang Zhou, Feiman Yang, Dairong Cao & Zhen Xing
  2. Department of Radiology, National Regional Medical Center, Binhai Campus of the First Affiliated Hospital, Fujian Medical University, Fuzhou, Fujian, 350212, China
    Lan Yu, Shujie Yu, Feng Wang, Xiaofang Zhou, Feiman Yang, Dairong Cao & Zhen Xing
  3. Department of Radiology, Fujian Key Laboratory of Precision Medicine for Cancer, The First Affiliated Hospital, Fujian Medical University, Fuzhou, 350005, China
    Dairong Cao
  4. Key Laboratory of Radiation Biology of Fujian Higher Education Institutions, The First Affiliated Hospital, Fujian Medical University, Fuzhou, 350005, China
    Dairong Cao

Authors

  1. Lan Yu
  2. Shujie Yu
  3. Feng Wang
  4. Xiaofang Zhou
  5. Feiman Yang
  6. Dairong Cao
  7. Zhen Xing

Contributions

All authors contributed to the study’s conception and design. Data collection and analysis were performed by Lan Yu, Shujie Yu, Feng Wang, Xiaofang Zhou and Feiman Yang. Lan Yu and Shujie Yu contributed equally to this study. Dairong Cao and Zhen Xing coordinated project activities. Dairong Cao and Zhen Xing contributed equally to this study. The first draft of the manuscript was written by Lan Yu and all authors commented on previous versions of the manuscript. All authors read and approved the final manuscript.

Corresponding authors

Correspondence toDairong Cao or Zhen Xing.

Ethics declarations

The study was approved by the Institutional Review Board of the First Affiliated Hospital of Fujian Medical University. Individual consent was waived due to the retrospective study design. All procedures performed in this study involving human participants were in accordance with the Declaration of Helsinki.

Not applicable.

Competing interests

The authors declare no competing interests.

Additional information

Publisher’s note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary Information

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Reprints and permissions

About this article

Cite this article

Yu, L., Yu, S., Wang, F. et al. Multiparametric MRI for differential diagnosis of primary central nervous system lymphoma and atypical glioblastoma: an analysis incorporating DWI, DCE-MRI, and contrast agent preload DSC-PWI.BMC Med Imaging 25, 345 (2025). https://doi.org/10.1186/s12880-025-01886-9

Download citation

Keywords