DSC-MRI is a “bolus tracking” technique that rapidly acquires gradient-echo (GRE) or spin-echo (SE) echo-planar images before (baseline), during, and after (tail) first-pass transit through the brain of an exogenous, paramagnetic gadolinium-based contrast agent (GBCA) that transiently decreases signal intensity. (7) Voxel-wise changes in contrast agent concentration are determined by applying susceptibility contrast physics to the signal intensity-time curves, and processed using tracer kinetic modeling and indicator dilution theory to estimate relative cerebral blood volume (CBV), which is the primary imaging marker obtained with DSC-MRI. (8-12) Because high-grade gliomas, for instance, produce pro-angiogenic factors, (13) relative CBV has putative value for differentiating tumor characterized by enlarged microvessels with high vascular density from treatment effects characterized by inflammatory or steroid-like behavior as in PsP or PsR.The indicator dilution technique assumes intravascular compartmentalization of non-diffusible tracer, which is violated for GBCA tracers in high-grade gliomas (HGG) with blood-brain barrier (BBB) disruption and avid contrast enhancement. GBCA extravasation yields T1 shortening, opposing the susceptibility contrast-induced T2 or T2* relaxivity change from intravascular GBCA that forms the basis for CBV estimation. Low to intermediate flip angles (i.e. 35°–60°) with longer TR (i.e. 1.2–1.7s) and TE (i.e. >20ms) can reduce T1 contamination due to GBCA extravasation. However, some parameter combinations may also reduce the signal-to-noise ratio (SNR) of the computed rCBV maps. (14). A “preload” dose of GBCA administered prior to the bolus dose of GBCA during dynamic imaging can help mitigate T1 contamination by partially saturating baseline T1-weighted signal contribution, (15) thereby diminishing T1-induced increased signal during bolus passage. Preload contrast administration combined with model-based post-processing leakage correction that diminishes both T1 and T2* extravasation-induced contamination effects improves the accuracy of rCBV estimates in HGGs. (16,17)Protocol decisions for DSC-MRI, including preload contrast administration, (15) acquisition parameters (flip angle, TE), and post-processing leakage correction, (16) are often chosen to minimize the effects of GBCA extravasation. Intravascular contrast agents like ferumoxytol eliminate contrast agent leakage effects, providing consistent CBV values across injections, whereas GBCA-based CBV is pre-load-dose dependent, which could impact longitudinal studies in clinical trials. (18,19) Several NCI-funded trials (e.g., NCT03264300, NCT03347617, and NCT00660543) use ferumoxytol-based CBV. However, application is limited because the standardized brain tumor imaging protocol (BTIP), (20) which is gaining traction for use in multi-center clinical trials, requires GBCA, and therefore this review focuses on GBCA-based DSC-MRI.Well-described size dependence relationships for GRE and SE relaxivity demonstrate that GRE relaxivity plateaus for large magnetic field perturber sizes, and SE relaxivity peaks for capillary-sized vessels. (11) Because of these relationships, GRE DSC-MRI has greater sensitivity to larger, disorganized microvessels seen in higher-grade tumors; greater signal changes for a given contrast agent dose; greater inherent accuracy of CBV estimates; and decreased sensitivity to changes in proton diffusion, (21) and therefore GRE DSC-MRI is recommended for use in neuro-oncology applications.CBV has been used to distinguish PsP and progressive disease (PD) at initial progressive contrast enhancement after chemoradiation, (22) and the literature is conflicting. For instance, Mangla et al found increased mean lesion CBV in PD and decreased CBV in PsP, with 77% sensitivity and 86% specificity. (23) Prager et al studied high-grade gliomas (HGG) at progressive contrast enhancement and found significant difference in median CBV between PsP and PD, with an optimal threshold of 1.3. (24) Kong et al found overall significant difference in mean CBV between PsP and PD for glioblastomas, but this difference applied to tumors with unmethylated but not with methylated MGMT. (25) However, a study of HGGs treated with paclitaxel poliglumex, a powerful radiation sensitizer, found no significant difference in mean CBV between PsP and PD at initial progressive enhancement. (26) These are just several examples of varied results in the literature.Literature results may conflict because DSC-MRI methodology varies greatly. Patel et al published a meta-analysis of studies using DSC-MRI for distinguishing recurrent tumor from treatment effect. (27) For the subgroup using mean lesion CBV, they found high-pooled sensitivity and specificity for recurrent tumor, but a wide range of optimal mean CBV thresholds for individual studies (0.9–2.15). This is likely due in large part to a wide range of DSC-MRI parameters including preload and bolus contrast agent dose, flip angle, TE and use of post-processing leakage correction. The authors concluded “standardization is needed before implementing any particular quantitative perfusion-weighted imaging strategy across institutions.”Efforts at standardization of these protocol choices have been made by several organizations including the ASFNR, which published a white paper recommending a 60-70o flip angle, field strength-dependent TE, and ¼–full dose preload with full-dose bolus. (28) Recently, the DSC-MRI Standardization Subcommittee of the Jumpstarting Brain Tumor Drug Development Coalition, representing the NBTS, SNO, and ABCC, used the ASFNR recommendations as a springboard for recommending DSC-MRI paradigms compliant with the standardized BTIP. (20) BTIP requires post-GBCA imaging to be done after one total dose of GBCA, either split between preload and DSC bolus before post-GBCA imaging, or fully given as preload with variable bolus dose DSC-MRI after post-GBCA imaging. This committee took a computational approach for determining optimal BTIP-compliant parameters.Ellingson and Leu used multi-compartment model-based simulation of DSC-MRI signal without versus with leakage to test a range of flip angles, TEs, and TRs with BTIP-compliant dosing schemes. They produced heat maps identifying the best performing strategies. Overall, the ASFNR parameters with 60° flip angle with full-dose preload/full-dose bolus had lowest mean error, but other schemes with high fidelity included low flip angle with no preload and low-intermediate flip angle with fractional dosing. Higher flip angles with no or fractional dosing performed relatively poorly. (29)Similarly, Quarles and Semmineh performed simulations using a validated population-based glioblastoma-trained digital reference object (30) and found the ASFNR parameters with 60o flip-angle to have excellent accuracy and precision for full-dose preload and bolus, but substantially degraded performance for fractional dosing, especially without preload at 1.5T. (31) A 30o low flip-angle scheme performed equally well for double-dosing (full-dose preload plus full-dose bolus), but also very well for fractional dosing, even without preload, and could be an attractive, generally applicable approach requiring less contrast agent.A recent study confirmed these simulation predictions with in vivo patient data, and demonstrated that a single-dose, low-flip angle DSC-MRI protocol without preload gives CBV estimates practically equivalent to the double-dose, intermediate FA reference standard that used a full-dose preload. A total of 84 patients diagnosed with a contrast-enhancing brain lesion were included in this three-institution study, which demonstrated practical equivalence between the methods supporting the idea that this low-dose approach should be considered for consensus protocol recommendation at least at 3T. Confirmation of equivalence at 1.5T requires a similar study.Spatial variation of lesion CBV is important. For example, glioblastomas with recurrent enhancement may have identical mean and median CBV, but very different fractional tumor burden (FTB), the percentage of voxels with CBV exceeding a threshold. FTB is thought to depict histopathology and correlate with OS better than mean or median CBV. (32,33)To test these concepts, a primary imaging trial entitled “Multi-site validation and application of a consensus DSC-MRI protocol” has been developed (NCT03401866). This NCI-funded, phase II imaging trial at four sites will directly compare the low flip-angle, no-preload and intermediate flip-angle, full-preload schemes in recurrent glioblastoma on standard therapy with stereotactic biopsy-guided histopathology validation, repeatability analysis, and survival prediction using FTB analysis. Repeatability is important because it impacts trial design. Analysis of double-baseline DSC-MRI in glioblastomas using six common post-processing methods revealed a wide range of repeatability coefficients for CBV. (34) Post-processing methodology impacted the estimated trial size required for detecting 10-20% change in CBV, with clear implications for DSC-MRI trial design.Pseudoresponse reflects decreased contrast enhancement independent of anti-tumor effect. (4) Bevacizumab yields high response rates and prolonged PFS without improved OS, (35) posing a limitation for early response assessment and prompting inclusion of non-enhancing tumor in modified response assessment criteria. (22) Regarding pseudoresponse and anti-angiogenic therapy, ACRIN 6677, the advanced imaging component of RTOG 0625, used DSC-MRI in recurrent glioblastoma treated with bevacizumab with or without irinotecan. Whereas OS was similar for decreasing and stable enhancement, (36) it differed significantly for patients with increasing versus decreasing CBV at week 2 or 16 of treatment compared to baseline. (37) Similar results were found in a single center study by Harris et al evaluating recurrent glioblastomas at baseline and within 2 months after starting bevacizumab. (38) Decreased CBV after therapy predicted improved OS, but pre- and post-treatment CBV independently did not. However, Schmainda et al evaluated recurrent HGGs 6-8 weeks after starting bevacizumab and found the opposite to be true: whereas pre- and early post-treatment CBV were independently predictive of OS, change in CBV was not. (39) And Kickingereder et al published results for bevacizumab-treated recurrent glioblastomas and found pre-treatment and 8-week follow-up CBV independently predicted OS, but change in CBV did not. (40) Therefore, as with PsP, there are controversies regarding CBV analysis of bevacizumab-treated glioblastoma. Should evaluation use single baseline or follow-up value, or change from baseline? What post-treatment time point should be used? And is mean or median sufficient, or should parametric response maps be used? Larger multi-center trials are needed to address these questions.EAF-151, an ECOG-ACRIN primary imaging trial for DSC-MRI entitled “Change in relative cerebral blood volume as a predictive biomarker for response to bevacizumab in patients with recurrent GBM,” (NCT03115333) aims to answer these questions. Using a standardized double-dose DSC-MRI protocol, EAF-151 aims to determine if binary changes (increase versus decrease) in CBV within enhancing tumor from baseline to two weeks post-bevacizumab initiation correlate with OS, with additional secondary aims that will help address some of these conflicts in the literature.There are additional DSC-MRI markers besides CBV that may help in clinical trials, including vessel size index, the ratio of measured gradient-echo to spin-echo relaxivity that monotonically increases with vessel size; (41) vessel architectural imaging, that parameterizes GRE versus SE relaxivity to produce vortex maps that change with treatment; (42) and measurement of transverse relaxivity at tracer equilibrium (TRATE), which has been shown to reflect tumor cytoarchitecture, including tumor cell density and cell size, potentially providing a unique structural signature unavailable with other techniques. (43) As advanced MRI perfusion markers evolve and prove useful for measuring different aspects of tumor physiology with predictive and prognostic value, it will become important to develop MRI pulse sequences and protocols that efficiently acquire the requisite data. To this end, Quarles et al have developed a multi-parametric perfusion MRI pulse sequence that combines DSC and dynamic contrast enhanced (DCE) acquisition using SAGE technique and a single bolus injection, permitting simultaneous assessment of CBV, VSI, permeability (ktrans), and measures of tumor cellularity and cytoarchitecture via TRATE. (44,45) Techniques such as this may prove useful for future clinical trials, but widespread dissemination of the pulse sequence and multi-site validation are required.So where do we stand with DSC-MRI for neuro-oncology trials? DSC-MRI CBV estimation is a mature, robust technique, and ample single-center data and some multi-center data exist, including application to glioma response assessment. However, DSC-MRI has thus far been acquired and post-processed with variable technique, as evidenced by often conflicting results in the literature. Recent efforts at standardization are very promising, including a single-dose, no-preload methodology using low flip-angle acquisition with excellent performance. Such a technique has much appeal given recent concerns over GBCA deposition in the brain. Multi-center trials are currently underway to validate these protocols and CBV accuracy, and to establish the utility of CBV for treatment response assessment. Other DSC-MRI markers for tumor biology including microvessel size, cell size, and perfusion efficiency are obtainable, permitting comprehensive assessment of tumor physiology pre- and post-therapy with use of advanced DSC-MRI pulse sequences. Advances in post-processing and analysis software will aid success.

Acknowledgements

No acknowledgement found.