Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 7-Day Trial for You or Your Team.

Learn More →

Evaluating Multisite rCBV Consistency from DSC-MRI Imaging Protocols and Postprocessing Software Across the NCI Quantitative Imaging Network Sites Using a Digital Reference Object (DRO)

Evaluating Multisite rCBV Consistency from DSC-MRI Imaging Protocols and Postprocessing Software... RESEARCH ARTICLE Evaluating Multisite rCBV Consistency from DSC-MRI Imaging Protocols and Postprocessing Software Across the NCI Quantitative Imaging Network Sites Using a Digital Reference Object (DRO) 1 1 2 2 2 Laura C. Bell , Natenael Semmineh , Hongyu An , Cihat Eldeniz , Richard Wahl , 3 3 4 4 5 Kathleen M. Schmainda , Melissa A. Prah , Bradley J. Erickson , Panagiotis Korfiatis , Chengyue Wu , 5 5 5 6 6 Anna G. Sorace , Thomas E. Yankeelov , Neal Rutledge , Thomas L. Chenevert , Dariya Malyarenko , 7 7 8 8 9,10 11 Yichu Liu , Andrew Brenner , Leland S. Hu , Yuxiang Zhou , Jerrold L. Boxerman , Yi-Fen Yen , 11 11 12 13 Jayashree Kalpathy-Cramer , Andrew L. Beers , Mark Muzi , Ananth J. Madhuranthakam , 13 13,14 1 Marco Pinho , Brian Johnson , and C. Chad Quarles 1 2 Division of Neuroimaging Research, Barrow Neurological Institute, Phoenix, AZ; Mallinckrodt Institute of Radiology, Washington University in St. Louis, St. Louis, MO; 4 5 Departments of Radiology and Biophysics, Medical College of Wisconsin, Wauwatosa, WI; Department of Radiology, Mayo Clinic, Rochester, MN; Department of 6 7 Diagnostic Medicine, University of Texas at Austin, Austin, TX; Department of Radiology, University of Michigan, Ann Arbor, MI; UT Health San Antonio, San Antonio, TX; 9 10 Department of Radiology, Mayo Clinic, Scottsdale, AZ; Department of Diagnostic Imaging, Rhode Island Hospital, Providence, RI; Alpert Medical School of Brown 11 12 University, Providence, RI; Department of Radiology, Massachusetts General Hospital, Boston, MA; Department of Radiology, University of Washington, Seattle, 13 14 Washington; UT Southwestern Medical Center, Dallas, TX; and Philips Healthcare, Gainesville, FL Corresponding Author: Key Words: DSC-MRI, relative cerebral blood volume, standardization, multisite consistency, Laura C. Bell, PhD reproducibility Division of Neuroimaging Research, Abbreviations: Relative cerebral blood volume (rCBV), postprocessing methods (PMs), imaging Barrow Neurological Institute, Phoenix, AZ, 85013; protocols (IPs), dynamic susceptibility contrast magnetic resonance imaging (DSC-MRI), Quantitative E-mail: [email protected] Imaging Network (QIN), American Society of Functional Neuroradiology (ASFNR), digital reference object (DRO), standard imaging protocol (SIP), intraclass correlation coefficient (ICC), limits of agreement (LOA), covariance (CV), echo time (TE), normal appearing white matter (NAWM) The use of rCBV as a response metric in clinical trials has been hampered, in part, due to variations in the biomarker consistency and associated interpretation across sites, stemming from differences in image acquisi- tion and post-processing methods. This study leveraged a dynamic susceptibility contrast magnetic resonance imaging digital reference object to characterize rCBV consistency across 12 sites participating in the Quanti- tative Imaging Network (QIN), specifically focusing on differences in site-specific imaging protocols (IPs; n 17), and PMs (n  19) and differences due to site-specific IPs and PMs (n  25). Thus, high agreement across sites occurs when 1 managing center processes rCBV despite slight variations in the IP. This result is most likely supported by current initiatives to standardize IPs. However, marked intersite disagreement was observed when site-specific software was applied for rCBV measurements. This study’s results have important implications for comparing rCBV values across sites and trials, where variability in PMs could confound the comparison of therapeutic effectiveness and/or any attempts to establish thresholds for categorical response to therapy. To overcome these challenges and ensure the successful use of rCBV as a clinical trial biomarker, we recommend the establishment of qualifying and validating site- and trial-specific criteria for scanners and acquisition methods (eg, using a validated phantom) and the software tools used for dynamic susceptibility contrast magnetic resonance imaging analysis (eg, using a digital reference object where the ground truth is known). INTRODUCTION aid in diagnosis (1), detecting treatment response (2, 3), guiding The relative cerebral blood volume (rCBV), derived from dy- biopsies (4, 5), and reliable differentiation of post-treatment namic susceptibility contrast magnetic resonance imaging radiation effects and tumor progression (6-10). It is also increas- (DSC-MRI), is an established biomarker of glioma status that can ingly leveraged as a biomarker of early therapeutic response in © 2019 The Authors. Published by Grapho Publications, LLC This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). ISSN 2379-1381 http://dx.doi.org/10.18383/j.tom.2018.00041 | | 110 TOMOGRAPHY.ORG VOLUME 5 NUMBER 1 MARCH 2019 ABSTRACT Multisite rCBV Consistency Using a DRO clinical trials (11, 12). However, variations in image acquisition each of these submitted site-specific IP DROs to evaluate and postprocessing methods (PMs) can limit rCBV reproducibil- differences owing to the IP provided. ity, potentially diminishing its clinical utility. To promote rCBV • Phase II (“constant IP w/site PM”) involved analysis of a reproducibility across institutions, many national initiatives are “standard imaging protocol” (SIP), which represents DSC- underway to standardize DSC-MRI acquisition and PMs, includ- MRI data acquired using the IP recommended by ASFNR 13). Each site was asked to process DSC-MRI DRO data ing National Cancer Institute’s Quantitative Imaging Network derived from the SIP. Some sites choose to use multiple (QIN), Radiological Society of North America’s Quantitative Im- commercially available software packages (site 03) and aging Biomarkers Alliance (QIBA), and the National Brain Tu- different rCBV definitions (sites 05, 06, 12), yielding a total mor Society’s Jumpstarting Brain Tumor Drug Development of 17 submitted rCBV maps. Coalition. Recent imaging protocol (IP) recommendations by the • Phase III (“site IP w/site PM”) required each site to calculate American Society of Functional Neuroradiology (ASFNR) has rCBV maps using their PM of choice and the site-specific served as the first step in standardizing DSC-MRI protocols for DRO data. Combining the possible permutations owing to clinical applications (13). choice of IP and PM from phases I–II, a total of 25 rCBV To aid in this effort, 12 institutions within the QIN aimed to maps were submitted. investigate and determine the current rCBV reproducibility us- ing a recently developed and validated in silico digital reference All sites but 1 completed all 3 phases of the challenge. Site 11 object (DRO) that is representative of a wide range of possible completed only phase I, and these results are included in this glioma magnetic resonance signals (14). Leveraging this DRO study. enables us as a community to determine the multisite consis- DRO Simulations tency in rCBV owing to varying permutations of imaging ac- The DSC-MRI signals for each IP were simulated using a recently quisition parameters and postprocessing steps. In specific, our developed and validated population-based DRO that was trained goals are to characterize rCBV consistency under conditions to generate realistic signals using in vivo data from 40 000 where there exist: (1) variations in the site-specific imaging voxels derived from patient data (14). The resulting DRO, which acquisition parameters (PMs held constant), (2) variations in contains 10 000 unique voxels, reflects the distribution of per- only site-specific PMs (IP held constant), and (3) variations fusion, permeability, precontrast T1, T2*, diffusion coefficients, owing to site-specific imaging and postprocessing protocols. and the vascular and cellular features found in patients with Results from this community-based challenge will help steer high-grade glioma. Using this DRO, the DSC-MRI signals and standardization of DSC-MRI rCBV protocols with the hope that resulting rCBV values can be computed for any combination of it can be successfully translated to the clinical setting. preload dosing scheme, contrast agent choice (by varying T1 relaxivities specific to the contrast agent), pulse sequence pa- MATERIALS AND METHODS rameters, and postprocessing protocol. For the purposes of this This National Cancer Institute QIN DSC-DRO challenge project study, the DRO consisted of tumor voxels simulated under two was proposed and organized by the investigators at Barrow blood-brain-barrier (BBB) conditions to recapitulate DSC-MRI trans Neurological Institute (BNI). Eleven centers participated in this signals from an intact-BBB (K 0) and a disrupted-BBB trans project: BNI (the managing center), Brown University (BU), Mas- (K 0). In addition to the tumor voxels, normal appearing trans sachusetts General Hospital (MGH), Mayo Clinic Arizona (Mayo white matter (NAWM) voxels (K 0) were simulated to AZ), Mayo Clinic Minnesota (Mayo MN), Medical College of normalize CBV. For the purposes of comparing site-to-site con- Wisconsin (MCW), University of Michigan (UM1), The Univer- sistency, the SIP that has been postprocessed by the managing sity of Texas Health at San Antonio (UTSA), University of Texas center was considered the reference standard where necessary. at Austin (UT), University of Texas Southwestern Medical Center In our recent study, focused on investigating the influence of IP at Dallas (UTSW), University of Washington (UW), and Wash- on CBV fidelity (15), the SIP yielded CBV values, when corrected ington University (WashU). Unless specifically named, these for contrast agent leakage, that were among the most accurate. participating sites have been anonymized, in no particular order, Site-Specific IP and PM Methods and will be referred to as sites 01–12 as seen in Table 1. Site-specific IP and PM methods are briefly listed in Table 1. This project comprised 3 phases, summarized in the last 3 Overall, IPs were similar across sites. Most sites submitted clin- columns of Table 1, to evaluate the influence of IPs and/or PMs ical DSC IPs for 3 T with 3 sites that also included a 1.5 T IP. on multisite consistency: Overall, the following were the imaging parameters [mode (min- • Phase I (“site IP w/constant PM”) involved each participat- max)]: repetition time  1500 milliseconds (1300 –2560 milli- ing site to submit their current clinical DSC IP to the seconds), echo time (TE)  30 milliseconds (18 –71 millisec- managing center. The managing center then simulated site- onds), flip angle  60° (60°–90°), preload dose  0.05 mmol/kg specific DROs reflecting the IP parameters provided. Some (0 – 0.1 mmol/kg), and injection dose  0.1 mmol/kg (0.05– 0.15 sites provided 1 IP owing to differences in field strengths mmol/kg). Five different gadolinium contrast agents were used (sites 01, 04, and 05), dosing schemes (sites 03 and 10), and across the 12 sites: gadobenate (n  5), gadobutrol (n  3), gad- acquisition method (site 04). In total, 19 different IPs were oterate (n  2), gadoteridol (n  1), and gadopentetate (n  1). For submitted. The managing center postprocessed (specific PMs, there was a mix of software options used, including in-house- based software scripts (n  4), IB Neuro (n  4), 3D Slicer (n  1), details below in “Site-specific IP and PM”) rCBV maps of | | TOMOGRAPHY.ORG VOLUME 5 NUMBER 1 MARCH 2019 111 Multisite rCBV Consistency Using a DRO Table 1. Summary of Participating Teams’ IPs and PMs Imaging Protocol (IP) Scan Protocol Dose Protocol ID Tag for Analysis Time Site IP Constant IP Site IP Site Field TR TE Preload Injection Between Processing Method w/Constant w/Site w/Site Number Strength (ms) (ms) Flip (mmol/kg) (mmol/kg) (min) CA (PM) PP PP PP 01 01:3.0 T 1500 30 60 0.05 0.10 3 Gadobenate 01: In-house processing S01_IP01 S01_PM01 S01_IP01_PM01 02:1.5 T 1500 30 60 0.05 0.10 3 Gadobenate S01_IP02 S01_IP02_PM01 02 01:3.0 T 1600 30 60 0 0.1 n/a Gadobenate 01: IB Neuro S02_IP01 S02_PM01 S02_IP01_PM01 03 01:3.0 T 1500 31 90 0.05 0.15 6.5 Gadoterate 01: 3DSlicer S03_IP01 S03_PM01 S03_IP01_PM01 02:3.0 T 1500 31 90 0.1 0.1 6.5 Gadoterate 02: nordicICE S03_IP02 S03_PM02 S03_IP01_PM02 03: PGUI S03_PM03 S03_IP01_PM03 S03_IP02_PM01 S03_IP02_PM02 S03_IP02_PM03 04 01:3.0 T 1500 30 80 0.10 0.10 5 Gadobutrol 01: IB Neuro S04_IP01 S04_PM01 S04_IP01_PM01 02:3.0 T 1500 2,35 80 0 0.10 n/a Gadobutrol S04_IP02 n/a 03:1.5 T 1500 30 72 0.10 0.10 5 Gadobutrol S04_IP03 S04_IP03_PM01 04:1.5 T 1500 2,35 72 0 0.10 n/a Gadobutrol S04_IP04 n/a 05 01:3.0 T 1300 30 60 0.025 0.10 5 Gadobutrol 01: IB Neuro (Integration S05_IP01 S05_PM01 S05_IP01_PM01 limits 1) 02:1.5 T 1300 30 60 0.025 0.10 5 Gadobutrol 02: IB Neuro (Integration S05_IP02 S05_PM02 S05_IP01_PM02 limits 2) S05_IP02_PM01 S05_IP02_PM02 06 01:3.0 T 1500 30 75 0.10 0.10 5 Gadoteridol 01: PGUI (rCBV definition 1) S06_IP01 S06_PM01 S06_IP01_PM01 02: PGUI (rCBV definition 2) S06_PM02 S06_IP01_PM01 07 01:3.0 T 1500 30 65 0.025 0.075 6 Gadobenate 01: In-house processing S07_IP01 S07_PM01 S07_IP01_PM01 08 01:3.0 T 1500 21 60 0.10 0.05 6 Gadobenate 01: In-house processing S08_IP01 S08_PM01 S08_IP01_PM01 09 01:3.0 T 1500 18 60 0.05 0.05 6 Gadobenate 01: IB Neuro S09_IP01 S09_PM01 S09_IP01_PM01 10 01:3.0 T 1900 36 90 0 0.10 n/a Gadoterate 01: In-house processing S10_IP01 S10_PM01 S10_IP01_PM01 02:3.0 T 1900 36 90 0.10 0.10 5 Gadoterate S10_IP02 S10_IP02_PM01 11 01:3.0 T 2560 71 90 0.025 0.10 2 Gadopentetate n/a S11_IP01 n/a n/a 12 01:3.0 T 1757 30 90 0.033 0.067 8 Gadobutrol 01: Philips ISP (rCBV S12_IP01 S12_PM01 S12_IP01_PM01 definition 1) 02: Philips ISP (rCBV S12_PM02 S12_IP01_PM02 definition 2) 03: Philips ISP (rCBV S12_PM03 S12_IP01_PM03 definition 3) Standard 01:3.0 T 1500 30 60 0.10 0.10 5 Gadopentetate n/a SIP n/a n/a Protocol Total :12 19 17 25 Excludes the standard protocol. nordicICE (n  1), PGUI (n  2), and Philips IntelliSpace Portal by oscillating singular value decompositions approach (rCBV (ISP; Philips Healthcare, Best, the Netherlands) (n  1). definition 2). S12 used 3 different rCBV definitions within the For PM methods, most sites defined rCBV  Philips ISP platform: a “model-free” option that integrates the t  t R ⁄ R area underneath the signal intensity curve (rCBV definition 1) and used the Boxerman–Schmain- 0 0 2,tumorBSW 2,NAWM 16). A few (17),a“-variate” option that integrates the area underneath the da–Weiskoff (BSW) method for leakage correction ( sites submitted results that deviated from this postprocessing signal intensity curve that has been fit to a -variate function convention by alternative rCBV definitions (S06, S12) and dif- (rCBV definition 2), and a “leakage correction” option that ferences in integration limits as determined by the software integrates the area underneath the computed delta R2* curve (S05). These differences are highlighted in Table 1. Site 06 after a modified BSW leakage correction method is applied defined CBV by the area under the curve of the deconvolved (rCBV definition 3). To be clear, the first 2 options of the Philips residue function. This deconvolved residue function was deter- ISP do not apply any sort of leakage correction algorithm to the mined by singular value decompositions (rCBV definition 1) and data. S05 included CBV maps calculated using the default inte- | | 112 TOMOGRAPHY.ORG VOLUME 5 NUMBER 1 MARCH 2019 Multisite rCBV Consistency Using a DRO Table 2. Summary of Participating Teams’ PMs Site Normalized Leakage Correction Number Software CBV Definition to NAWM? Integration Limits Method Comments 01 01: In-house processing AUC of the R2* time No Time points: 2 to 64 (93 sec) BSW leakage correction Manual inspection of pre- course method and post- contrast points for rCBV integration 02 01: IB Neuro AUC of the R2* time Yes automatically detected BSW leakage correction Default IB Neuro settings course (default option) method for rCBV 03 01: 3DSlicer AUC of the R2* time No 118 seconds BSW leakage correction No thresholding course method 02: nordicICE AUC of the R2* time Yes Time points: 2 to 121 BSW leakage correction course (178.5 sec) method 03: PGUI AUC of the R2* time No Time points: 2 to 121 BSW leakage correction No thresholding, but course (178.5 sec) method smoothing applied 04 01: IB Neuro AUC of the R2* time Yes automatically detected BSW leakage correction course (default option) method 05 01: IB Neuro (Integration limits 1) AUC of the R2* time Yes automatically detected BSW leakage correction course (default option) method 02: IB Neuro (Integration limits 2) AUC of the R2* time Yes 180 seconds (all time BSW leakage correction course points) method 06 01: PGUI (rCBV definition 1) Deconvolution of the No Time points: 5 to 121 BSW leakage correction residue function (174 sec) method (SVD) 02: PGUI (rCBV definition 2) Deconvolution of the No Time points: 5 to 121 BSW leakage correction residue function (174 sec) method (oSVD) 07 01: In-house processing AUC of the R2* time No automatically detected BSW leakage correction course (default option) method 08 01: In-house processing AUC of the R2* time Yes 90 sec BSW leakage correction course method 09 01: IB Neuro AUC of the R2* time Yes automatically detected BSW leakage correction Did not use the entire course (default option) method NAWM ROI - instead useda6mm  6mm (225 pixels) ROI 10 01: In-house processing AUC of the R2* time No 171 sec BSW leakage correction R2* maps were course method smoothed with a 5 5 Gaussian window that had an FWHM value of 3 mm 11 n/a 12 01: Philips ISP (rCBV definition 1) AUC of the SI time No Based on the characteristics No leakage correction course of signal time curves method 02: Philips ISP (rCBV definition 2) AUC of the SI time No Based on the characteristics No leakage correction course fitted to a of signal time curves method gamma-variate 03: Philips ISP (rCBV definition 3) AUC of the R2* time No 180 s BSW leakage correction course method gration limits set by IB Neuro (integration limits 1) and manu- 15). No thresholding, smoothing, or quality assess- where applicable ( ally chose all time points in IB Neuro (integration limits 2). A ment was done before rCBV calculations when analyzed by the man- little less than 50% of the submitted rCBV maps were normalized aging center. to the NAWM. To compare maps, the managing site normalized tumor CBV to the mean NAWM CBV of all pixels when neces- Statistics sary. Specifics on site-specific postprocessing steps are outlined To evaluate the consistency of rCBV across sites owing to dif- in Table 2. ferences between IP and PM, the intraclass correlation coeffi- The managing center postprocessed the site-specific DROs with cient (ICC) was calculated. Furthermore, to evaluate the agree- 120sec  120sec an in-house script by defining rCBV  R ⁄ ment of rCBV between sites and a reference (SIP), the 95% limits 0 0 2,tumorBSW R , where the conventional R of agreement (LOA) were extracted from a Bland–Altman anal- curves in the tumor were 2,NAWM 2 ysis. Variability of rCBV was assessed across a distribution of corrected for leakage effects using the BSW method. In our recent rCBV values by calculating the covariance (CV) across sites. study, the CBV was found to be the most accurate by using these Lastly, Lin’s correlation coefficient was calculated for rCBV specific PM steps, and thus was chosen to be used as the reference | | TOMOGRAPHY.ORG VOLUME 5 NUMBER 1 MARCH 2019 113 Multisite rCBV Consistency Using a DRO Table 3. Intraclass Correlation Coefficient Results for Each Phase of the Study for Computed rCBV from the Simulated Intact-BBB and Disrupted-BBB DRO Site-Specific IP Constant IP Site-Specific IP w/Constant PM w/Site-Specific PM w/Site-Specific PM Intact-BBB 0.970 0.690 0.641 Disrupted-BBB 0.879 0.439 0.380 trans between the intact-BBB and disrupted-BBB DROs for each per- LOA are observed for even the K 0 case, where no leakage mutation of IP and PM to determine the agreement of rCBV after correction is applied during postprocessing. The analysis software leakage correction was applied. All statistical calculations were that show the smallest 95% LOA with the reference are in-house done in MATLAB R2018a (The MathWorks Inc., Natick, MA) by processing scrips (S01_PM01, S08_PM01), IB Neuro (S02_PM01, the managing center. S04_PM01, S05_PM01, S05_PM02, S09_PM01), nordicICE (S03_PM02), and the “model-free option” in Philips ISP (S12_PM01). For phase III (Figure 1C), 9 out of the 24 sites show a RESULTS trans tight 95% LOA and relatively no bias when compared to the SP In general, the ICC decreases when K 0, that is, disrupted- reference (S01_IP01_PM01, S01_IP01_PM02, S03_IP02_PM02, BBB (Table 3) for all the 3 phases of this study. High agreement S04_IP01_PM01, S04_IP03_PM01, S05_IP01_PM01, S05_IP01_ is observed across sites when a constant PM is applied to site- trans PM02, S05_IP02_PM01, S05_IP02_PM02) for the K  0 case. specific IP (ICC  0.879). However, when site-specific PMs are These 4 sites implemented nordicICE, IB Neuro, and an in-house applied to either a constant IP or to their site-specific IP, the postprocessing script. agreement is quite poor (ICC  0.439 and 0.380, respectively). Figure 2 illustrates the CV as a function of rCBV across Figure 1 shows consistency in rCBV measurements for all 3 DROs for all voxels. The covariance across DROs (n  19, phases of this study when compared with the reference. For each Phase I trans site, the 95% LOA of both K n  17 n  25) was calculated in the 10 000 tumor 0 (gray lines, intact-BBB) and Phase II Phase III trans K - voxels and plotted against the mean rCBV of each voxel across 0 (black lines, disrupted-BBB) are indicated in compar trans trans ison to the reference (see Table 1 for site ID descriptions). For DROs. The DRO simulated with K  0 (gray circles) and K  0 phase I (Figure 1A), the 95% LOA are generally fairly narrow and (black circles) is plotted along with the mean CV (horizontal line trans centered around the mean rCBV of the reference for the K plots) across all voxels. This figure does not assume a reference 0 case. A few exceptions (S02_IP01, S10_IP01, and S12_IP01) for calculations. In general, the CV increases for each phase show larger 95% LOA and a negative bias compared with the when more freedom is allowed in the rCBV calculations for both other sites. The first 2 sites (S02 and S10) did not use a contrast IP and PM. For phase I (Figure 2A), the average CV is 4% and it trans agent preload unlike the other sites, while the third site (S12) remains fairly flat over the rCBV distribution for K  0. trans used 1/3 standard dose for a preload. Sites S09_IP01 and However, when K  0, the average CV rose to 17% and S10_IP01, although centered around the reference’s mean rCBV, exponentially decreased from roughly 60% to 10% as rCBV also express wider ranges of 95% LOA compared with other sites. increased. For phases II and III (Figure 2, B and C, respectively), trans These 2 sites have markedly lower TE and use less than a full the CV is observed to exponentially decrease for both K trans standard dose compared with the other sites. Much larger LOA are cases. For phase II, the average CV is 18% and 30% for K seen for phase II in Figure 1B) than for that in Figure 1A. Large 95% 0 and 0, respectively. As rCBV increases, the CV exponentially Figure 1. Bland–Altman limits of agreement (LOA) against the standard imaging protocol (SIP) plotted for site-specific IP w/constant postprocessing method (PM) (A), constant IP w/site-specific PM (B), and site-specific IP w/site-specific PM (C). The vertical dashed line is the mean rCBV across 10 000 voxels for the SIP. | | 114 TOMOGRAPHY.ORG VOLUME 5 NUMBER 1 MARCH 2019 Multisite rCBV Consistency Using a DRO Figure 2. The covariance (CV%) across all relative cerebral blood volume (rCBV) maps for each of the 10 000 voxels plot- ted across the mean rCBV of the voxels for site-specific IP w/constant PM (A), constant IP w/site-specific PM (B), and site-spe- trans trans cific IP w/site-specific PM (C). Results from the K  0 (light gray) and K  0 (black) are included with their mean CV% across all 10 000 voxels indicated for the horizontal lines. For all 3 phases of this study, the largest variation in rCBV occurs trans at the low rCBV range for K  0, and CV% increases when more freedom was introduced in the choice of IPs and PMs. trans decreases from roughly 80% to 20% for both K cases. For the third black bar is the SIP and has a high LCC value, which is trans phase III, the average CV is 21% and 39% for K  0 and  consistent with previous results (15) and therefore used as ref- 0, respectively. As rCBV increases, the CV exponentially de- erence in Figure 1. Here we observed that most of the sites’ IPs trans creases from roughly 120% to 35% for both K cases. are able to accurately compute CBV—most likely because these Figure 3 examines the agreement between the intact-BBB sites already use IPs similar to the SIP. Three site protocols had trans trans (K  0) and disrupted-BBB (K  0) DRO for each pro- an LCCC  0.8, indicating low rCBV accuracy when leakage cessed rCBV map. The LCC for each analysis combination was effects are introduced: S02_IP01, S10_IP01, and S12_IP01. sorted from the highest (perfect agreement  1) to the lowest (no These protocols also resulted in large LOA and a negative bias as agreement  0) for each of the 3 phases. A high agreement seen in Figure 1. These results indicate that the IP is highly indicates that the processed CBV from the simulated disrupted- sensitive to contrast agent leakage effects even when a leakage BBB DRO had high accuracy when compared to the simulated correction PM algorithm is applied. Constant IP with site-spe- intact DRO where no leakage occurs. Site-specific IP with con- cific PM results are indicated in the dark gray bars in the bar stant PM is shown by the black bars in the bar graph. Note that graph. Here we observe 10 software programs that clearly show Figure 3. A bar plot of Lin’s correlation coefficient (LCC) for each rCBV map for site-specific IP w/constant PM (black), constant IP w/site-specific PM (medium gray), and site-specific IP w/site-specific PM (light gray). Each phase is sorted by the resulting LCC from the highest to the lowest value. A horizontal bar at LCC  0.8 is placed to evaluate agreement good agreement (LCC  0.8). | | TOMOGRAPHY.ORG VOLUME 5 NUMBER 1 MARCH 2019 115 Multisite rCBV Consistency Using a DRO high agreement: in-house scripts (n  4), IB Neuro (n  4), These differences most likely average out when hotspot types of nordicICE (n  1), and “model-free” option in the Philips ISP. analyses are performed. Although most likely sufficient for Lastly, site-specific IP with site-specific PM resulted in 50% of diagnosis of tumor grade, this might not be ideal for longitudi- the rCBV maps with LCC  0.8, most likely owing to a combi- nal assessment of treatment response where voxel-wise analysis nation of variations in IPs and postprocessing as deduced from and/or CBV difference quantification has shown to be more the earlier 2 phases. beneficial (11, 12). Despite this, our results indicating incon- sistent CBV values as more freedom is allowed to the IP and DISCUSSION AND CONCLUSION processing methods is not surprising. Kelm et al. compared Reproducibility in DSC-MRI rCBV is crucial for the success of rCBV measurements using 3 software platforms (IB Neuro, multisite clinical trials. In this study, we have evaluated rCBV FuncTool, and nordicICE) and also found significant varia- consistency owing to differences in both IPs and PMs across 12 tion in rCBV (19). QIN sites using a DRO. The results outlined in this manuscript A limitation to our study is that the ROIs for brain tumor show that standardization of both is warranted. and NAWM have been clearly outlined and predetermined for Our prior DRO investigation highlighted the significant in- analysis. In the context of patient data, allowing users to define fluence of IPs (including preload dosing and pulse sequence ROIs would likely contribute to greater rCBV inconsistency. parameters) on CBV accuracy (15). The findings of this study Schmainda et al. showed high mean CBV agreement when ROIs strongly indicate that differences in the PM can also confound were predetermined (18). In addition, sites were not required to multisite CBV consistency and accuracy. High agreement when determine an AIF for the CBV calculations within this manu- site-specific IP were processed by the managing center most script. likely reflects the similarity of the IP parameters across all the Results from this study and our prior DRO analysis, which sites owing to previous initiatives from the ASFNR that aimed to focused on IP optimization (15), highlight the IPs and PMs that standardize IPs (13). However, it was observed that when no maximize rCBV accuracy and multisite consistency. First, IPs preload was used in the IP (S02_IP01 and S10_IP01), a system- that yield the highest rCBV accuracy and multisite concordance atic negative bias relative to the SIP occurs. Furthermore, a utilize a full-dose contrast agent preload and a full-dose bolus slight negative bias is observed for the sites that administered injection, low (30°) or moderate (65°) flip angle,30 millisecond less than a full standard dose as the main injection (S07_IP01, TE, and a 1.5 millisecond TR. In both studies, the use of lower S08_IP01, S09_IP01, S12_IP01). These 2 findings underscore bolus dose injections (eg, 1/2 dose) were found to substantially potential challenges to comparing CBV changes in a clinical reduce both consistency and accuracy, likely owing to the lower trial from sites that use dissimilar preload and bolus dosing CNR. Second, the 2 studies further show that, even with opti- protocols. Three sites (S01, S04, S05) provided clinical IPs for 1.5 mized IPs, leakage correction should be applied to DSC-MRI T. Sites 01 and 05 used the same IP at both field strengths, and data in brain tumors. Further, the correction algorithms should it was observed that the LOA did widen when compared to the be based on the underlying biophysics and kinetics, such as the 3.0 T protocol. Site 04 used a smaller flip angle at 1.5 T than at BSW correction, as they maximize both accuracy and precision. 3 T; however, a widening of LOA was still observed. Differences Generic leakage correction algorithms (like gamma variate fit- here may warrant further investigation into a standardized 1.5 T ting) that arbitrarily modify the shape of DSC-MRI data to IP; however, for the scope of this paper, high agreement was remove T1 and/or T2* leakage effects are not recommended. It observed when both field strengths were compared together. should be noted that in the IP optimization study (15), a low flip When each site was asked to postprocess the SIP, agreement angle approach (30° with a 30 millisecond TE) with a full-dose decreased substantially as indicated by the ICC and the 95% bolus injection, no preload, and application of BSW leakage LOAs. Interestingly, the disagreement across sites is not isolated correction provided accuracy slightly less than that using the to differences in the leakage correction method, as poor agree- ideal protocol. Studies are currently underway to validate the trans trans ment is also observed with the K clinical potential of this protocol as it could be a compelling 0 case. For the K 0 case, 1 potential source of disagreement in rCBV arises from single-dose option for routine surveillance scans and in clinical whether smoothing is implemented in the software and the CBV trials. definition. Methods 01 and 02 from S12 deviated from the Although great efforts have been made to standardize DSC- traditional CBV definition, as these methods calculated CBV MRI imaging acquisition protocols, this study highlights that from the signal intensity curves, potentially losing the biophys- poor CBV agreement can arise when there are variations in trans ics and kinetic properties. For the K processing platforms. Highest agreement is observed when site- 0 case, potential sources of disagreement in rCBV may be attributed to smoothing specific CBV maps are processed by 1 managing center, as might and the algorithms and/or implementation of algorithms used be expected in a clinical trial setting where acquisition and PMs for leakage correction. are predetermined, and/or raw data are sent to a single site for It is challenging to compare the results from this current analysis. However, differences in CBV, especially at low values, study directly to prior ones since we performed a voxel-wise as would be expected with effective therapy, arise when differ- analysis across the DRO, whereas most other studies, like the ent platforms are used. This finding has important implications recent DSC-MRI challenge (18), report comparisons between for comparing CBV values across trials, where variability in mean region of interest tumor values across data that likely trial-specific PMs could confound the comparison of therapeutic exhibits patient-specific rCBV distributions. As seen in Figure 2, effectiveness and/or any attempts to establish thresholds for there is greater variation across platforms at low rCBV values. categorical response (eg, predetermined percent changes in | | 116 TOMOGRAPHY.ORG VOLUME 5 NUMBER 1 MARCH 2019 Multisite rCBV Consistency Using a DRO rCBV values that could be used to refine RANO criteria). To to that in the RSNA DCE-MRI Profile (20), for scanners and overcome these challenges and to ensure the successful use of acquisition methods to be used in clinical trials (eg, using a rCBV as a clinical trial biomarker, it is critical that the DSC-MRI validated phantom) and the software used for DSC-MRI analysis community establish qualifying and validating criteria, similar (eg, using a DRO where the ground truth is known). ACKNOWLEDGMENTS NIH/NCI R01CA213158 (LCB, NS, CCQ); NIH/NCI U01CA207091 (AJM, MCP); Disclosure: No disclosures to report. NIH/NCI U01CA166104 and P01CA085878 (DM, TLC); NIH/NCI U01CA142565 Conflicts of Interest: The authors have no conflicts of interest to declare. (CW, AGS, TEY, NR); NIH/NCI U01 CA176110 (KMS, MAP). REFERENCES 1. Fink JR, Muzi M, Peck M, Krohn KA. Multimodality brain tumor imaging: MR im- 11. Galbán CJ, Lemasson B, Hoff BA, Johnson TD, Sundgren PC, Tsien C, Chenevert aging, PET, and PET/MR imaging. J Nucl Med. 2015;56:1554–1561. TL, Ross BD. Development of a multiparametric voxel-based magnetic resonance 2. Boxerman JL, Ellingson BM, Jeyapalan S, Elinzano H, Harris RJ, Rogg JM, Pope imaging biomarker for early cancer therapeutic response assessment. Tomogra- WB, Safran H. Longitudinal DSC-MRI for distinguishing tumor recurrence from phy. 2015;1:44–52. pseudoprogression in patients with a high-grade glioma. Am J Clin Oncol. 2014; 12. Boxerman JL, Zhang Z, Safriel Y, Larvie M, Snyder BS, Jain R, Chi TL, Sorensen 0:1–7. AG, Gilbert MR, Barboriak DP. Early post-bevacizumab progression on contrast- 3. Ellingson BM, Zaw T, Cloughesy TF, Naeini KM, Lalezari S, Mong S, Lai A, Ng- enhanced MRI as a prognostic marker for overall survival in recurrent glioblas- hiemphu PL, Pope WB. Comparison between intensity normalization techniques toma: results from the ACRIN 6677/RTOG 0625 Central Reader Study. Neuro for dynamic susceptibility contrast (DSC)-MRI estimates of cerebral blood volume Oncol. 2013;15:945–954. (CBV) in human gliomas. J Magn Reson Imaging. 2012;35:1472–1477. 13. Welker K, Boxerman J, Kalnin A, Kaufmann T, Shiroishi M, Wintermark M; 4. Chaskis C, Stadnik T, Michotte A, Van Rompaey K, D’Haens J. Prognostic value American Society of Functional Neuroradiology MR Perfusion Standards and of perfusion-weighted imaging in brain glioma: a prospective study. Acta Neuro- Practice Subcommittee of the ASFNR Clinical Practice Committee. ASFNR recom- chir (Wien). 2006;148:277–285. mendations for clinical performance of MR dynamic susceptibility contrast perfu- 5. Maia ACM, Malheiros SMF, da Rocha AJ, Stávale JN, Guimarães IF, Borges LR, sion imaging of the brain. AJNR. Am J Neuroradiol. 2015;36:E41–E51. Santos AJ, da Silva CJ, de Melo JG, Lanzoni OP, Gabbai AA, Ferraz FA. Stereo- 14. Semmineh NB, Stokes AM, Bell LC, Boxerman JL, Quarles CC. A population- tactic biopsy guidance in adults with supratentorial nonenhancing gliomas: role based digital reference object (DRO) for optimizing dynamic susceptibility con- of perfusion-weighted magnetic resonance imaging. J Neurosurg. 2004;101: trast (DSC)-MRI methods for clinical trials. Tomography. 2017;3:41–49. 970–976. 15. Semmineh N, Bell L, Stokes A, Hu L, Boxerman J, Quarles C. Optimization of 6. Danchaivijitr N, Waldman AD, Tozer DJ, Benton CE, Brasil Caseiras G, Tofts PS, acquisition and analysis methods for clinical dynamic susceptibility contrast (DSC) Rees JH, Jäger HR. Low-grade gliomas: do changes in rCBV measurements at lon- MRI using a population-based digital reference object. AJNR Am J Neuroradiol. gitudinal perfusion-weighted MR imaging predict malignant transformation? Radi- 2018;39:1981–1988. ology. 2008;247:170–178. 16. Boxerman JL, Schmainda KM, Weisskoff RM. Relative cerebral blood volume 7. Rossi Espagnet MC, Romano A, Mancuso V, Cicone F, Napolitano A, Scaringi maps corrected for contrast agent extravasation significantly correlate with gli- C, Minniti G, Bozzao A. Multiparametric evaluation of low grade gliomas at fol- oma tumor grade, whereas uncorrected maps do not. AJNR Am J Neuroradiol. low-up: comparison between diffusion and perfusion MR with 2006;27:859–867. F-FDOPA PET. Br J Radiol. 2016;89:20160476. 17. Meyer-Baese A, Lange O, Wismueller A, Hurdal MK. Analysis of dynamic sus- 8. Boxerman JL, Paulson ES, Prah MA, Schmainda KM. The effect of pulse sequence ceptibility contrast MRI time series based on unsupervised clustering methods. parameters and contrast agent dose on percentage signal recovery in DSC-MRI: IEEE Trans Inf Technol Biomed. 2007;11:563–573. implications for clinical applications. AJNR Am J Neuroradiol. 2013;34:1364– 18. Schmainda KM, Prah MA, Rand SD, Liu Y, Logan B, Muzi M, Rane SD, Da X, 1369. Yen YF, Kalpathy-Cramer J, Chenevert TL, Hoff B, Ross B, Cao Y, Aryal MP, Erick- 9. Hu LS, Baxter LC, Smith KA, Feuerstein BG, Karis JP, Eschbacher JM, Coons SW, son B, Korfiatis P, Dondlinger T, Bell L, Hu L, Kinahan PE, Quarles CC. Multisite Nakaji P, Yeh RF, Debbins J, Heiserman JE. Relative cerebral blood volume val- concordance of DSC-MRI analysis for brain tumors: results of a National Cancer ues to differentiate high-grade glioma recurrence from posttreatment radiation Institute Quantitative Imaging Network Collaborative Project. AJNR Am J Neuro- effect: direct correlation between image-guided tissue histopathology and local- radiol. 2018;39:1008–1016. ized dynamic susceptibility-weighted contrast-enhanced perfusion MR imaging 19. Kelm ZS, Korfiatis PD, Lingineni RK, et al. Variability and accuracy of different measurements. Am J Neuroradiol. 2009;30:552–558. software packages for dynamic susceptibility contrast magnetic resonance imag- 10. Barajas RF, Chang JS, Segal MR, Parsa AT, McDermott MW, Berger MS, Cha S. ing for distinguishing glioblastoma progression from pseudoprogression. J Med Differentiation of recurrent glioblastoma multiforme from radiation necrosis after imaging (Bellingham). 2015;2:26001. external beam radiation therapy with dynamic susceptibility-weighted contrast- 20. DCE MRI Technical Committee. DCE MRI Quantification Profile, Quantitative Imaging enhanced perfusion MR imaging. Radiology. 2009;253:486–496. Biomarkers Alliance. 2012; Version 1.0. Available from: http://rsna.org/QIBA_.aspx. | | TOMOGRAPHY.ORG VOLUME 5 NUMBER 1 MARCH 2019 117 http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Tomography Multidisciplinary Digital Publishing Institute

Evaluating Multisite rCBV Consistency from DSC-MRI Imaging Protocols and Postprocessing Software Across the NCI Quantitative Imaging Network Sites Using a Digital Reference Object (DRO)

Download PDF
8 pages

Loading next page...
 
/lp/multidisciplinary-digital-publishing-institute/evaluating-multisite-rcbv-consistency-from-dsc-mri-imaging-protocols-rTw5hrzsdB

References

References for this paper are not available at this time. We will be adding them shortly, thank you for your patience.

Publisher
Multidisciplinary Digital Publishing Institute
Copyright
© 1996-2021 MDPI (Basel, Switzerland) unless otherwise stated Disclaimer The statements, opinions and data contained in the journals are solely those of the individual authors and contributors and not of the publisher and the editor(s). MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations. Terms and Conditions Privacy Policy
ISSN
2379-139X
DOI
10.18383/j.tom.2018.00041
Publisher site
See Article on Publisher Site

Abstract

RESEARCH ARTICLE Evaluating Multisite rCBV Consistency from DSC-MRI Imaging Protocols and Postprocessing Software Across the NCI Quantitative Imaging Network Sites Using a Digital Reference Object (DRO) 1 1 2 2 2 Laura C. Bell , Natenael Semmineh , Hongyu An , Cihat Eldeniz , Richard Wahl , 3 3 4 4 5 Kathleen M. Schmainda , Melissa A. Prah , Bradley J. Erickson , Panagiotis Korfiatis , Chengyue Wu , 5 5 5 6 6 Anna G. Sorace , Thomas E. Yankeelov , Neal Rutledge , Thomas L. Chenevert , Dariya Malyarenko , 7 7 8 8 9,10 11 Yichu Liu , Andrew Brenner , Leland S. Hu , Yuxiang Zhou , Jerrold L. Boxerman , Yi-Fen Yen , 11 11 12 13 Jayashree Kalpathy-Cramer , Andrew L. Beers , Mark Muzi , Ananth J. Madhuranthakam , 13 13,14 1 Marco Pinho , Brian Johnson , and C. Chad Quarles 1 2 Division of Neuroimaging Research, Barrow Neurological Institute, Phoenix, AZ; Mallinckrodt Institute of Radiology, Washington University in St. Louis, St. Louis, MO; 4 5 Departments of Radiology and Biophysics, Medical College of Wisconsin, Wauwatosa, WI; Department of Radiology, Mayo Clinic, Rochester, MN; Department of 6 7 Diagnostic Medicine, University of Texas at Austin, Austin, TX; Department of Radiology, University of Michigan, Ann Arbor, MI; UT Health San Antonio, San Antonio, TX; 9 10 Department of Radiology, Mayo Clinic, Scottsdale, AZ; Department of Diagnostic Imaging, Rhode Island Hospital, Providence, RI; Alpert Medical School of Brown 11 12 University, Providence, RI; Department of Radiology, Massachusetts General Hospital, Boston, MA; Department of Radiology, University of Washington, Seattle, 13 14 Washington; UT Southwestern Medical Center, Dallas, TX; and Philips Healthcare, Gainesville, FL Corresponding Author: Key Words: DSC-MRI, relative cerebral blood volume, standardization, multisite consistency, Laura C. Bell, PhD reproducibility Division of Neuroimaging Research, Abbreviations: Relative cerebral blood volume (rCBV), postprocessing methods (PMs), imaging Barrow Neurological Institute, Phoenix, AZ, 85013; protocols (IPs), dynamic susceptibility contrast magnetic resonance imaging (DSC-MRI), Quantitative E-mail: [email protected] Imaging Network (QIN), American Society of Functional Neuroradiology (ASFNR), digital reference object (DRO), standard imaging protocol (SIP), intraclass correlation coefficient (ICC), limits of agreement (LOA), covariance (CV), echo time (TE), normal appearing white matter (NAWM) The use of rCBV as a response metric in clinical trials has been hampered, in part, due to variations in the biomarker consistency and associated interpretation across sites, stemming from differences in image acquisi- tion and post-processing methods. This study leveraged a dynamic susceptibility contrast magnetic resonance imaging digital reference object to characterize rCBV consistency across 12 sites participating in the Quanti- tative Imaging Network (QIN), specifically focusing on differences in site-specific imaging protocols (IPs; n 17), and PMs (n  19) and differences due to site-specific IPs and PMs (n  25). Thus, high agreement across sites occurs when 1 managing center processes rCBV despite slight variations in the IP. This result is most likely supported by current initiatives to standardize IPs. However, marked intersite disagreement was observed when site-specific software was applied for rCBV measurements. This study’s results have important implications for comparing rCBV values across sites and trials, where variability in PMs could confound the comparison of therapeutic effectiveness and/or any attempts to establish thresholds for categorical response to therapy. To overcome these challenges and ensure the successful use of rCBV as a clinical trial biomarker, we recommend the establishment of qualifying and validating site- and trial-specific criteria for scanners and acquisition methods (eg, using a validated phantom) and the software tools used for dynamic susceptibility contrast magnetic resonance imaging analysis (eg, using a digital reference object where the ground truth is known). INTRODUCTION aid in diagnosis (1), detecting treatment response (2, 3), guiding The relative cerebral blood volume (rCBV), derived from dy- biopsies (4, 5), and reliable differentiation of post-treatment namic susceptibility contrast magnetic resonance imaging radiation effects and tumor progression (6-10). It is also increas- (DSC-MRI), is an established biomarker of glioma status that can ingly leveraged as a biomarker of early therapeutic response in © 2019 The Authors. Published by Grapho Publications, LLC This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). ISSN 2379-1381 http://dx.doi.org/10.18383/j.tom.2018.00041 | | 110 TOMOGRAPHY.ORG VOLUME 5 NUMBER 1 MARCH 2019 ABSTRACT Multisite rCBV Consistency Using a DRO clinical trials (11, 12). However, variations in image acquisition each of these submitted site-specific IP DROs to evaluate and postprocessing methods (PMs) can limit rCBV reproducibil- differences owing to the IP provided. ity, potentially diminishing its clinical utility. To promote rCBV • Phase II (“constant IP w/site PM”) involved analysis of a reproducibility across institutions, many national initiatives are “standard imaging protocol” (SIP), which represents DSC- underway to standardize DSC-MRI acquisition and PMs, includ- MRI data acquired using the IP recommended by ASFNR 13). Each site was asked to process DSC-MRI DRO data ing National Cancer Institute’s Quantitative Imaging Network derived from the SIP. Some sites choose to use multiple (QIN), Radiological Society of North America’s Quantitative Im- commercially available software packages (site 03) and aging Biomarkers Alliance (QIBA), and the National Brain Tu- different rCBV definitions (sites 05, 06, 12), yielding a total mor Society’s Jumpstarting Brain Tumor Drug Development of 17 submitted rCBV maps. Coalition. Recent imaging protocol (IP) recommendations by the • Phase III (“site IP w/site PM”) required each site to calculate American Society of Functional Neuroradiology (ASFNR) has rCBV maps using their PM of choice and the site-specific served as the first step in standardizing DSC-MRI protocols for DRO data. Combining the possible permutations owing to clinical applications (13). choice of IP and PM from phases I–II, a total of 25 rCBV To aid in this effort, 12 institutions within the QIN aimed to maps were submitted. investigate and determine the current rCBV reproducibility us- ing a recently developed and validated in silico digital reference All sites but 1 completed all 3 phases of the challenge. Site 11 object (DRO) that is representative of a wide range of possible completed only phase I, and these results are included in this glioma magnetic resonance signals (14). Leveraging this DRO study. enables us as a community to determine the multisite consis- DRO Simulations tency in rCBV owing to varying permutations of imaging ac- The DSC-MRI signals for each IP were simulated using a recently quisition parameters and postprocessing steps. In specific, our developed and validated population-based DRO that was trained goals are to characterize rCBV consistency under conditions to generate realistic signals using in vivo data from 40 000 where there exist: (1) variations in the site-specific imaging voxels derived from patient data (14). The resulting DRO, which acquisition parameters (PMs held constant), (2) variations in contains 10 000 unique voxels, reflects the distribution of per- only site-specific PMs (IP held constant), and (3) variations fusion, permeability, precontrast T1, T2*, diffusion coefficients, owing to site-specific imaging and postprocessing protocols. and the vascular and cellular features found in patients with Results from this community-based challenge will help steer high-grade glioma. Using this DRO, the DSC-MRI signals and standardization of DSC-MRI rCBV protocols with the hope that resulting rCBV values can be computed for any combination of it can be successfully translated to the clinical setting. preload dosing scheme, contrast agent choice (by varying T1 relaxivities specific to the contrast agent), pulse sequence pa- MATERIALS AND METHODS rameters, and postprocessing protocol. For the purposes of this This National Cancer Institute QIN DSC-DRO challenge project study, the DRO consisted of tumor voxels simulated under two was proposed and organized by the investigators at Barrow blood-brain-barrier (BBB) conditions to recapitulate DSC-MRI trans Neurological Institute (BNI). Eleven centers participated in this signals from an intact-BBB (K 0) and a disrupted-BBB trans project: BNI (the managing center), Brown University (BU), Mas- (K 0). In addition to the tumor voxels, normal appearing trans sachusetts General Hospital (MGH), Mayo Clinic Arizona (Mayo white matter (NAWM) voxels (K 0) were simulated to AZ), Mayo Clinic Minnesota (Mayo MN), Medical College of normalize CBV. For the purposes of comparing site-to-site con- Wisconsin (MCW), University of Michigan (UM1), The Univer- sistency, the SIP that has been postprocessed by the managing sity of Texas Health at San Antonio (UTSA), University of Texas center was considered the reference standard where necessary. at Austin (UT), University of Texas Southwestern Medical Center In our recent study, focused on investigating the influence of IP at Dallas (UTSW), University of Washington (UW), and Wash- on CBV fidelity (15), the SIP yielded CBV values, when corrected ington University (WashU). Unless specifically named, these for contrast agent leakage, that were among the most accurate. participating sites have been anonymized, in no particular order, Site-Specific IP and PM Methods and will be referred to as sites 01–12 as seen in Table 1. Site-specific IP and PM methods are briefly listed in Table 1. This project comprised 3 phases, summarized in the last 3 Overall, IPs were similar across sites. Most sites submitted clin- columns of Table 1, to evaluate the influence of IPs and/or PMs ical DSC IPs for 3 T with 3 sites that also included a 1.5 T IP. on multisite consistency: Overall, the following were the imaging parameters [mode (min- • Phase I (“site IP w/constant PM”) involved each participat- max)]: repetition time  1500 milliseconds (1300 –2560 milli- ing site to submit their current clinical DSC IP to the seconds), echo time (TE)  30 milliseconds (18 –71 millisec- managing center. The managing center then simulated site- onds), flip angle  60° (60°–90°), preload dose  0.05 mmol/kg specific DROs reflecting the IP parameters provided. Some (0 – 0.1 mmol/kg), and injection dose  0.1 mmol/kg (0.05– 0.15 sites provided 1 IP owing to differences in field strengths mmol/kg). Five different gadolinium contrast agents were used (sites 01, 04, and 05), dosing schemes (sites 03 and 10), and across the 12 sites: gadobenate (n  5), gadobutrol (n  3), gad- acquisition method (site 04). In total, 19 different IPs were oterate (n  2), gadoteridol (n  1), and gadopentetate (n  1). For submitted. The managing center postprocessed (specific PMs, there was a mix of software options used, including in-house- based software scripts (n  4), IB Neuro (n  4), 3D Slicer (n  1), details below in “Site-specific IP and PM”) rCBV maps of | | TOMOGRAPHY.ORG VOLUME 5 NUMBER 1 MARCH 2019 111 Multisite rCBV Consistency Using a DRO Table 1. Summary of Participating Teams’ IPs and PMs Imaging Protocol (IP) Scan Protocol Dose Protocol ID Tag for Analysis Time Site IP Constant IP Site IP Site Field TR TE Preload Injection Between Processing Method w/Constant w/Site w/Site Number Strength (ms) (ms) Flip (mmol/kg) (mmol/kg) (min) CA (PM) PP PP PP 01 01:3.0 T 1500 30 60 0.05 0.10 3 Gadobenate 01: In-house processing S01_IP01 S01_PM01 S01_IP01_PM01 02:1.5 T 1500 30 60 0.05 0.10 3 Gadobenate S01_IP02 S01_IP02_PM01 02 01:3.0 T 1600 30 60 0 0.1 n/a Gadobenate 01: IB Neuro S02_IP01 S02_PM01 S02_IP01_PM01 03 01:3.0 T 1500 31 90 0.05 0.15 6.5 Gadoterate 01: 3DSlicer S03_IP01 S03_PM01 S03_IP01_PM01 02:3.0 T 1500 31 90 0.1 0.1 6.5 Gadoterate 02: nordicICE S03_IP02 S03_PM02 S03_IP01_PM02 03: PGUI S03_PM03 S03_IP01_PM03 S03_IP02_PM01 S03_IP02_PM02 S03_IP02_PM03 04 01:3.0 T 1500 30 80 0.10 0.10 5 Gadobutrol 01: IB Neuro S04_IP01 S04_PM01 S04_IP01_PM01 02:3.0 T 1500 2,35 80 0 0.10 n/a Gadobutrol S04_IP02 n/a 03:1.5 T 1500 30 72 0.10 0.10 5 Gadobutrol S04_IP03 S04_IP03_PM01 04:1.5 T 1500 2,35 72 0 0.10 n/a Gadobutrol S04_IP04 n/a 05 01:3.0 T 1300 30 60 0.025 0.10 5 Gadobutrol 01: IB Neuro (Integration S05_IP01 S05_PM01 S05_IP01_PM01 limits 1) 02:1.5 T 1300 30 60 0.025 0.10 5 Gadobutrol 02: IB Neuro (Integration S05_IP02 S05_PM02 S05_IP01_PM02 limits 2) S05_IP02_PM01 S05_IP02_PM02 06 01:3.0 T 1500 30 75 0.10 0.10 5 Gadoteridol 01: PGUI (rCBV definition 1) S06_IP01 S06_PM01 S06_IP01_PM01 02: PGUI (rCBV definition 2) S06_PM02 S06_IP01_PM01 07 01:3.0 T 1500 30 65 0.025 0.075 6 Gadobenate 01: In-house processing S07_IP01 S07_PM01 S07_IP01_PM01 08 01:3.0 T 1500 21 60 0.10 0.05 6 Gadobenate 01: In-house processing S08_IP01 S08_PM01 S08_IP01_PM01 09 01:3.0 T 1500 18 60 0.05 0.05 6 Gadobenate 01: IB Neuro S09_IP01 S09_PM01 S09_IP01_PM01 10 01:3.0 T 1900 36 90 0 0.10 n/a Gadoterate 01: In-house processing S10_IP01 S10_PM01 S10_IP01_PM01 02:3.0 T 1900 36 90 0.10 0.10 5 Gadoterate S10_IP02 S10_IP02_PM01 11 01:3.0 T 2560 71 90 0.025 0.10 2 Gadopentetate n/a S11_IP01 n/a n/a 12 01:3.0 T 1757 30 90 0.033 0.067 8 Gadobutrol 01: Philips ISP (rCBV S12_IP01 S12_PM01 S12_IP01_PM01 definition 1) 02: Philips ISP (rCBV S12_PM02 S12_IP01_PM02 definition 2) 03: Philips ISP (rCBV S12_PM03 S12_IP01_PM03 definition 3) Standard 01:3.0 T 1500 30 60 0.10 0.10 5 Gadopentetate n/a SIP n/a n/a Protocol Total :12 19 17 25 Excludes the standard protocol. nordicICE (n  1), PGUI (n  2), and Philips IntelliSpace Portal by oscillating singular value decompositions approach (rCBV (ISP; Philips Healthcare, Best, the Netherlands) (n  1). definition 2). S12 used 3 different rCBV definitions within the For PM methods, most sites defined rCBV  Philips ISP platform: a “model-free” option that integrates the t  t R ⁄ R area underneath the signal intensity curve (rCBV definition 1) and used the Boxerman–Schmain- 0 0 2,tumorBSW 2,NAWM 16). A few (17),a“-variate” option that integrates the area underneath the da–Weiskoff (BSW) method for leakage correction ( sites submitted results that deviated from this postprocessing signal intensity curve that has been fit to a -variate function convention by alternative rCBV definitions (S06, S12) and dif- (rCBV definition 2), and a “leakage correction” option that ferences in integration limits as determined by the software integrates the area underneath the computed delta R2* curve (S05). These differences are highlighted in Table 1. Site 06 after a modified BSW leakage correction method is applied defined CBV by the area under the curve of the deconvolved (rCBV definition 3). To be clear, the first 2 options of the Philips residue function. This deconvolved residue function was deter- ISP do not apply any sort of leakage correction algorithm to the mined by singular value decompositions (rCBV definition 1) and data. S05 included CBV maps calculated using the default inte- | | 112 TOMOGRAPHY.ORG VOLUME 5 NUMBER 1 MARCH 2019 Multisite rCBV Consistency Using a DRO Table 2. Summary of Participating Teams’ PMs Site Normalized Leakage Correction Number Software CBV Definition to NAWM? Integration Limits Method Comments 01 01: In-house processing AUC of the R2* time No Time points: 2 to 64 (93 sec) BSW leakage correction Manual inspection of pre- course method and post- contrast points for rCBV integration 02 01: IB Neuro AUC of the R2* time Yes automatically detected BSW leakage correction Default IB Neuro settings course (default option) method for rCBV 03 01: 3DSlicer AUC of the R2* time No 118 seconds BSW leakage correction No thresholding course method 02: nordicICE AUC of the R2* time Yes Time points: 2 to 121 BSW leakage correction course (178.5 sec) method 03: PGUI AUC of the R2* time No Time points: 2 to 121 BSW leakage correction No thresholding, but course (178.5 sec) method smoothing applied 04 01: IB Neuro AUC of the R2* time Yes automatically detected BSW leakage correction course (default option) method 05 01: IB Neuro (Integration limits 1) AUC of the R2* time Yes automatically detected BSW leakage correction course (default option) method 02: IB Neuro (Integration limits 2) AUC of the R2* time Yes 180 seconds (all time BSW leakage correction course points) method 06 01: PGUI (rCBV definition 1) Deconvolution of the No Time points: 5 to 121 BSW leakage correction residue function (174 sec) method (SVD) 02: PGUI (rCBV definition 2) Deconvolution of the No Time points: 5 to 121 BSW leakage correction residue function (174 sec) method (oSVD) 07 01: In-house processing AUC of the R2* time No automatically detected BSW leakage correction course (default option) method 08 01: In-house processing AUC of the R2* time Yes 90 sec BSW leakage correction course method 09 01: IB Neuro AUC of the R2* time Yes automatically detected BSW leakage correction Did not use the entire course (default option) method NAWM ROI - instead useda6mm  6mm (225 pixels) ROI 10 01: In-house processing AUC of the R2* time No 171 sec BSW leakage correction R2* maps were course method smoothed with a 5 5 Gaussian window that had an FWHM value of 3 mm 11 n/a 12 01: Philips ISP (rCBV definition 1) AUC of the SI time No Based on the characteristics No leakage correction course of signal time curves method 02: Philips ISP (rCBV definition 2) AUC of the SI time No Based on the characteristics No leakage correction course fitted to a of signal time curves method gamma-variate 03: Philips ISP (rCBV definition 3) AUC of the R2* time No 180 s BSW leakage correction course method gration limits set by IB Neuro (integration limits 1) and manu- 15). No thresholding, smoothing, or quality assess- where applicable ( ally chose all time points in IB Neuro (integration limits 2). A ment was done before rCBV calculations when analyzed by the man- little less than 50% of the submitted rCBV maps were normalized aging center. to the NAWM. To compare maps, the managing site normalized tumor CBV to the mean NAWM CBV of all pixels when neces- Statistics sary. Specifics on site-specific postprocessing steps are outlined To evaluate the consistency of rCBV across sites owing to dif- in Table 2. ferences between IP and PM, the intraclass correlation coeffi- The managing center postprocessed the site-specific DROs with cient (ICC) was calculated. Furthermore, to evaluate the agree- 120sec  120sec an in-house script by defining rCBV  R ⁄ ment of rCBV between sites and a reference (SIP), the 95% limits 0 0 2,tumorBSW R , where the conventional R of agreement (LOA) were extracted from a Bland–Altman anal- curves in the tumor were 2,NAWM 2 ysis. Variability of rCBV was assessed across a distribution of corrected for leakage effects using the BSW method. In our recent rCBV values by calculating the covariance (CV) across sites. study, the CBV was found to be the most accurate by using these Lastly, Lin’s correlation coefficient was calculated for rCBV specific PM steps, and thus was chosen to be used as the reference | | TOMOGRAPHY.ORG VOLUME 5 NUMBER 1 MARCH 2019 113 Multisite rCBV Consistency Using a DRO Table 3. Intraclass Correlation Coefficient Results for Each Phase of the Study for Computed rCBV from the Simulated Intact-BBB and Disrupted-BBB DRO Site-Specific IP Constant IP Site-Specific IP w/Constant PM w/Site-Specific PM w/Site-Specific PM Intact-BBB 0.970 0.690 0.641 Disrupted-BBB 0.879 0.439 0.380 trans between the intact-BBB and disrupted-BBB DROs for each per- LOA are observed for even the K 0 case, where no leakage mutation of IP and PM to determine the agreement of rCBV after correction is applied during postprocessing. The analysis software leakage correction was applied. All statistical calculations were that show the smallest 95% LOA with the reference are in-house done in MATLAB R2018a (The MathWorks Inc., Natick, MA) by processing scrips (S01_PM01, S08_PM01), IB Neuro (S02_PM01, the managing center. S04_PM01, S05_PM01, S05_PM02, S09_PM01), nordicICE (S03_PM02), and the “model-free option” in Philips ISP (S12_PM01). For phase III (Figure 1C), 9 out of the 24 sites show a RESULTS trans tight 95% LOA and relatively no bias when compared to the SP In general, the ICC decreases when K 0, that is, disrupted- reference (S01_IP01_PM01, S01_IP01_PM02, S03_IP02_PM02, BBB (Table 3) for all the 3 phases of this study. High agreement S04_IP01_PM01, S04_IP03_PM01, S05_IP01_PM01, S05_IP01_ is observed across sites when a constant PM is applied to site- trans PM02, S05_IP02_PM01, S05_IP02_PM02) for the K  0 case. specific IP (ICC  0.879). However, when site-specific PMs are These 4 sites implemented nordicICE, IB Neuro, and an in-house applied to either a constant IP or to their site-specific IP, the postprocessing script. agreement is quite poor (ICC  0.439 and 0.380, respectively). Figure 2 illustrates the CV as a function of rCBV across Figure 1 shows consistency in rCBV measurements for all 3 DROs for all voxels. The covariance across DROs (n  19, phases of this study when compared with the reference. For each Phase I trans site, the 95% LOA of both K n  17 n  25) was calculated in the 10 000 tumor 0 (gray lines, intact-BBB) and Phase II Phase III trans K - voxels and plotted against the mean rCBV of each voxel across 0 (black lines, disrupted-BBB) are indicated in compar trans trans ison to the reference (see Table 1 for site ID descriptions). For DROs. The DRO simulated with K  0 (gray circles) and K  0 phase I (Figure 1A), the 95% LOA are generally fairly narrow and (black circles) is plotted along with the mean CV (horizontal line trans centered around the mean rCBV of the reference for the K plots) across all voxels. This figure does not assume a reference 0 case. A few exceptions (S02_IP01, S10_IP01, and S12_IP01) for calculations. In general, the CV increases for each phase show larger 95% LOA and a negative bias compared with the when more freedom is allowed in the rCBV calculations for both other sites. The first 2 sites (S02 and S10) did not use a contrast IP and PM. For phase I (Figure 2A), the average CV is 4% and it trans agent preload unlike the other sites, while the third site (S12) remains fairly flat over the rCBV distribution for K  0. trans used 1/3 standard dose for a preload. Sites S09_IP01 and However, when K  0, the average CV rose to 17% and S10_IP01, although centered around the reference’s mean rCBV, exponentially decreased from roughly 60% to 10% as rCBV also express wider ranges of 95% LOA compared with other sites. increased. For phases II and III (Figure 2, B and C, respectively), trans These 2 sites have markedly lower TE and use less than a full the CV is observed to exponentially decrease for both K trans standard dose compared with the other sites. Much larger LOA are cases. For phase II, the average CV is 18% and 30% for K seen for phase II in Figure 1B) than for that in Figure 1A. Large 95% 0 and 0, respectively. As rCBV increases, the CV exponentially Figure 1. Bland–Altman limits of agreement (LOA) against the standard imaging protocol (SIP) plotted for site-specific IP w/constant postprocessing method (PM) (A), constant IP w/site-specific PM (B), and site-specific IP w/site-specific PM (C). The vertical dashed line is the mean rCBV across 10 000 voxels for the SIP. | | 114 TOMOGRAPHY.ORG VOLUME 5 NUMBER 1 MARCH 2019 Multisite rCBV Consistency Using a DRO Figure 2. The covariance (CV%) across all relative cerebral blood volume (rCBV) maps for each of the 10 000 voxels plot- ted across the mean rCBV of the voxels for site-specific IP w/constant PM (A), constant IP w/site-specific PM (B), and site-spe- trans trans cific IP w/site-specific PM (C). Results from the K  0 (light gray) and K  0 (black) are included with their mean CV% across all 10 000 voxels indicated for the horizontal lines. For all 3 phases of this study, the largest variation in rCBV occurs trans at the low rCBV range for K  0, and CV% increases when more freedom was introduced in the choice of IPs and PMs. trans decreases from roughly 80% to 20% for both K cases. For the third black bar is the SIP and has a high LCC value, which is trans phase III, the average CV is 21% and 39% for K  0 and  consistent with previous results (15) and therefore used as ref- 0, respectively. As rCBV increases, the CV exponentially de- erence in Figure 1. Here we observed that most of the sites’ IPs trans creases from roughly 120% to 35% for both K cases. are able to accurately compute CBV—most likely because these Figure 3 examines the agreement between the intact-BBB sites already use IPs similar to the SIP. Three site protocols had trans trans (K  0) and disrupted-BBB (K  0) DRO for each pro- an LCCC  0.8, indicating low rCBV accuracy when leakage cessed rCBV map. The LCC for each analysis combination was effects are introduced: S02_IP01, S10_IP01, and S12_IP01. sorted from the highest (perfect agreement  1) to the lowest (no These protocols also resulted in large LOA and a negative bias as agreement  0) for each of the 3 phases. A high agreement seen in Figure 1. These results indicate that the IP is highly indicates that the processed CBV from the simulated disrupted- sensitive to contrast agent leakage effects even when a leakage BBB DRO had high accuracy when compared to the simulated correction PM algorithm is applied. Constant IP with site-spe- intact DRO where no leakage occurs. Site-specific IP with con- cific PM results are indicated in the dark gray bars in the bar stant PM is shown by the black bars in the bar graph. Note that graph. Here we observe 10 software programs that clearly show Figure 3. A bar plot of Lin’s correlation coefficient (LCC) for each rCBV map for site-specific IP w/constant PM (black), constant IP w/site-specific PM (medium gray), and site-specific IP w/site-specific PM (light gray). Each phase is sorted by the resulting LCC from the highest to the lowest value. A horizontal bar at LCC  0.8 is placed to evaluate agreement good agreement (LCC  0.8). | | TOMOGRAPHY.ORG VOLUME 5 NUMBER 1 MARCH 2019 115 Multisite rCBV Consistency Using a DRO high agreement: in-house scripts (n  4), IB Neuro (n  4), These differences most likely average out when hotspot types of nordicICE (n  1), and “model-free” option in the Philips ISP. analyses are performed. Although most likely sufficient for Lastly, site-specific IP with site-specific PM resulted in 50% of diagnosis of tumor grade, this might not be ideal for longitudi- the rCBV maps with LCC  0.8, most likely owing to a combi- nal assessment of treatment response where voxel-wise analysis nation of variations in IPs and postprocessing as deduced from and/or CBV difference quantification has shown to be more the earlier 2 phases. beneficial (11, 12). Despite this, our results indicating incon- sistent CBV values as more freedom is allowed to the IP and DISCUSSION AND CONCLUSION processing methods is not surprising. Kelm et al. compared Reproducibility in DSC-MRI rCBV is crucial for the success of rCBV measurements using 3 software platforms (IB Neuro, multisite clinical trials. In this study, we have evaluated rCBV FuncTool, and nordicICE) and also found significant varia- consistency owing to differences in both IPs and PMs across 12 tion in rCBV (19). QIN sites using a DRO. The results outlined in this manuscript A limitation to our study is that the ROIs for brain tumor show that standardization of both is warranted. and NAWM have been clearly outlined and predetermined for Our prior DRO investigation highlighted the significant in- analysis. In the context of patient data, allowing users to define fluence of IPs (including preload dosing and pulse sequence ROIs would likely contribute to greater rCBV inconsistency. parameters) on CBV accuracy (15). The findings of this study Schmainda et al. showed high mean CBV agreement when ROIs strongly indicate that differences in the PM can also confound were predetermined (18). In addition, sites were not required to multisite CBV consistency and accuracy. High agreement when determine an AIF for the CBV calculations within this manu- site-specific IP were processed by the managing center most script. likely reflects the similarity of the IP parameters across all the Results from this study and our prior DRO analysis, which sites owing to previous initiatives from the ASFNR that aimed to focused on IP optimization (15), highlight the IPs and PMs that standardize IPs (13). However, it was observed that when no maximize rCBV accuracy and multisite consistency. First, IPs preload was used in the IP (S02_IP01 and S10_IP01), a system- that yield the highest rCBV accuracy and multisite concordance atic negative bias relative to the SIP occurs. Furthermore, a utilize a full-dose contrast agent preload and a full-dose bolus slight negative bias is observed for the sites that administered injection, low (30°) or moderate (65°) flip angle,30 millisecond less than a full standard dose as the main injection (S07_IP01, TE, and a 1.5 millisecond TR. In both studies, the use of lower S08_IP01, S09_IP01, S12_IP01). These 2 findings underscore bolus dose injections (eg, 1/2 dose) were found to substantially potential challenges to comparing CBV changes in a clinical reduce both consistency and accuracy, likely owing to the lower trial from sites that use dissimilar preload and bolus dosing CNR. Second, the 2 studies further show that, even with opti- protocols. Three sites (S01, S04, S05) provided clinical IPs for 1.5 mized IPs, leakage correction should be applied to DSC-MRI T. Sites 01 and 05 used the same IP at both field strengths, and data in brain tumors. Further, the correction algorithms should it was observed that the LOA did widen when compared to the be based on the underlying biophysics and kinetics, such as the 3.0 T protocol. Site 04 used a smaller flip angle at 1.5 T than at BSW correction, as they maximize both accuracy and precision. 3 T; however, a widening of LOA was still observed. Differences Generic leakage correction algorithms (like gamma variate fit- here may warrant further investigation into a standardized 1.5 T ting) that arbitrarily modify the shape of DSC-MRI data to IP; however, for the scope of this paper, high agreement was remove T1 and/or T2* leakage effects are not recommended. It observed when both field strengths were compared together. should be noted that in the IP optimization study (15), a low flip When each site was asked to postprocess the SIP, agreement angle approach (30° with a 30 millisecond TE) with a full-dose decreased substantially as indicated by the ICC and the 95% bolus injection, no preload, and application of BSW leakage LOAs. Interestingly, the disagreement across sites is not isolated correction provided accuracy slightly less than that using the to differences in the leakage correction method, as poor agree- ideal protocol. Studies are currently underway to validate the trans trans ment is also observed with the K clinical potential of this protocol as it could be a compelling 0 case. For the K 0 case, 1 potential source of disagreement in rCBV arises from single-dose option for routine surveillance scans and in clinical whether smoothing is implemented in the software and the CBV trials. definition. Methods 01 and 02 from S12 deviated from the Although great efforts have been made to standardize DSC- traditional CBV definition, as these methods calculated CBV MRI imaging acquisition protocols, this study highlights that from the signal intensity curves, potentially losing the biophys- poor CBV agreement can arise when there are variations in trans ics and kinetic properties. For the K processing platforms. Highest agreement is observed when site- 0 case, potential sources of disagreement in rCBV may be attributed to smoothing specific CBV maps are processed by 1 managing center, as might and the algorithms and/or implementation of algorithms used be expected in a clinical trial setting where acquisition and PMs for leakage correction. are predetermined, and/or raw data are sent to a single site for It is challenging to compare the results from this current analysis. However, differences in CBV, especially at low values, study directly to prior ones since we performed a voxel-wise as would be expected with effective therapy, arise when differ- analysis across the DRO, whereas most other studies, like the ent platforms are used. This finding has important implications recent DSC-MRI challenge (18), report comparisons between for comparing CBV values across trials, where variability in mean region of interest tumor values across data that likely trial-specific PMs could confound the comparison of therapeutic exhibits patient-specific rCBV distributions. As seen in Figure 2, effectiveness and/or any attempts to establish thresholds for there is greater variation across platforms at low rCBV values. categorical response (eg, predetermined percent changes in | | 116 TOMOGRAPHY.ORG VOLUME 5 NUMBER 1 MARCH 2019 Multisite rCBV Consistency Using a DRO rCBV values that could be used to refine RANO criteria). To to that in the RSNA DCE-MRI Profile (20), for scanners and overcome these challenges and to ensure the successful use of acquisition methods to be used in clinical trials (eg, using a rCBV as a clinical trial biomarker, it is critical that the DSC-MRI validated phantom) and the software used for DSC-MRI analysis community establish qualifying and validating criteria, similar (eg, using a DRO where the ground truth is known). ACKNOWLEDGMENTS NIH/NCI R01CA213158 (LCB, NS, CCQ); NIH/NCI U01CA207091 (AJM, MCP); Disclosure: No disclosures to report. NIH/NCI U01CA166104 and P01CA085878 (DM, TLC); NIH/NCI U01CA142565 Conflicts of Interest: The authors have no conflicts of interest to declare. (CW, AGS, TEY, NR); NIH/NCI U01 CA176110 (KMS, MAP). REFERENCES 1. Fink JR, Muzi M, Peck M, Krohn KA. Multimodality brain tumor imaging: MR im- 11. Galbán CJ, Lemasson B, Hoff BA, Johnson TD, Sundgren PC, Tsien C, Chenevert aging, PET, and PET/MR imaging. J Nucl Med. 2015;56:1554–1561. TL, Ross BD. Development of a multiparametric voxel-based magnetic resonance 2. Boxerman JL, Ellingson BM, Jeyapalan S, Elinzano H, Harris RJ, Rogg JM, Pope imaging biomarker for early cancer therapeutic response assessment. Tomogra- WB, Safran H. Longitudinal DSC-MRI for distinguishing tumor recurrence from phy. 2015;1:44–52. pseudoprogression in patients with a high-grade glioma. Am J Clin Oncol. 2014; 12. Boxerman JL, Zhang Z, Safriel Y, Larvie M, Snyder BS, Jain R, Chi TL, Sorensen 0:1–7. AG, Gilbert MR, Barboriak DP. Early post-bevacizumab progression on contrast- 3. Ellingson BM, Zaw T, Cloughesy TF, Naeini KM, Lalezari S, Mong S, Lai A, Ng- enhanced MRI as a prognostic marker for overall survival in recurrent glioblas- hiemphu PL, Pope WB. Comparison between intensity normalization techniques toma: results from the ACRIN 6677/RTOG 0625 Central Reader Study. Neuro for dynamic susceptibility contrast (DSC)-MRI estimates of cerebral blood volume Oncol. 2013;15:945–954. (CBV) in human gliomas. J Magn Reson Imaging. 2012;35:1472–1477. 13. Welker K, Boxerman J, Kalnin A, Kaufmann T, Shiroishi M, Wintermark M; 4. Chaskis C, Stadnik T, Michotte A, Van Rompaey K, D’Haens J. Prognostic value American Society of Functional Neuroradiology MR Perfusion Standards and of perfusion-weighted imaging in brain glioma: a prospective study. Acta Neuro- Practice Subcommittee of the ASFNR Clinical Practice Committee. ASFNR recom- chir (Wien). 2006;148:277–285. mendations for clinical performance of MR dynamic susceptibility contrast perfu- 5. Maia ACM, Malheiros SMF, da Rocha AJ, Stávale JN, Guimarães IF, Borges LR, sion imaging of the brain. AJNR. Am J Neuroradiol. 2015;36:E41–E51. Santos AJ, da Silva CJ, de Melo JG, Lanzoni OP, Gabbai AA, Ferraz FA. Stereo- 14. Semmineh NB, Stokes AM, Bell LC, Boxerman JL, Quarles CC. A population- tactic biopsy guidance in adults with supratentorial nonenhancing gliomas: role based digital reference object (DRO) for optimizing dynamic susceptibility con- of perfusion-weighted magnetic resonance imaging. J Neurosurg. 2004;101: trast (DSC)-MRI methods for clinical trials. Tomography. 2017;3:41–49. 970–976. 15. Semmineh N, Bell L, Stokes A, Hu L, Boxerman J, Quarles C. Optimization of 6. Danchaivijitr N, Waldman AD, Tozer DJ, Benton CE, Brasil Caseiras G, Tofts PS, acquisition and analysis methods for clinical dynamic susceptibility contrast (DSC) Rees JH, Jäger HR. Low-grade gliomas: do changes in rCBV measurements at lon- MRI using a population-based digital reference object. AJNR Am J Neuroradiol. gitudinal perfusion-weighted MR imaging predict malignant transformation? Radi- 2018;39:1981–1988. ology. 2008;247:170–178. 16. Boxerman JL, Schmainda KM, Weisskoff RM. Relative cerebral blood volume 7. Rossi Espagnet MC, Romano A, Mancuso V, Cicone F, Napolitano A, Scaringi maps corrected for contrast agent extravasation significantly correlate with gli- C, Minniti G, Bozzao A. Multiparametric evaluation of low grade gliomas at fol- oma tumor grade, whereas uncorrected maps do not. AJNR Am J Neuroradiol. low-up: comparison between diffusion and perfusion MR with 2006;27:859–867. F-FDOPA PET. Br J Radiol. 2016;89:20160476. 17. Meyer-Baese A, Lange O, Wismueller A, Hurdal MK. Analysis of dynamic sus- 8. Boxerman JL, Paulson ES, Prah MA, Schmainda KM. The effect of pulse sequence ceptibility contrast MRI time series based on unsupervised clustering methods. parameters and contrast agent dose on percentage signal recovery in DSC-MRI: IEEE Trans Inf Technol Biomed. 2007;11:563–573. implications for clinical applications. AJNR Am J Neuroradiol. 2013;34:1364– 18. Schmainda KM, Prah MA, Rand SD, Liu Y, Logan B, Muzi M, Rane SD, Da X, 1369. Yen YF, Kalpathy-Cramer J, Chenevert TL, Hoff B, Ross B, Cao Y, Aryal MP, Erick- 9. Hu LS, Baxter LC, Smith KA, Feuerstein BG, Karis JP, Eschbacher JM, Coons SW, son B, Korfiatis P, Dondlinger T, Bell L, Hu L, Kinahan PE, Quarles CC. Multisite Nakaji P, Yeh RF, Debbins J, Heiserman JE. Relative cerebral blood volume val- concordance of DSC-MRI analysis for brain tumors: results of a National Cancer ues to differentiate high-grade glioma recurrence from posttreatment radiation Institute Quantitative Imaging Network Collaborative Project. AJNR Am J Neuro- effect: direct correlation between image-guided tissue histopathology and local- radiol. 2018;39:1008–1016. ized dynamic susceptibility-weighted contrast-enhanced perfusion MR imaging 19. Kelm ZS, Korfiatis PD, Lingineni RK, et al. Variability and accuracy of different measurements. Am J Neuroradiol. 2009;30:552–558. software packages for dynamic susceptibility contrast magnetic resonance imag- 10. Barajas RF, Chang JS, Segal MR, Parsa AT, McDermott MW, Berger MS, Cha S. ing for distinguishing glioblastoma progression from pseudoprogression. J Med Differentiation of recurrent glioblastoma multiforme from radiation necrosis after imaging (Bellingham). 2015;2:26001. external beam radiation therapy with dynamic susceptibility-weighted contrast- 20. DCE MRI Technical Committee. DCE MRI Quantification Profile, Quantitative Imaging enhanced perfusion MR imaging. Radiology. 2009;253:486–496. Biomarkers Alliance. 2012; Version 1.0. Available from: http://rsna.org/QIBA_.aspx. | | TOMOGRAPHY.ORG VOLUME 5 NUMBER 1 MARCH 2019 117

Journal

Tomography – Multidisciplinary Digital Publishing Institute

Published: Mar 1, 2019

Keywords: DSC-MRI; relative cerebral blood volume; standardization; multisite consistency; reproducibility

There are no references for this article.