Accuracy of percentage of signal intensity recovery and relative cerebral blood volume derived from dynamic susceptibility-weighted, contrast-enhanced MRI in the preoperative diagnosis of cerebral tumours

Abstract

Conventional magnetic resonance imaging (MRI) is the technique of choice for diagnosis of cerebral tumours, and has become an increasingly powerful tool for their evaluation; however, the diagnosis of common contrast-enhancing lesions can be challenging, as it is sometimes impossible to differentiate them using conventional imaging. Histopathological analysis of biopsy specimens is the gold standard for diagnosis; however, there are significant risks associated with the invasive procedure and definitive diagnosis is not always achieved. Early accurate diagnosis is important, as management differs accordingly. Advanced MRI techniques have increasing utility for aiding diagnosis in a variety of clinical scenarios. Dynamic susceptibility-weighted contrast-enhanced (DSC) MRI is a perfusion imaging technique and a potentially important tool for the characterisation of cerebral tumours. The percentage of signal intensity recovery (PSR) and relative cerebral blood volume (rCBV) derived from DSC MRI provide information about tumour capillary permeability and neoangiogenesis, which can be used to characterise tumour type and grade, and distinguish tumour recurrence from treatment-related effects. Therefore, PSR and rCBV potentially represent a non-invasive means of diagnosis; however, the clinical utility of these parameters has yet to be established. We present a review of the literature to date.

Introduction

Magnetic resonance imaging (MRI) is the technique of choice for evaluation of cerebral tumours.¹ Conventional T1- and T2-weighted MRIs are highly sensitive for detection of cerebral tumours; and can indicate the presence of oedema, mass effect, haemorrhage and tissue necrosis²; however, conventional MRI is sometimes unable to differentiate between certain contrast-enhancing lesions, making diagnosis challenging.

High-grade cerebral tumours, such as glioblastoma multiforme (GBM), typically disrupt the blood-brain barrier (BBB), leading to contrast agent leakage into the tumour region; and thus, usually contrast-enhancement on conventional MRI.² However, as some low-grade tumours may also enhance, contrast-enhancement does not correlate with grade.³ Knowledge of tumour grade is important for determining appropriate management. Additionally, there are some clinical scenarios in which conventional MRI does not allow precise diagnosis, such as diagnosing ring-enhancing lesions; and differentiating between tumour recurrence and treatment-related changes, such as radiation necrosis and pseudoprogression.^4–6 Furthermore, other contrast-enhancing tumours, such as primary central nervous system lymphoma (PCNSL) and solitary metastasis can sometimes appear similar to primary gliomas, making precise diagnosis on conventional MRI impossible.⁷

Definitive diagnosis of cerebral tumours currently relies on histopathological analysis of surgically-obtained specimens, requiring an invasive procedure with surgical and anaesthetic risk^8,9; however, histopathology is not always definitive, with non-diagnostic results of up to 10%.⁸ Inconclusive results may occur due to heterogeneous histology, diffuse tumour infiltration and difficult access due to anatomical site, resulting in inadequate or unrepresentative samples.^8,9 A noninvasive imaging-based means of accurate diagnosis would, therefore, be advantageous. Currently, this is not available.

Advanced MRI techniques have demonstrated increasing utility for aiding diagnosis of cerebral lesions in various clinical scenarios.^10,11 In particular, perfusion MRI techniques such as dynamic susceptibility-weighted contrast-enhanced (DSC) imaging are considered useful for viewing various pathologies that disrupt tissue perfusion.¹ DSC MRI has been applied to diagnosis and grading of cerebral tumours, distinguishing tumour recurrence from radiation necrosis, and evaluating the response to therapy through deriving two clinically useful parameters: Percentage of signal intensity recovery (PSR) and relative cerebral blood volume (rCBV)^10,12; however, despite its potential clinical utility, DSC MRI remains largely a research tool.

The purpose of this review is to summarise the clinical applications of PSR and rCBV for preoperative assessment of cerebral tumours; and to discuss some of the factors which have hitherto prevented PSR and rCBV from being more widely utilised in clinical practice.

DSC MRI

DSC MRI is recognised as a potentially important tool for the evaluation of cerebral tumours, based on tumour neoangiogenesis and capillary permeability.⁷ It tracks the first pass of a bolus of gadolinium-based contrast through brain tissue.¹ Passage of contrast agent through the tissue vasculature typically causes a transient decrease in brain signal intensity (SI). This signal loss is used to assess the concentration of contrast agent as a function of time, in each voxel of the image. After the first pass of the contrast agent, the SI gradually returns to baseline, which enables the PSR and rCBV to be obtained.¹²

PSR is a semi-quantitative measure of capillary permeability and is expressed as the percentage of SI (%SI) recovered at the end of the first pass of contrast through brain vasculature, relative to baseline. It is calculated from T2* SI-time curves, generated from the DSC imaging set.¹ PSR reflects BBB integrity and contrast agent leakage from tumour capillaries; thus, it provides information about alterations in capillary permeability. The rCBV is a unitless, semi-quantitative measure of microvascular density, and it provides an estimation of the degree of tumour neoangiogenesis, relative to normal brain tissue.¹³ It is calculated from CBV maps that are also derived from the DSC imaging set. Regions of interest (ROIs) are drawn around the lesion using image-processing software, generating SI-time curves and CBV maps.¹³

DSC MRI is obtained using T2*-weighted echo planar imaging (EPI), with either gradient echo EPI (GRE-EPI) or spin-echo EPI (SE-EPI) sequences.¹⁴ SE-EPI is most sensitive for small vessels, while GRE-EPI is sensitive for a broad range of vessels and is more useful for assessing tumour vessels, which are characteristically enlarged and morphologically abnormal.¹² Both 1.5T and 3T MRI scanners can be used; however, contrast agent dose and pulse sequence parameters may vary, depending on the field strength.¹² Gadolinium-based contrast agents must be injected at a relatively high rate (typically around 5 mL/s) in order to achieve a sufficiently concentrated contrast bolus, for analysis of the first pass.¹⁴ A dose of 0.1 mmol/kg is recommended by some authors, and is acceptable at both 1.5T and 3T.¹² Studies quantifying PSR alone have typically used lower contrast doses; however, as rCBV requires a higher contrast dose than PSR, this has been recommended if PSR and rCBV are to be calculated together.¹⁵ At our institution, GRE-EPI sequences are acquired during bolus injection of 0.1 mmol/kg gadolinium contrast, at a rate of 4 mL/s.

Postprocessing and calculation of the parameters derived from DSC MRI is based on simple deconvolution and pharmacokinetic modelling, using commercially available software.^12,14 Various techniques for analysing PSR and rCBV have been proposed.¹⁶ Absolute quantification of perfusion parameters requires measurement of the arterial input function (AIF), which represents the concentration of contrast agent entering the tissue at a given time point; however, accurate measurement of AIF is challenging and no standardised approach exists.¹² Due to these difficulties, semiquantitative measures like PSR and rCBV are generally used. CBV values from the area of interest are normalised (and hence ‘relative’) to an area of normal tissue, such as the contralateral normal-appearing white matter. Similarly, PSR is calculated relative to the baseline SI, and is thus semiquantitative.¹²

Contrast agent extravasation is a well-known challenge to DSC-based imaging, as the pharmacokinetic models used in post-processing assume that contrast agent remains strictly intravascular.¹⁷ When contrast agent leaks into the extravascular-extracellular (interstitial) space, the expected loss in SI on the T2-weighted image may be masked by an increase in signal intensity in the interstitial space; where there is significant T1 effect, known as the ‘susceptibility effect’.¹⁸ This can result in either overestimation or underestimation of rCBV, and overestimation of PSR in tumours; which cause breakdown of the BBB.¹⁵ As a result, various methods to minimise the effects of contrast leakage have been proposed. These include: administering a loading dose of contrast agent before injection of the main bolus, using small flip angle gradient echo or double echo acquisitions, and post-processing mathematical correction.^17–19 There are limitations to each method, and there is currently no universally accepted method of T1 leakage correction. At our institution, we use a loading dose of 0.05 mmol/kg of contrast agent, as this has been found to be the most robust method for diminishing the T1 leakage effect.^7,16,18

Clinical applications of PSR and rCBV

Diagnosing solitary contrast-enhancing lesions

Solitary contrast-enhancing cerebral lesions are often difficult to diagnose on conventional MRI. Tumours such as high-grade glioma (HGG), PCNSL and solitary metastasis may have similar appearances. GBM and metastasis can both appear as ring-enhancing lesions on T2-weighted MRI, with a perilesional area that is non-enhancing on T1-weighted post-contrast images.²⁰ PCNSL is typically hypointense on T1 and T2 MRIs with homogeneous contrast-enhancement; however, it may sometimes demonstrate T2 hyperintensity.²¹ Furthermore, PCNSL may spread across the corpus callosum, mimicking GBM; and while ring-enhancement is rare, it may occur.²¹ A non-invasive method of accurately diagnosing these lesions would be of significant value.

Several studies have found rCBV and PSR to be potentially useful in distinguishing between PCNSL, GBM and solitary metastasis.^7,22–25 Characteristic features include high PSR and low rCBV in PCNSL, intermediate PSR with high rCBV in GBM, and low PSR in metastasis (Table 1); rCBV has not been extensively evaluated in metastasis. One retrospective study of 20 cases found significantly lower rCBV and higher PSR figures for PCNSL, compared to HGG; however, there was considerable overlap between the two groups and sensitivity and specificity values were not provided.²² Other studies show similar trends, though reported rCBV and PSR values vary considerably. For example, Xing et al.²⁴ reported a PSR of 89% as being 100% sensitive and 88.5% specific for distinguishing PCNSL from GBM; whilst Mangla et al.⁷ reported a PSR of 114% as 82% sensitive, but did not provide specificity. Similarly, while the trends are consistent, Mangla et al.⁷ reported quite different rCBV values to Toh et al.²⁵; though without providing sensitivity nor specificity values. Therefore, the results are difficult to compare, possibly reflecting differences in image acquisition and post-processing methods.

Table 1.

Diagnosing contrast-enhancing lesions: Summary of literature results.

Study (n)	Group (n)	Mean rCBV	Mean PSR (%)	rCBV threshold (Sn%; Sp%)	PSR threshold (%) (Sn%; Sp%)	Comments	Level of evidenceb
Hartmann 2003 (24)	GBM (12)	4.99a	NA	Not provided	NA	Retrospective. No cutoffs, sensitivities or specificities provided.	4
	PCNSL (12)	1.29a	NA
Cha 2007 (43)	GBM (27)	NA	80.9a	NA	66 (67; 100)	Retrospective. Some cases not histopathologically confirmed.	4
	MET (16)	NA	62.5a
Liao 2009 (20)	HGG (11)	4.86a	93a	Not provided	Not provided	Retrospective. No cutoffs, sensitivities or specificities provided.	4
	PCNSL (9)	1.72a	175a
Mangla 2011 (66)	GBM (22)	4.72a	87a	Not provided	94 (PCNSL vs. other) (100; 69) 75 (MET vs. other) (100; 83)	Retrospective. Some cases not histopathologically confirmed.	4
	PCNSL (22)	2.43a	158a
	MET (22)	2.93a	55a
Xing 2013 (38)	HGG (26)	5.05a	68a	2.56 (96.2; 90)	89 (100; 88.5)	Retrospective	4
	PCNSL (12)	1.69a	137a
Toh 2013 (35)	GBM (20)	5.47a	NA	3.01 (90; 93.3)	NA	Retrospective	4
	PCNSL (15)	2.28a	NA

^ap < 0.05.

^bSee Appendix A.

GBM: glioblastoma; HGG: high-grade glioma; MET: metastasis; NA: not studied; PCNSL: primary central nervous system lymphoma; PSR: percentage of signal intensity recovery; rCBV: relative cerebral blood volume; Sn: sensitivity; Sp: specificity; vs.: versus

The significant differences in rCBV and PSR found between PCNSL, GBM and metastasis are attributed to variations in capillary architecture and permeability, and microvascular density. GBM capillaries usually exhibit BBB disruption and morphological variety, including glomeruloid capillaries, simple vascular hyperplasia and neocapillaries resembling normal brain vasculature.²⁶ This heterogeneity in capillary architecture results in a normal-to-increased range of permeability.²⁶ In contrast, lymphomas have a characteristically angiocentric growth pattern, without prominent neovascularization. Thus, GBM demonstrates high rCBV and low PSR, because there is higher transient SI loss after the first pass and subsequent return to baseline, while PCNSL yields low rCBV, but comparatively high PSR.²³ Metastatic lesions tend to exhibit prominent capillary fenestration and completely lack BBB components. This results in uniformly increased permeability and generally low PSR.²⁶

Glioma grading

Gliomas are the most common primary malignant brain tumour in adults and are graded on a spectrum from Grade I to Grade IV.^27,28 The most common primary malignant brain tumours in adults are graded on a spectrum from Grade I to Grade IV, including GBM, which usually requires resection and adjuvant chemotherapy or radiotherapy. Determination of grade currently relies on a histopathological analysis; however, this has several limitations, as discussed earlier.^27,29 The degree of malignancy correlates with neovascularisation, among other features; and thus, while conventional MRI has low sensitivity for grading, DSC MRI may provide useful information.³⁰

Several studies show rCBV to be significantly higher in HGG, compared to LGG^{18,27,29–31}; however, reported rCBV values, specificities and sensitivities differ widely (Table 2). For example, Law et al.²⁹ report that a rCBV of 1.75 significantly differentiates between LGG and HGG with 95% sensitivity and 57.5% specificity; while Hakyemez et al.³⁰ report a rCBV cutoff value of 2.00 to be 100% sensitive and 90.9% specific. These results are also difficult to compare, as Law et al.²⁹ report max rCBV values, while Hakyemez et al.³⁰ quote mean rCBV. Another study found that rCBV significantly differentiated between Grade II and Grade IV gliomas, but not Grade III and Grade IV tumours, though sensitivity and specificity values were not quoted for the proposed threshold.¹⁸ Furthermore, the inclusion of anaplastic oligodendroglioma (Grade III) in the HGG group in these studies is controversial. Though high-grade anaplastic oligodendrogliomas, specifically those possessing the 1p19q co-deletion, are more chemosensitive and have a better prognosis than other Grade III and Grade IV gliomas.^32,33 The clinical utility of categorising all Grade III gliomas in the HGG group in these studies is controversial.

Table 2.

Glioma grading: Summary of literature results.

Study (n)	Group (n)	Mean rCBV	Mean PSR (%)	rCBV threshold (Sn%; Sp%)	PSR threshold (%) (Sn%; Sp%)	Comments	Level of evidenceb
Law 2003 (160)	HGG (120)	5.18a (max)	NA	2.97 (72.5; 87.5)	NA	Retrospective study, though comparatively large. Mean rCBV not reported; max rCBV values referred to.	4
	LGG (40)	2.14a (max)	NA
Lev 2004 (30)	HGG (17)	2.9a		1.5 (97; 55)	NA	LGG rCBV lower (1.3) when oligodendroglioma excluded.	4
	LGG (13)	1.5a
Hakyemez 2005 (33)	HGG (22)	6.5a	NA	2.00 (100; 90.9)	NA	Prospective study. Inclusion of anaplastic oligodendroglioma in HGG group and oligodendroglioma in LGG group is controversial.	4
	LGG (11)	1.69a	NA
Boxerman 2006 (43)	Grade II (11)	1.52a	NA	Not provided	NA	No significant difference between Grade III and Grade IV tumours found.	4
	Grade III (9)	2.84a	NA
	Grade IV (23)	3.96a	NA
Server 2011 (79)	HGG (61)	6.6a	NA	2.94 (98.4; 72.2)	NA	Prospective. Some patients on steroids. Inclusion of oligodendroglioma in LGG controversial.	4
	LGG (18)	2.42a	NA
Aprile 2015 (49)	HGG (31)	5.0	65.8a	2.6 (77.7; 74.1)	89 (94.4; 96.8)	Retrospective. Difference in rCBV between LGG and HGG not statistically significant (p > 0.05).	4
	LGG (18)	2.3	94.9a
Smitha 2015 (64)	HGG (25)	5.29a	81.4a	Not provided		Prospective. All patients treatment naïve.	4
	LGG (39)	2.60a	94.8a

^aP < 0.05.

^bSee Appendix A.

HGG: high-grade glioma; LGG: low-grade glioma; NA: not studied; PSR: percentage of signal intensity recovery; rCBV: relative cerebral blood volume; Sn: sensitivity; Sp: specificity; vs.: versus

More recently, one study reported that rCBV values in the peri-enhancing region of the tumour were significantly higher in HGG, compared to LGG²⁷; however, a large proportion of patients were on steroid treatment at the time of rCBV measurement, which is a recognised confounder and decreases permeability and tumour blood volume. Furthermore, the authors defined LGG as World Health Organisation (WHO) Grade II tumours, including oligodendrogliomas, which are highly vascular tumours and may impact rCBV values.^31,34 Indeed, Cha et al.³⁴ found that tumour rCBV was significantly higher in low-grade oligodendroglioma than in low-grade astrocytoma, the most common subtypes of LGG. Thus, the categorization of ‘low-grade’ versus ‘high-grade’ is not necessarily useful, possibly over-representing the relationship between rCBV and glioma grade.

To date, only two studies have investigated the utility of PSR for glioma grading, with both PSR to be significantly higher in LGG than in HGG.^35,36 Contrary to previous studies, Aprile et al.³⁵ did not find a significant difference in rCBV between the groups (p > 0.05), despite having a larger sample size than some previous studies. Interestingly, Smitha et al.³⁶ analysed PSR using two separate software packages, and they found that that only one of them was able to distinguish Grade III and Grade IV gliomas. This raises interesting questions regarding the degree to which the commercial software package used may influence results. While preliminary results are encouraging, further evaluation is required to draw an informed conclusion.

Tumour recurrence versus radiation necrosis

Radiation therapy is used to treat various cerebral tumours, including HGG and metastatic brain tumours, and can prolong survival for many patients.³⁷ High doses of radiation are delivered to a specified region, in order to kill tumour cells. Patients are subsequently monitored for either tumour recurrence or delayed radiation necrosis, a complication of therapy.⁴ On conventional MRI, both may appear as a mass lesion exhibiting progressive contrast-enhancement and edema. Histopathological analysis of tissue samples is currently the only method of definitive diagnosis.^4,37 Histopathologically, radiation necrosis typically consists of fibrinoid necrosis, vascular dilation and endothelial injury to the surrounding cerebral vasculature, while recurrent tumour is generally characterised by vascular proliferation and increased vascular density.⁴ DSC MRI may be useful in differentiating these entities, based on their respective vascular characteristics.

While rCBV was more widely studied, with respect to distinguishing tumour recurrence from radiation necrosis, both rCBV and PSR are potentially useful. Several studies show significantly higher rCBV and lower PSR values for recurrent HGG, compared to delayed radiation necrosis (Table 3).^5,37–39 Hu et al.⁴⁰ report a rCBV cutoff value of 0.71 to distinguish the two groups, with 100% specificity and 91.7% specificity; however, the mean rCBV and P values for each group were not provided, and there was significant variation in the radiotherapy dose.⁴⁰ Furthermore, histopathological confirmation is lacking in three of these studies, with some diagnoses based on clinical and radiological follow-up.^5,38,39 Given that both entities may have identical clinical and radiological characteristics, this is a notable limitation.⁴¹

Table 3.

Radiation necrosis versus recurrent HGG: Summary of literature results.

Study (n)	Group (n)	Mean rCBV		Mean PSR (%)	rCBV threshold (Sn%; Sp%)	PSR threshold (%) (Sn%; Sp%)	Comments	Level of evidenceb
Di Constanzo 2008 (29)	GBM (21)	1.73a		NA	Not provided	NA	Retrospective. Significant variation in treatment history and regimes. Some patients on steroids. Not all cases histopathologically confirmed.	4
	RN (8)	0.86a		NA
Hu 2009 (40)	HGG (24)	0.55 - 4.64		NA	0.71 (91.7; 100)	NA	Retrospective. Mean rCBV and P values not provided. Significant variation in radiotherapy dose.	4
	RN (16)	0.21 -0.71		NA
Barajas, Chang, Segal 2009 (57)	GBM (40)	2.38a		80.2a	1.75 (78.92; 71.58)	87.3 (78.26; 76.19)	Retrospective. Not all cases were histopathologically confirmed.	4
	RN (17)	0.82a		89.3a
Fink 2012 (40)	HGG (30)	3.62a	NA		2.08 (86.2; 90)	NA	Retrospective. Not all cases were histopathologically confirmed.	4
	RN (10)	1.31a	NA

^aP < 0.05.

^bSee Appendix A.

GBM: glioblastoma; HGG: high-grade glioma; MET: metastasis; NA: not studied; PSR: percentage of signal intensity recovery; rCBV: relative cerebral blood volume; RN: radiation necrosis; Sn: sensitivity; Sp: specificity; vs.: versus

Studies comparing recurrent metastasis to radiation necrosis have also found significantly higher rCBV and lower PSR values in the tumour recurrence groups; however, it is worth noting that proposed threshold values for PSR and rCBV differ widely between studies and that no prospective studies have been published to date (Table 4).^4,42,43 Mitsuya et al.⁴² found an rCBV cutoff of 2.1 to provide 100% sensitivity and 95.2% specificity. Interestingly, Barajas et al.⁴ reported PSR to be more significant than rCBV in differentiating between recurrent metastasis and radiation necrosis following radiotherapy, with a cutoff of 76.3% yielding a sensitivity of 95.65% and a specificity of 100%; while Huang et al.⁴³ found rCBV, but not PSR, to differ significantly. This difference may be explained by the fact that Huang et al.⁴³ used different pulse sequence parameters than Barajas et al.⁴ It is known that MRI pulse sequence parameters, such as echo time (TE) and flip angle, can impact measured PSR values.⁴⁴ Furthermore, all studies show significant overlap in rCBV and PSR values between groups. Thus, larger prospective studies are required for adequate evaluation.

Table 4.

Radiation necrosis vs. recurrent metastasis: Summary of literature results.

Study (n)	Group (n)	Mean rCBV		Mean PSR (%)	rCBV threshold (Sn%; Sp%)	PSR threshold (%) (Sn%; Sp%)	Comments	Level of evidenceb
Hatzoglou 2013 (12)	GBM/MET (5)	Not provided			1.8 (100; 71)	74 (60; 100)	Retrospective. GBM and MET mixed. Very small study size. Mean rCBV, PSR and P values for each group not provided.	4
	RN (7)	Not provided
Barajas, Chang, Sneed 2009 (34)	MET (23)	2.38a		60.64a	1.54 (91.3; 72.73)	76.3 (95.65; 100)	Retrospective. Not all cases histopathologically confirmed.	4
	RN (11)	1.54a		83.33a
Mitsuya 2010 (28)	MET (7)	1.0a	NA		2.1 (100; 95.2)	NA	Retrospective. Lesions not histopathologically confirmed.	4
	RN (21)	3.5a	NA
Huang 2011 (27)	MET (18)	2.49a	81		2.0 (56; 100)	Not provided	Retrospective. PSR not statistically significant between groups.	4
	RN (9)	1.03a	80

^ap < 0.05.

^bSee Appendix A.

GBM: glioblastoma; HGG: high-grade glioma; MET: metastasis; NA: not studied; PSR: percentage of signal intensity recovery; rCBV: relative cerebral blood volume; RN: radiation necrosis; Sn: sensitivity; Sp: specificity; vs.: versus

Evaluation of treatment response and prognosis

The response of cerebral tumours to radiotherapy and chemotherapy is typically monitored by serial conventional MRI.⁴⁵ Response is evaluated by successive measurements of the contrast-enhancing area of the tumour, which assumes this area represents the tumour size; however, any process that affects BBB permeability can affect contrast-enhancement, independent of tumour activity.⁴⁶ New therapies such as bevacizumab, used for salvage chemotherapy in GBM, can decrease BBB permeability, resulting in a phenomenon called ‘pseudoresponse’, describing a decrease in the contrast-enhancing area of the tumour without actual decrease in tumour size.⁴⁶ This can cause overestimation of treatment response. Furthermore, in up to 40% of GBM patients who receive chemoradiotherapy, tissue injury can increase BBB permeability, causing progressive contrast-enhancement without tumour progression. This phenomenon is called ‘pseudoprogression’.⁶ PSR and rCBV may assist in monitoring true glioma behavior following therapy, and may also be of prognostic importance.

Three recent studies found HGG to demonstrate similarly high rCBV, compared to pseudo-progression; however, this was statistically significant in only two of them (Table 5).^6,19,47–50 Only one study included PSR, which was found to be significantly lower in true tumour progression, compared to pseudo-progression.⁵⁰ PSR and rCBV in the pseudo-response has not yet been directly studied.^46,51 Pseudo-response correlates with shorter overall survival (OS), compared to true tumour response; and thus, the utility of rCBV as a prognostic indicator could be a valuable topic for future research.^45,51 High rCBV levels have also been found to have a significant inverse correlation with OS, in anaplastic astrocytoma (Grade III) and GBM (Grade IV), suggesting that rCBV may be useful for prognostication in HGG^19,48,49; however, histopathological confirmation was lacking in all studies. Furthermore, there was significant variation in treatment administration and dosage between patients, and the degree to which this may have affected the results is not clear.

Table 5.

Pseudoprogression versus true progression: Summary of literature results.

Study (n)	Group (n)	Mean rCBV	Mean PSR (%)	rCBV threshold (Sn%; Sp%)	PSR threshold (%) (Sn%; Sp%)	Comments	Level of evidenceb
Kong 2011 (90)	TP (33)	2.85a	NA	1.49 (81.5; 77.8)	NA	Prospective. Not all cases histopathologically confirmed.	4
	PsP (26)	1.49a	NA
Gahramanov 2013 (19)	TP (10)	> 1.5	NA	1.5 (not provided)	NA	Prospective. Assumption re overall survival rather than histopathological confirmation used to categorise groups. Mean rCBV and P values for each group not provided.	4
	PsP (9)	< 1.5	NA
Young 2013 (20)	TP (16)	2.75a	84a	1.8 (100; 75)	90 (100; 63)	Prospective. Cases not histopathologically confirmed.	4
	PsP (4)	1.5a	101a

^ap < 0.05.

^bSee Appendix A.

GBM: glioblastoma; HGG: high-grade glioma; MET: metastasis; NA: not studied; PSP: pseudo-progression; PSR: percentage of signal intensity recovery; rCBV: relative cerebral blood volume; Sn: sensitivity; Sp: specificity; TP: true tumour progression; vs.: versus

Two recent studies suggest that rCBV may also be useful for predicting the natural history of low-grade gliomas.^52,53 Law et al.⁵² found that lower rCBV values were associated with longer overall survival, with rCBV values > 1.75 being significantly associated with rapid tumour progression and possible malignant transformation. A small study of 13 patients with low-grade glioma found rCBV to be a significant predictor of malignant transformation, at least 12 months before transformation; however, histopathological confirmation was lacking for some cases.⁵³ Another retrospective study found rCBV is significantly correlated with patient survival for grades II, III and IV of glioma, though the patients with HGG were receiving steroid treatment at the time of the study, potentially impacting the results.⁵⁴ Thus, while the literature is encouraging, the few studies published to date have several limitations and more evaluation is necessary.

Challenges and implications for future practice

While PSR and rCBV possess considerable clinical potential for the non-invasive assessment of cerebral tumours, routine use for guiding patient management has yet to be established.⁵⁵ As discussed, the majority of data to date is derived from small, single-centre, retrospective studies. Although some prospective studies have been published, these are mainly single-centre, with small patient populations. Furthermore, reproducibility and technical standardisation for obtaining PSR and rCBV measurements have yet to be established. Thus, while their potential clinical applications are numerous, rCBV and PSR remain largely research tools without clinical validation.¹²

Quality of studies

Almost all studies published on the subject are observational studies of the case series type, as defined by the Oxford Centre for Evidence-based Medicine (CEBM).⁵⁶ Case series are considered Level 4 evidence, according to the CEBM Levels of Evidence (Appendix A). These studies are considered to yield low-quality evidence, as they lack controls and are susceptible to confounders and many types of bias.⁵⁶ Given that < 30 relevant studies on the subject have been published, these are also likely to suffer from publication bias⁵⁷; however, it is worth noting that despite their numerous limitations, most of the studies have reported largely consistent results. This is encouraging for future, more rigorous studies.

Technical standardisation

Despite a significant body of literature, there is still no standardised technique for measurement of rCBV and PSR.¹² It is known that pulse-sequence parameters, such as TE, flip angle and field strength, as well as contrast dose, can impact measured PSR values through influencing the T1- and T2*-weighting.¹⁸ While an in-depth technical comparison of the literature is beyond the scope of this review, a wide range of pulse-sequence parameters has been published, some of which were discussed earlier in this review. Furthermore, differences in the commercial software used for post-processing is considered the largest source of variability in perfusion parameters, among and between studies.¹³ The degree to which technical variations have impacted the results is unknown; however, until standardised guidelines for image acquisition and post-processing methods exist, widespread adoption in a clinical setting is unlikely.¹

Reproducibility

Few studies to date have examined the reproducibility of common techniques used for rCBV and PSR measurement.^12,58 Measurement generally relies on accurate placement of ROIs around the lesion in question. These ROIs are operator-defined, and thus incorporate a degree of subjectivity. There is no single best method of determining ROIs.¹² Although using multiple ROIs has been deemed clinically acceptable and adopted in most of the studies reviewed, considerable inter- and intra-observer variability exists.^12,58 Tumour heterogeneity may distort calculations, and accurate ROI placement can be difficult in lesions with complex borders.^12,18 Furthermore, to date all studies were conducted by experienced radiologists working in a research environment, potentially limiting reproducibility in clinical settings. Methods enabling even inexperienced operators to obtain reliable results could increase their reproducibility and clinical applicability.⁵⁰ Alternatives to ROI-based analysis, such as histogram analysis and parametric response mapping, have been proposed; however, these are highly technical and still under development.¹²

Conclusions

This review has discussed some of the clinical applications of PSR and rCBV to the preoperative assessment of cerebral tumours. These parameters, derived from DSC MRI, have considerable potential for widespread clinical value; however, the majority of studies to date have been small and retrospective in nature. This and the lack of technical standardisation have hitherto prevented PSR and rCBV from being more widely utilised in clinical practice. Future large, prospective and multi-centre studies are necessary, before PSR and rCBV can be justified for routine patient assessment and management.

Appendix A. Oxford Centre for Evidence-based Medicine: Levels of Evidence

The Oxford Centre for Evidence-based Medicine (CEBM) Levels of Evidence describes a systematic approach for stratifying evidence, based on quality and strength, and is widely used for evaluating the quality of research. These are summarized in the table below, which is adapted from the CEBM.⁵⁶

Level	Type of evidence
1a	Systematic review of randomised controlled trials
1b	Individual randomised controlled trials
1c	All or none case-series
2a	Systematic review of cohort studies
2b	Individual cohort study
2c	Outcomes research
3a	Systematic review of case-control studies
3b	Individual case-control study
4	Case series
5	Expert opinion without critical appraisal
Adapted from Phillips et al., 1998.

Funding

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Conflict of interest

The authors declare that there are no conflicts of interest.