Gliomas possess complex and heterogeneous vasculatures with abnormal hemodynamics. Despite considerable advances in diagnostic and therapeutic techniques for improving tumor management and patient care in recent years, the prognosis of malignant gliomas remains dismal. Perfusion-weighted magnetic resonance imaging techniques that could noninvasively provide superior information on vascular functionality have attracted much attention for evaluating brain tumors. However, nonconsensus imaging protocols and postprocessing analysis among different institutions impede their integration into standard-of-care imaging in clinic. And there have been very few studies providing a comprehensive evidence-based and systematic summary. This review first outlines the status of glioma theranostics and tumor-associated vascular pathology and then presents an overview of the principles of dynamic contrast-enhanced MRI (DCE-MRI) and dynamic susceptibility contrast-MRI (DSC-MRI), with emphasis on their recent clinical applications in gliomas including tumor grading, identification of molecular characteristics, differentiation of glioma from other brain tumors, treatment response assessment, and predicting prognosis. Current challenges and future perspectives are also highlighted.
Gliomas are the most common primary brain tumors in adults with varying malignancy ranging from pilocytic astrocytoma to glioblastoma multiforme (GBM) [
Magnetic resonance imaging (MRI) is currently the prior choice for clinical applications in brain tumors [
The versatile clinical applications of contrast-enhanced perfusion MRI techniques in gliomas.
Despite that numerous studies have explored PW-MRI for evaluating gliomas, there have been very few studies providing a comprehensive evidence-based and systematic summary. This review first outlines the status of glioma theranostics and tumor-associated vascular pathology and then presents an overview of the principles of DCE-MRI and DSC-MRI, with emphasis on their recent clinical applications in gliomas including tumor grading, identification of molecular characteristics, differentiation of glioma from other brain tumors, treatment response assessment, and predicting prognosis. Current challenges and future perspectives are also highlighted.
Malignant gliomas possess exuberant neovascularization characterized by disorganized, irregular, and tortuous vessels with arteriovenous shunting [
Cerebral vascular hemodynamics can be assessed with PW-MRI, including DCE-MRI, DSC-MRI, and arterial spin-labeling (ASL) techniques. Using exogenous gadolinium-based contrast agents (GBCAs), PW-MRI can characterize tumor vascular perfusion and permeability with multiple parameters, by emphasizing either the T1 relaxivity properties of GBCAs through T1-weighted DCE-MRI or their susceptibility effects through T2/T2
DCE-MRI is based on T1 relaxivity of GBCAs with fast imaging acquisition. Due to the BBB disruption and vascular hyperpermeability in gliomas, the GBCAs administered intravenously are easy to leak from intravascular compartment to extravascular extracellular space (EES), leading to an increase in T1 signal intensity induced by paramagnetic effect [
Model-free parameters are calculated based on signal intensity-acquisition time curve, reflecting an overall kinetics of GBCAs perfusion (Figure
An illustration of parameters derived from DCE-MRI and DSC-MRI. (a) Semiquantitative parameters from signal intensity curve in DCE-MRI. (b) Schematic diagram of ETK model from DCE-MRI. (c) Calculation of PSR and PH from DSC-MRI. (d) Contrast concentration-time course curve of DSC-MRI. CBV is proportional to determined area under contrast concentration-time course curve (blue shaded area), and CBF is easily calculated given the relationship of MTT and CBV.
Model-dependent parameters can be calculated by fitting various mathematical PK models. Common-used PK models for brain tumors include classic Tofts-Kermode (TK) model and extended TK (ETK) model [
Main perfusion parameters derived from DCE-MRI and DSC-MRI.
Parameters | Full name | Definition and meaning |
---|---|---|
|
||
|
Volume transfer constant between blood plasma and EES | It describes the leakage rate of GBCAs from the blood plasma towards EES |
|
Extravascular extracellular volume fraction | Quantification of cellularity and necrosis in EES. |
|
Blood plasma volume | Quantification of the volume of blood plasma |
|
Transfer constant from EES into blood plasma | It is determined by the equation |
|
||
CBV | Cerebral blood volume | The blood volume in a given region of brain tissue (unit, mL/100 g). It is calculated by integrating the area under the CC-TCC |
CBF | Cerebral blood flow | The blood volume passing through a given region of brain tissue per unit of time (unit, mL/min/100 g) |
MTT | Mean transit time | The average time in which blood passes through a given region of brain tissue (unit, s). It is estimated from the CC-TCC as width of the curve at half maximum height |
PH | Peak height | The maximal drop of signal intensity from precontrast baseline during the first-pass bolus phase of GBCAs. It is correlated with CBV and reflects total blood volume |
PSR | Percentage of signal intensity recovery | It reflects capillary permeability indirectly, providing information like |
rCBV | Relative cerebral blood volume | Measurement of the relative lesion blood volume compared with that of contralateral white matter. It is proportional to the area under the CC-TCC, providing an estimate of MVD and angiogenesis |
rCBF | Relative cerebral blood flow | Measurement of the relative lesion blood flow compared with that of contralateral white matter |
|
Leakage coefficient | Quantification of the degree of vascular permeability using algorithm method for leakage effect correction |
DSC-MRI is based on a dynamic series acquisition of T2/T2
Although promising in vascular perfusion evaluation, DSC-MRI has some limitations. The T2
Accurate glioma grading is of great importance for clinical decision making and personalized management. Histopathologic biopsy is currently the gold standard for glioma grading in clinical practice. However, it encounters inherent sampling bias, invasive procedure, and interobserver variability. Moreover, biopsy specimen may not be representative of the tumor panorama characteristics due to the improper resection and intratumoral heterogeneity. It is crucial to establish an accurate diagnosis without biopsy if (
Examples of perfusion MRI for glioma grading.
Study (year) (ref) | Group ( |
Average age (years) | Imaging modality (method or model; parameter analysis) | Indexes | Results | Limitations |
---|---|---|---|---|---|---|
Maia et al. (2005) [ |
Grade II (13) |
36 | DSC-MRI (leakage effect uncorrected; ROI-based analysis) | rCBV | Positive correlation between rCBV and tumor grade and VEGF expression | Impact of leakage effect on rCBV accuracy; small sample size |
|
||||||
Boxerman et al. (2006) [ |
Grade II (11) |
52 | DSC-MRI (algorithm for leakage correction; ROI-based analysis) | rCBV | Significant correlation between tumor grade and corrected rCBV | rCBV threshold to discriminate tumor grade was not provided |
|
||||||
Law et al. (2007) [ |
Grade II (31) |
43 | DSC-MRI ( |
rCBV | Positive correlation between all parameters and tumor grade; more specific than rCBV |
Histogram was based on whole tumor ROI probably including normal brain tissues |
|
||||||
Emblem et al. (2008) [ |
LGG (24) |
52 | DSC-MRI ( |
rCBV | Increased diagnosis accuracy and interobserver agreement were obtained using histogram method | Only the peak height of histogram distribution was measured |
|
||||||
Server et al. (2011) [ |
Grade II (18) |
57 | DSC-MRI (algorithm for leakage correction; ROI-based analysis) | rCBV |
All parameters were correlated with tumor grade; the diagnostic power of rCBV was better than |
Influence of steroid treatment on correlation between |
|
||||||
Yoon et al. (2014) [ |
LGG (12) |
50 | DSC-MRI ( |
rCBV | Significant difference of rCBV between HGG and LGG | Subjectivity and neglect of the heterogeneity using ROI-based analysis |
|
||||||
Aprile et al. (2015) [ |
HGG (31) |
55 | DSC-MRI (preload for leakage correction; ROI-based analysis) | PSR |
Both the two parameters were significantly different between LGG and HGG; PSR was better than rCBV for grading | The relative small sample number of grade III glioma |
|
||||||
Smitha et al. (2015) [ |
HGG (25) |
38 | DSC-MRI (leakage effect uncorrected; ROI-based analysis) | rPSR |
Positive correlation between all parameters and tumor grade; the diagnosis performance of rPSR was better than rCBV and rCBF | Impact of leakage effect on rCBV accuracy |
|
||||||
Choi et al. (2013) [ |
LGG (10) |
51 | DCE-MRI (ETK model; ROI-based analysis), DWI |
|
Significant difference in |
Small sample size |
|
||||||
Zhao et al. (2015) [ |
LGG (9) |
46 | DCE-MRI (TK model; ROI-based analysis), DWI |
|
Significant difference of all parameters between LGG and HGG; |
Small sample size; lack of correlation between histopathology and imaging biomarkers |
|
||||||
Jung et al. (2014) [ |
Grade II (7) |
49 | DCE-MRI (ETK model; histogram analysis) |
|
Positive correlation between all parameters and tumor grade | Small sample size of LGG; lack the percentile of parameters ranging from 0 to 50 |
|
||||||
Li et al. (2015) [ |
Grade II (15) |
42 | DCE-MRI (TK model; ROI-based analysis), SWI |
|
All parameters could distinguish tumor grade except for grade III and grade IV | Small sample size; lack of voxel-to-voxel correlation between imaging features and pathological specimens |
|
||||||
Nguyen et al. (2015) [ |
Grade II (9) |
57 | DCE-MRI (ETK model, phase-derived AIF; ROI-based analysis), DSC-MRI (bookend method; ROI-based analysis) |
|
Significant difference between all parameters and tumor grade; improved diagnostic power of parameters using phase-derived AIF method | Only 2 flip angles were used for estimation of the precontrast T1 map; sampling error of histopathological biopsy; some patients received steroids before imaging |
|
||||||
Santarosa et al. (2016) [ |
Grade II (9) |
55 | DCE-MRI (ETK model; histogram/ROI-based analysis), DSC-MRI (algorithm for leakage correction; histogram/ROI-based analysis) |
|
Significant difference of all parameters between HGG and LGG; histogram analysis is better than ROI-based method | Small sample size |
Early studies demonstrated that increased rCBV was correlated with more active angiogenesis and aggressive tumor malignancy, being a potential imaging biomarker for preoperative tumor grading [
Increased vascular permeability is another predominant characteristic of tumor vessels, playing an adjuvant role for glioma grading. PSR was found to be inversely correlated with vascular permeability [
In spite of serving as potential imaging biomarkers for glioma grading, the perfusion parameters are overlapped to some extent among different tumor grades. The thresholds of perfusion indexes, specificity, and sensitivity from different institutions vary considerably, making the comparison difficult. This may be partly attributed to the difference in sample sizes, enrollment criteria, and especially imaging methods. Although there have been a variety of imaging strategies (e.g., bookend technique and phase-derived arterial input function) for improving the accuracy and reproducibility of indexes estimation, standardization and improvement of the imaging acquisition methodology are indispensable for further clinical applications.
Recent in-depth molecular/genetic investigations have led to a profound shift in glioma theranostics based on the substantial progress in genetic alteration profiles. The latest 2016 WHO classification for central nervous system (CNS) tumors integrates the molecular/genetic criteria into histological diagnostics [
Examples of perfusion MRI for identifying molecular characterization.
Study (year) (ref) | Group ( |
Average age (year) | Imaging modality (method or model; parameter analysis) | Indexes | Results | Limitations |
---|---|---|---|---|---|---|
Kickingereder et al. (2015) [ |
Grades II and III: |
49 | DSC-MRI (algorithm for leakage correction; histogram analysis) | rCBV | rCBV was significantly different between IDH mutation and wild-type tumors | Only including grades II and III tumors |
|
||||||
Lee et al. (2015) [ |
HGG: |
50 | DSC-MRI (algorithm for leakage correction; histogram analysis) | nCBV | Significant difference of nCBV between IDH mutation and wildtype; higher heterogeneity in mutation tumor than the wild-type | Not including LGG. Not excluding influence of MGMT mutation |
|
||||||
Tykocinski et al. (2012) [ |
GBM: |
61 | DSC-MRI (preload for leakage correction; ROI-based analysis) | rCBV | Strong correlation between rCBV and EGFRvIII status | Relative small sample size of EGFRvIII-positive tumors |
|
||||||
Gupta et al. (2015) [ |
GBM: |
66 | DSC-MRI ( |
PSR |
Higher rCBV and lower PSR were associated with EGFR |
Pathologic sampling may not be consistent with ROI selection |
|
||||||
Arevalo-Perez et al. (2015) [ |
GBM: |
66 | DCE-MRI (ETK model; histogram analysis) |
|
Strong correlation between both parameters and EGFRvIII status; |
Not eliminating influence of other molecular mutations |
|
||||||
Jung et al. (2013) [ |
GBM: |
52 | DSC-MRI ( |
nCBV | nCBV was higher in MGMT-negative tumors than in MGMT-positive tumors | Small sample size |
|
||||||
Moon et al. (2012) [ |
HGG: |
51 | DSC-MRI (leakage effect uncorrected; ROI-based analysis) DTI | rCBV |
No significant correlation between rCBV and MGMT | Small sample size; impact leakage effect of rCBV accuracy |
|
||||||
Ahn et al. (2014) [ |
GBM: |
58 | DCE-MRI (TK model; ROI-based analysis); DTI |
|
Only |
Subjectivity of ROI-based method |
|
||||||
Jenkinson et al. (2006) [ |
Grades II and III: |
44 | DSC-MRI (leakage effect uncorrected; ROI-based analysis) | rCBV | rCBV was associated with 1p/19q codeletion in oligodendroglioma | Subjectivity of ROI-based method; impact leakage effect of rCBV accuracy |
|
||||||
Emblem et al. (2008) [ |
Grades II and III: |
52 | DSC-MRI (algorithm for leakage correction; histogram analysis) | rCBV | Histogram analysis of rCBV could differentiate 1p/19q genotype in astrocytic and oligodendroglial tumors | Small sample size; only the peak height of histogram distribution was assessed |
IDH (IDH-1/IDH-2) enzymes catalyze isocitrate oxidative decarboxylation to form
Water suppressed proton-magnetic resonance spectroscopy (1H-MRS) has been explored to noninvasively detect 2-HG in gliomas for identification of IDH-1 gene mutation [
DSC-MRI for identification of IDH mutation status in GBM. Six sets of representative FLAIR and corresponding rCBV images from IDH1/2 mutant and wild-type GBM. Histogram analysis demonstrates that IDH1/2 mutant tumors have substantially lower rCBV value than the wild-type. Reproduce with permission from Kickingereder et al. [
EGFR is a transmembrane glycoprotein belonging to receptor tyrosine kinase (RTK) family [
Previous studies showed that higher contrast enhancement volume and enhancement/necrosis ratio on conventional MRI were associated with EGFR overexpression [
MGMT is a ubiquitous DNA repair enzyme in glioma cells. The MGMT promoter methylation could induce epigenetic silencing of this gene and consequently result in DNA damage and cell death [
Some conventional imaging features (such as enhancement pattern, tumor margin characteristic and T2/FLAIR signal intensity) appear to be associated with MGMT promoter methylation status but have some discrepancies among institutions [
The unbalanced translocation between chromosome arm 1p and 19q results in loss of heterozygosity (LOH) [
Jenkinson et al. [
Above-mentioned studies demonstrate that PW-MRI parameters hold great potential implications for reflecting glioma-associated molecular characteristics. However, given the intrinsic limitations of PW-MRI imaging technique, the physiologic description or significance of perfusion parameters is intricate at molecular level and is difficult to recapitulate a certain molecule/gene characterization. For example, EGFR amplification and mutation can result in the overexpression of various downstream effector molecules such as VEGF, interleukin-18, and angiopoietin-like 4 to make synergic effect on tumor neovascularization, consequently altering the vascular structure and function [
The therapeutics and prognosis of different CNS tumors are of extreme disparity. Preoperative differentiation of gliomas from other brain tumors is important for preoperative staging, intraoperative management, and postoperative treatment. Conventional MRI cannot provide pathophysiological information for identifying glioma, solitary brain metastasis (MET), and primary central nervous system lymphoma (PCNSL), due to their similar imaging performance such as space-occupying and enhancing patterns [
Differential diagnosis in glioma, metastasis, and PCNSL.
Study (year) (ref) | Tumor type ( |
Average age (year) | Imaging modality (method or model; parameter analysis) | Indexes | Results | Limitations |
---|---|---|---|---|---|---|
Law et al. (2002) [ |
HGG (24) |
52 | DSC-MRI (leakage effect uncorrected; ROI-based analysis) | rCBV | rCBV in peritumoral region was significantly different between HGG and MET | The peritumoral region was not defined clearly; the threshold value was not provided |
|
||||||
Cha et al. (2007) [ |
GBM (27) |
52 | DSC-MRI (alteration of |
PSR |
Significant difference of all parameters between GBM and MET; PSR was the most powerful with 100% specificity | Small sample size; some cases were not confirmed by histopathology |
|
||||||
Mangla et al. (2011) [ |
GBM (22) |
54 | DSC-MRI (preload for leakage correction; ROI-based analysis) | rCBV |
PSR was better than rCBV for differentiation | Small sample size; impact of steroid treatment on parameter evaluation |
|
||||||
Toh et al. (2013) [ |
GBM (20) |
60 | DSC-MRI (algorithm for leakage correction; ROI-based analysis) | rCBV |
Uncorrected rCBV is much better for differentiating | Lack of direct correlation between parameters and histopathologic features |
|
||||||
Xing et al. (2014) [ |
HGG (26) |
51 | DSC-MRI (leakage effect uncorrected; ROI-based analysis) | rCBV |
The combination of rCBV with PSR might help in more accurate differentiation | Impact of leakage effect on parameter measurements |
|
||||||
Kickingereder et al. (2014) [ |
GBM (60) |
N/A | DCE-MRI (TK model; ROI-based analysis) |
|
|
Relative small sample size of PCNSL |
|
||||||
Kickingereder et al. (2014) [ |
GBM (28) |
66 | DSC-MRI (preload for leakage correction; ROI-based analysis), DWI, SWI | rCBV |
Multiparametric MRI allowed differentiation of GBM from PCNSL | Small sample size |
|
||||||
Zhao et al. (2015) [ |
LGG (9) |
46 | DCE-MRI (TK model; ROI-based analysis) |
|
All parameters were significantly different between LGG, HGG, and MET. IAUC had the most diagnostic power | Small sample size; subjectivity of ROI selection |
|
||||||
Jung et al. (2016) [ |
GBM (26) |
N/A | DCE-MRI (ETK model, ROI-based analysis) |
|
Semiquantitative parameters could differentiate between GBM and hypovascular metastasis | Subjectivity of ROI selection |
GBM and metastatic brain tumor are the two most common malignant intracranial tumors representing similar imaging appearances and enhancing patterns on conventional MRI [
DSC-MRI could differentiate subtle differences of vascular perfusion. Higher rCBV
DSC-MRI (a) and DCE-MRI (b) for differentiation of GBM, PCNSL, and metastasis. rCBV maps demonstrate different characteristic features in the three distinct entities, with significantly higher rCBV value of GBM compared with metastasis and PCNSL. The
Although PW-MRI provides valuable information for antidiastole between gliomas and solitary brain metastases, it is undeniable that the threshold of indexes for diagnosis varies among the studies because of different origin of metastases except for various imaging acquisitions. More importantly, DCE-MRI is weak in differentiating GBM and highly vascular brain metastasis such as melanoma metastasis on account of their similar vascular function. DWI-derived ADC value could be an alternative and complementary imaging biomarker to differentiate the two tumor entities [
PCNSL is a rare neoplasm constituting up to 6% of intracranial malignant tumors [
Higher rCBV and lower PSR were suggestive of GBM (Figure
The current standard of care for GBM is concomitant and adjuvant chemoradiotherapy following maximum safe surgical resection. The treatment options are influenced by various factors and need to be timely adjusted at different stages of care. Accurate treatment response assessment is greatly important to clinical decision making and personalized medicine. Macdonald Criteria is based on treatment response assessment via evaluation of the contrast-enhancing areas on MRI [
Approximately up to 50% of glioma patients treated with chemoradiotherapy can develop transient new areas of increasing contrast enhancement or edema, termed pseudoprogression (PsP), which is easily confounded with true progressive disease (PD) [
Differentiation of pseudoprogression from true progression.
Study (year) (ref) | Group ( |
Average age (year) | Imaging modality (method or model; parameter analysis) | Indexes | Threshold (Sp%, Sn%) | limitations |
---|---|---|---|---|---|---|
Mangla et al. (2010) [ |
PsP (7) |
61 | DSC-MRI (algorithm for leakage effect correction; ROI-based analysis) | rCBV | Percentage change in rCBV for discrimination of PsP and TP (85.7%, 76.9%) | Retrospective; different treatment management; small sample size |
|
||||||
Martínez-Martínez and Martínez-Bosch (2014) [ |
PsP (17) |
48 | DSR-MRI (leakage effect uncorrected; ROI-based analysis) | rCBV |
rPH = 1.37 (82.2%, 88%) |
Retrospective; small sample size; impact of corticoid therapy on parameter evaluation; lack of histological confirmation |
|
||||||
Prager et al. (2015) [ |
PsP (8) |
55 | DSC-MRI ( |
rCBV |
rCBV |
Retrospective; small sample size of PsP; MGMT in some patients may affect the perfusion parameters |
|
||||||
Baek et al. (2012) [ |
PsP (37) |
49 | DSC-MRI ( |
nCBV | Percent change of skewness: 1.27% (79.2%, 85.7%) |
Different therapies in patients; results were obtained from only one observer |
|
||||||
Tsien et al. (2010) [ |
PsP (13) |
52 | DSC-MRI (leakage effect uncorrected, parametric response map) | rCBV |
Not provided; patients with progressive had reduced rCBV | Leakage effect may underestimate rCBV value |
|
||||||
Gahramanov et al. (2013) [ |
PsP (9) |
N/A | DSC-MRI (ferumoxytol for leakage correction) | rCBV | rCBV = 1.5 (Sp%, Sn% not provided) | Lack of histopathologic confirmation; small sample size |
|
||||||
Suh et al. (2013) [ |
PsP (36) |
50 | DCE-MRI (nonmodel fitting; histogram analysis) | AUCR |
mAUCR |
Lack of correlation between imaging measurements and specimen histology |
|
||||||
Yun et al. (2015) [ |
PsP (16) |
55 | DCE-MRI (ETK model; histogram analysis) |
|
|
Small relative sample size; lack of histological confirmation |
PD demonstrated higher rCBV and lower PSR, while PsP exhibited decreased rCBV and rPH [
Discrimination of PsP from PD using DSC-MRI and DCE-MRI. (a) Contrast-enhanced T1WI of GBM treated with temozolomide demonstrates increased contrast enhancement suspicious for both PsP
DSC-MRI has intrinsic sensitivity to susceptibility artifact, commonly caused by posttreatment hemorrhage and calcification [
While a number of studies have employed PW-MRI to discriminate PsP from PD in GBM, cut-off values of parameters with specificity and sensitivity across institutions are somewhat different even not comparable because of small sample size, as well as lack of standardization of imaging protocols and accordant inclusive criterion of individuals. Accuracy and reproducibility of perfusion parameters are inevitably affected by technical aspects (e.g., leakage correction, types of GBCAs, and PK model fitting) and parameter analysis (e.g., ROI-based/histogram analysis and parametric response map). The inclusion of patients who have already received corticoid therapy may bias the results of parameters evaluation. In addition, the initial and end timing for imaging monitoring, types, and doses of drug are inconsistent. Therefore, more well-controlled studies and coregistration of PW-MRI with corresponding histological mapping are urgently needed for reconfirmation of these results.
Antiangiogenic therapies (such as bevacizumab and cediranib) could induce early decrease in contrast enhancement and edema on conventional MRI due to the restored BBB integrity and reduced endothelial permeability, resulting in prolonged progress-free survival (PFS) but modest benefit of overall survival (OS) [
PW-MRI may help differentiate true response from PD by predicting OS. A multicenter trial investigated the efficacy of standardized rCBV (sRCBV) and mean tumor rCBV normalized to white matter (nRCBV) for predicting OS in recurrent GBM after treatment initiation [
Due to the diverse imaging protocols applied, the use of standardized parameters (sRCBV) and model-free parameters (IAUC) could be alternative to reduce variability and improve accuracy and reproducibility when comparing results from multiple institutions or using different acquisition strategies.
The current standard response assessment of glioma is lined with the RANO criteria, especially including the abnormal hyperintensity of T2/FLAIR in nonenhancing regions [
Radiation-induced brain injuries are mainly classified into three stages based on the occurrence time: acute (during radiation), subacute (within 3 months after radiation), and late (months to years after radiation). The acute and early subacute injuries are mainly caused by vasodilation, BBB disruption, and edema, usually present as relatively unchanged MR appearance [
Discrimination of recurrent glioma from radiation necrosis.
Study (year) (ref) | Group ( |
Average age (year) | Imaging modality (method or model; parameter analysis) | Indexes | Threshold (Sp%, Sn%) | Limitations |
---|---|---|---|---|---|---|
Barajas et al. (2009) [ |
RN (17) |
54 | DSC-MRI (alteration of |
rCBV |
rPH = 1.38 (81.38%, 89.32%) |
Impact of partial volume averaging effect on parameter evaluation |
|
||||||
Hu et al. (2009) [ |
rHGG (24) |
47 | DSC-MRI (baseline subtraction method for leakage correction; ROI-based analysis) | rCBV | rCBV = 0.71 (100%, 91.7%) | Various tumor types; inconsistent radiation dose and different therapies |
|
||||||
Bisdas et al. (2011) [ |
rHGG (12) |
N/A | DCE-MRI (TK model; ROI-based analysis) |
|
|
Small sample size; lack of histopathologic confirmation in some cases |
|
||||||
Shin et al. (2014) [ |
Recurrent glioma (19) |
55 | DCE-MRI (TK model; ROI-based analysis), DSC-MRI (preload for leakage corrected; ROI-based analysis) | r |
rCBV = 2.33 (70%, 72.2%) |
Relative small sample size; ROI-based method was not comprehensive |
|
||||||
Larsen et al. (2013) [ |
Recurrent glioma (11) |
56 | DCE-MRI (deconvolution technique) | CBV | CBV = 2.0 ml/100 g (100%, 100%) | Small sample size; sample bias in histological analysis; various tumor types |
|
||||||
Masch et al. (2016) [ |
Recurrent glioma (16) |
51 | DSC-MRI (preload for leakage correction; ROI-based analysis) | rCBV | Not provided; elevated rCBV in recurrent lesion compared with RN | Various tumor types; lack of histological confirmation in some cases |
Several studies demonstrated that recurrent glioma owned higher rCBV and lower PSR compared with radiation injury [
Discrimination of RN from recurrent GBM using DCE-MRI (a) and DSC-MRI (b). Contrast-enhanced T1WI demonstrates similar contrast enhancement in recurrent glioblastoma
Nonstandardized imaging acquisition renders a wide range of sensitivity and specificity using PW-MRI. Some other factors such as different inclusive criteria, tumor grades and radiation timing, and dose may also disturb the diagnostic accuracy of perfusion parameters. Moreover, histopathological validation lacks in published studies. Further work on significant improvement of imaging method and correlation between the imaging and histologic features is warranted to draw a definite conclusion.
Initial patient stratification is clinically important for optimized and individualized therapeutic regimens. Multiple efforts are ongoing for survival prediction in glioma patients. Glioma is characterized by abnormal vasculature with active angiogenesis. Perfusion MRI techniques providing physiologic information have been widely investigated for noninvasive prognosis prediction in glioma patients.
rCBV has demonstrated predictive value for gliomas regardless of treatment [
Contrast-enhanced PW-MRI techniques are becoming increasingly common approaches for clinical applications in gliomas. They facilitate better understanding of a variety of hemodynamic pathologies and the underlying mechanisms of tumor neovascularization. However, there are still some unresolved issues when implementing PW-MRI in contemporary radiology practice. We noted that perfusion parameters are inevitably influenced by various hemodynamic factors, types of GBCAs, and total acquisition time. For example,
Perfusion parameters are affected by a complex interaction of factors. In multicenter clinical trials, even minor differences of benchmarked standards may result in significant changes in perfusion parameters. These variables include (
Despite some clinical limitations and unsolved issues, the current evidence available demonstrated the tremendous foreground of PW-MRI for improving glioma management. Imaging protocols standardization is urgently demanded for accelerating the translation of PW-MRI into routine clinical applications. For DSC-MRI, sustained and focused efforts on exploiting novel imaging sequences, contrast agents, and better algorithm to maximally eliminate T1 and T2
Glioblastoma multiforme
Perfusion-weighted magnetic resonance imaging
Dynamic contrast-enhanced MRI
Dynamic susceptibility contrast-MRI
Glioblastoma multiforme
Fluid attenuated inversion recovery
Vascular endothelial growth factor
Low grade glioma
Blood brain barrier
High grade glioma
Microvessel density
Endothelial cells
Arterial spin-labeling
Gadolinium-based contrast agents
Extravascular extracellular space
Initial area under the concentration-time curve
Area under the concentration-time curve
Peak signal intensity
Tofts-Kermode
Extended Tofts-Kermode
The volume transfer constant between blood plasma and EES
The volume of EES per unit volume of tissue
Rate constant between EES and blood plasma
Fractional plasma volume per unit of tissue volume
Spin echo-echo planar imaging
Gradient echo-echo planar imaging
Signal intensity-time course curve
Contrast concentration-time course curve
Cerebral blood volume
Cerebral blood flow
Peak height
Mean transit time
Percentage of signal intensity recovery
Relative CBV
Region of interest
Permeability surface-area product
Central nervous system
Isocitrate dehydrogenase
Epidermal growth factor receptor
EGFR amplification
Methyl-guanine methyltransferase
2-Hydroxyglutarate
Water suppressed proton-magnetic resonance spectroscopy
Normalized CBV
Apparent diffusion coefficient
EGFR variant III
Chimeric antigen receptor T-cell
Relative peak height
Reverse-transcription polymerase chain reaction
Loss of heterozygosity
Metastasis
Primary central nervous system lymphoma
Intratumoral susceptibility signals
Response Assessment in Neuro-Oncology
Parametric response map
The initial and final area under the time-signal intensity curves ratio
The mean AUCR at a higher curve
The 50th cumulative AUCR histogram parameter
rCBV normalized to white matter
Standardized rCBV
Pseudoprogression
Progressive disease
Radiation necrosis
Percentage of specificity
Percentage of sensitivity
Apparent transfer constant
Progress-free survival
Overall survival
Relative cerebral blood flow
Brain Tumor Imaging Protocol
American Society of Functional Neuroradiology
Quantitative Imaging Biomarkers Alliance of the Radiological Society of North American.
The authors declared that they have no conflicts of interest.
Junfeng Zhang and Heng Liu contributed equally to this work.
This work was supported by National Natural Science Foundation of China (nos. 81271626 and 81511660), Chongqing Science and Technology R&D Base Construction (International Cooperation) Project (cstc2014gjhz110002), and Clinical Scientific Research Fund of Daping Hospital, Institute of Surgery Research, Third Military Medical University (2014YLC03).