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

Learn More →

A systematic review and meta-analysis on the differentiation of glioma grade and mutational status by use of perfusion-based magnetic resonance imaging

A systematic review and meta-analysis on the differentiation of glioma grade and mutational... Background: Molecular characterization plays a crucial role in glioma classification which impacts treatment strategy and patient outcome. Dynamic susceptibility contrast (DSC) and dynamic contrast enhanced (DCE) perfusion imaging have been suggested as methods to help characterize glioma in a non-invasive fashion. This study set out to review and meta-analyze the evidence on the accuracy of DSC and/or DCE perfusion MRI in predicting IDH genotype and 1p/19q integrity status. Methods: After systematic literature search on Medline, EMBASE, Web of Science and the Cochrane Library, a qualita- tive meta-synthesis and quantitative meta-analysis were conducted. Meta-analysis was carried out on aggregated AUC data for different perfusion metrics. Results: Of 680 papers, twelve were included for the qualitative meta-synthesis, totaling 1384 patients. It was observed that CBV, ktrans, Ve and Vp values were, in general, significantly higher in IDH wildtype compared to IDH mutated glioma. Meta-analysis comprising of five papers (totaling 316 patients) showed that the AUC of CBV, ktrans, Ve and Vp were 0.85 (95%-CI 0.75–0.93), 0.81 (95%-CI 0.74–0.89), 0.84 (95%-CI 0.71–0.97) and 0.76 (95%-CI 0.61–0.90), respectively. No conclusive data on the prediction of 1p/19q integrity was available from these studies. Conclusions: Future research should aim to predict 1p/19q integrity based on perfusion MRI data. Additionally, correlations with other clinically relevant outcomes should be further investigated, including patient stratification for treatment and overall survival. Keywords: Dynamic contrast enhancement magnetic resonance perfusion imaging, Dynamic susceptibility contrast magnetic resonance perfusion imaging, Glioma, Molecular classification Key points • Perfusion MR imaging shows a promising method to characterize glioma non-invasively. • Significant higher perfusion metrics are observed in IDH-wildtype glioma. • The effects of 1p/19q mutations on perfusion metrics *Correspondence: [email protected] are understudied and remain unelucidated. Department of Medical Imaging, Radboud University Medical Center, Geert Grooteplein Zuid 10, 6525 EZ Nijmegen, The Netherlands Full list of author information is available at the end of the article © The Author(s) 2022. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/. van Santwijk et al. Insights into Imaging (2022) 13:102 Page 2 of 12 To this end, artificial intelligence applied to con - Introduction ventional MRI sequences (i.e., pre- and post-contrast Following the 2016 World Health Organization (WHO) T1-weighted, T2-weighted and T2-weighted FLAIR classification system of tumors of the central nervous sys - images) to predict mutational status has provided prom- tem, the high-grade glioma group can be divided in two ising results in recent years (for a review, see [8]). In addi- subgroups. One subgroup comprises the anaplastic oli- tion, various signs have been identified which can help godendroglioma IDH mutant and 1p/19q codeleted, and the radiologist to predict the molecular status of glioma the anaplastic oligodendroglioma not otherwise speci- in the daily clinical setting. For example, the T2-FLAIR fied. The second subgroup comprises the IDH mutant mismatch sign has been found to be a reliable non-inva- glioblastoma, the IDH wildtype glioblastoma, and the sive marker for identification of IDH mutant astrocyto - glioblastoma not otherwise specified [1]. Knowledge on mas [9]. the exact mutational status of glioma is not only impor- Bearing in mind the pathophysiological differences tant for classification, it also has significant impact on between various glioma subtypes and the related changes prognosis [2] and treatment strategy [3–5]. With regard in the gliomas vasculature, perfusion-based imaging to low grade gliomas, two groups of gliomas can be dis- could increase the diagnostic accuracy of non-invasive tinguished. The first groups consist of oligodendroglial characterization of glioma subtypes. For example, oligo- tumors which are isocitrate dehydrogenase (IDH) mutant dendroglial tumors are characterized by a branching pat- and 1p/19q codeleted The second groups consist of astro - tern of vascularization, whereas astrocytic glioma shows cytic tumors. It is comprised of (1) IDH mutated, 1p/19q a distinctively different vascularization [10]. Therefore, non-codeleted diffuse astrocytoma, (2) the IDH wildtype perfusion based MR imaging (either dynamic suscepti- astrocytoma, and (3) the diffuse astrocytoma not other - bility contrast (DSC) or dynamic contrast enhancement wise specified [1]. (DCE) perfusion MR imaging) has been the subject of However, the recently published WHO 2021 classi- research to non-invasively identify molecular character- fication system has placed even more emphasis on the istics [11, 12]. molecular characteristics of glioma subtypes. The group DSC-perfusion MR imaging relies on the susceptibility of diffuse astrocytic and oligodendroglial gliomas can be induced signal loss on T2*-weighted sequences, resulting subdivided based on the IDH mutations. IDH wildtype from a bolus of gadolinium-based contrast agent pass- tumors are classified as high-grade gliomas, without ing through the capillaries. The most commonly used exception. In order to be classified as glioblastoma (IDH DSC perfusion parameter is Cerebral Blood Volume wildtype; grade 4), nuclear ATRX loss has to be present. (CBV). CBV can be estimated by use of the area under Additionally, IDH wild-type diffuse astrocytic tumors in the curve (AUC) of the signal intensity-time curve [13, adults without the histological features of glioblastoma, 14]. However, more recent studies compute CBV maps but with one or more of three genetic parameters (tel- by integrating the transverse relaxivity changes which omerase reverse transcriptase gene [TERT] promoter occur dynamically over a first-pass injection followed by mutation, epidermal growth factor receptor [EGFR] gene leakage correction due to the leaky blood–brain barrier amplification, or combined gain of entire chromosome 7 in most tumors (for a recent overview and recommenda- and loss of entire chromosome 10 [+ 7/ − 10]) are now tions, see [15]). DCE-perfusion MR imaging relies on the also classified as glioblastoma. In the 2021 classification, evaluation of T1 shortening induced by a gadolinium- all IDH-mutant diffuse astrocytic tumors with intact based contrast agent bolus leaking from the blood vessels 1p/19q chromosomes are considered a single type called into the tissue. Pharmacokinetic modeling can be used to astrocytoma, IDH-mutant with WHO grades ranging derive various values including, Ve and Vp. ktrans repre- from 2 to 4.  Grading of these tumors takes into account sents the capillary permeability; Ve represents the frac- molecular findings such as the homozygous deletion tional volume of the gadolinium-based contrast agent in of CDKN2A/B, which is associated with a worse prog- the extravascular-extracellular space; Vp represents the nosis. IDH-mutant astrocytomas with these molecular fractional volume of the of the gadolinium-based con- alterations will be classified as WHO grade of 4, even trast agent in the plasma space [13]. if microvascular proliferation or necrosis is absent [6]. Although various studies with different methodologies Additionally, IDH mutant oligodendroglial gliomas with and outcomes have been published since the release of codeleted 1p/19q chromosomes are considered oligoden- the WHO 2016 classification system of glioma, a com - drogliomas. While the establishment of the sophisticated prehensive overview of the accuracy of perfusion based molecular markers to classify gliomas is an important MR imaging to predict the molecular characteristics of advance in glioma diagnosis, all of the literature which glioma is still lacking. In addition, a systematic overview is covered within this review is based on the 2016 WHO of the literature on this topic could help to shape future classification of central nervous system tumors [6, 7]. v an Santwijk et al. Insights into Imaging (2022) 13:102 Page 3 of 12 research and daily clinical practice to focus on the most were cross-checked afterward, and discrepancies were promising technique (either DSC- or DCE-perfusion resolved in consensus. MRI). The aim of the current paper was therefore to pro - vide an overview of the relevant literature with regard to Qualitative meta‑synthesis and quantitative meta‑analysis the use of DSC and DCE perfusion imaging used to dif- Eligible literature was synthesized qualitatively follow- ferentiate glioma grade and mutational status. ing the PICO-strategy as proposed by Eriksen et  al. [17]. Also, quality of primary diagnostic accuracy stud- Materials and methods ies was assessed using the QUADAS-2. Meta-analysis Search strategy and inclusion/exclusion methodology was conducted on the AUC and the 95% Confidence This systematic review and meta-analysis was conducted Interval (95%-CI) using a random effects model. From following the Preferred Reporting Items for System- the included studies, perfusion metrics and the afore- atic Reviews and Meta-Analyses (PRISMA) statement mentioned statistics were extracted. If one of these vari- [16]. Databases searched for literature were: Medline ables was missing, the researchers aimed to re-calculate (accessed through PubMed), EMBASE, Web of Science, the value when possible [18]. In addition, corresponding and the Cochrane Library. The full search strategies for authors were contacted to provide missing details, with each database are made available in the Additional file  1. up to two reminders send by e-mail. When not all neces- Cross-referencing was used to add relevant literature to sary data could be acquired, studies could not be meta- the database. Searches were conducted between May 1, analyzed. Meta-analyses were conducted on different 2020 and January 1, 2021. Inclusion criteria were: (1) the subgroups of target conditions. Meta-analysis was per- use of either DSC or DCE perfusion MRI; (2) the inclu- formed with the use of OpenMetaAnalyst (MetaAnalyst, sion of patients suffering from glioma; (3) glioma grading Tufts Medical Center) [19] and/or SPSS (version 25; IBM and classification by use of the WHO 2016 classification Corp., Armonk, NY) and results were displayed in forest system [1]; and (4) the aim of the study needed to com- plots. The Higgins test was used to test for heterogene - prise the non-invasive classification of histopathological ity between included studies. Low heterogeneity between features and/or molecular characteristics (WHO grade, groups is marked with an I < 40%, whereas considerable IDH genotype and/or 1p/19q codeletion status). Besides, heterogeneity is indicated by I > 75% [18]. papers needed to report results as quantitative measures (e.g., sensitivity, specificity, mean accuracy and/or mean Results area under the receiver operator curve (AUC)). Papers A total of 552 studies were identified after systematic were excluded if they were based on animals or non- searching. Duplicates were removed and 379 papers were human samples or a pediatric population. Letters, pre- systematically screened on title and abstract resulting in prints, case reports, congress proceedings, and narrative the inclusion of 34 papers for full-text analysis. Reasons reviews were excluded as well. for exclusion of the 345 papers are provided in Fig.  1. All papers were independently assessed by two After full-text analysis, the investigators met to discuss researchers in three steps. First, screening on title and the identified non-consensus papers to resolve disagree - abstract was carried out. Second, full-text analysis was ments and to reach consensus. Of the 34 papers, 12 could employed to assess whether the papers met the inclu- be included in the qualitative meta-synthesis. Twenty- sion- and/or exclusion criteria. Finally, information was two papers were therefore excluded (details provided in extracted from the included papers. Researchers met Fig.  1). No discrepancies between the judgement of the periodically to discuss their findings and resolve dis - two researchers remained after discussion, resulting in crepancies. Standardized tables were used to acquire the the final inclusion of 12 papers for the qualitative meta- information of interest from the included articles by two synthesis [20–31] (Fig. 1). Five papers provided sufficient researchers (LvS and DH) independently. Data extracted data to be included in the quantitative meta-analysis [22, from each study were (a) first author and year of publi - 24, 26, 27, 29] (Table 1). cation, (b) number of patients included, (c) mean age Using the QUADAS2 (QUality Assessment tool for of the included participants, (d) gender of the included Diagnostic Accuracy Studies), the most current version participants, (e) use of DSC and/or DCE, (f ) which his- of the QUADAS tool of the QUADAS task force, the topathological/molecular outcome was assessed, (g) risk of bias was considered low in all included studies perfusion based MR imaging metrics and (h) accompa- (Table 2). nying statistics (e.g., AUC value, standard deviation, 95% confidence interval (CI) and/or standard error). Perfor - Qualitative meta‑synthesis mance was expressed in accuracy, AUC and/or sensi- The twelve included studies [20–31] totaled 1384 patients tivity and specificity for each outcome. Extracted data (792 males; 592 females) suffering from glioma. Gliomas van Santwijk et al. Insights into Imaging (2022) 13:102 Page 4 of 12 Fig. 1 PRISMA flow diagram v an Santwijk et al. Insights into Imaging (2022) 13:102 Page 5 of 12 Table 1 Overview of the included studies Authors (year) N Age (years) M/F MRI perfusion Glioma types and Outcome assessed Major findings method + details on analysis grades included Brendle et al. (2020) [30] 56 Mean age 48.0 ± 16.0 33/23 DSC WHO grade II: 29 IDH mutation status & 1p/19q The mean rCBV was significantly Gradient echo sequence WHO grade III: 20 codeletion status different between the astrocytic Pre-bolus of contrast agent WHO grade IV: 7 tumors, oligodendrogliomas and was applied (0.025 mmol/kg IDHmut: 32 IDHmut astrocytic tumors and gadobutrol) IDHwt: 24 oligodendrogliomas and IDHwt Mean rCBV values astrocytic tumors BSW-model based leakage correction Choi et al. (2019) [20] 463 Mean age 52.2 ± 14.8 272/191 DSC WHO grade II: 32 IDH mutation status The IDH mutation status predic- Gradient echo sequence WHO grade III: 142 tions had an accuracy, sensitivity, Pre-bolus of 0.1 mmol/kg WHO grade IV: 289 and specificity of 92.8%, 92.6%, gadobutrol IDHmut: 328 and 93.1%, respectively, in the Mean rCBV values IDHwt: 125 validation set with an AUC of 0.9 No information on post- 1p/19q codel: 56 (95%-CI 0.969–0.991). In the test processing with regard to 1p/19q non-codel: 407 set, the IDH genotype prediction leakage-correction had an accuracy, sensitivity, and specificity of 91.7%, 92.1%, and 91.5%, respectively, with an AUC of 0.95 (95%-CI 0.898–0.982) Hempel et al. (2019) [21] 100 Mean age 51.4 ± 14.7 55/45 DSC WHO grade II: 40 IDH mutation status rCBV was significantly lower in Gradient echo sequence WHO grade III: 30 patients with IDHmut than in Pre-bolus of 0.1 mmol/kg WHO grade IV: 30 those with the IDHwt. Mean gadobutrol IDHmut: 31 rCBV values showed a sensitivity/ Mean rCBV values* IDHwt: 46 specificity of 52/91 for the pre - BSW-model based leakage 1p/19q codel: 23 diction of IDH mutation status correction with an AUC of 0.780 Hilario et al. (2019) [22] 49 Range 28/21 DSC LGG: 8 IDH mutation status Significant differences in the val- (X) 16–78 Gradient echo sequence HGG: 41 ues of leakage (p = 0.01), Ktrans Pre-bolus of 0.1 mmol/kg IDHmut: 10 (p = 0.002), Vp (p = 0.032) and Ve gadobutrol IDHwt: 31 (p < 0.001) between high-grade No rCBV values provided and low-grade diffuse gliomas BSW-model based leakage were observed correction The highest AUC was demon- strated by the DCE permeability parameters Ktrans (AUC = 0.838, CI95% 0.710–0.967, p = 0.003) and Ve (AUC = 0.878, CI95% 0.768–0.988, p = 0.001). Among IDHmut and IDHwt highgrade gliomas, there were significant differences in leakage (p = 0.004) and Ktrans values (p = 0.028) showing lower leakage and Ktrans values van Santwijk et al. Insights into Imaging (2022) 13:102 Page 6 of 12 Table 1 (continued) Authors (year) N Age (years) M/F MRI perfusion Glioma types and Outcome assessed Major findings method + details on analysis grades included DCE Dynamic gradient echo sequence 5 mL of gadobutrol at a rate of 3 mL/s Mean Ktrans, Vp, Ve and Kep values Model based leakage correc- tion Lee et al. (2018) [23] 39 Mean age 43.6 (range 21–82) 19/20 DSC WHO grade II: 19 WHO grade Ktrans, Kep, and Ve showed Gradient echo sequence WHO grade III: 20 tendencies toward higher values No prebolus administration in oligodendroglial tumors than described astrocytic tumors Mean rCBV values* BSW-model based leakage correction DCE Dynamic gradient echo sequence 0.1 mmol/kg gadobutrol at a rate of 4 mL/s Mean Ktrans Model based leakage correc- tion Lee et al. (2020) [24] 110 IDHmut.- 1p/19q noncodel: 56/54 DSC WHO grade II: 45 IDH mutation status & 1p/19q When using nCBV skewness, (X) mean age 40.7 ± 12.8; Gradient echo sequence WHO grade III: 65 codeletion status the AUC was found to be 0.690 IDHwt: mean age 51.2 ± 14.0; Pre-bolus of 0.1 mmol/kg IDHmut: 65 (95%-CI: 0.573, 0.807) with a IDHmut-1p/19q codel: mean gadoterate meglumine IDHwt: 45 sensitivity of 84.2 and specificity age 46.5 ± 11.7 Mean normalized rCBV values* 1p/19q codel: 46 of 59.3 to distinguish IDHmut- BSW-model based leakage 1p/19q non-codel: 19 1p/19q noncodel from the other correction two groups v an Santwijk et al. Insights into Imaging (2022) 13:102 Page 7 of 12 Table 1 (continued) Authors (year) N Age (years) M/F MRI perfusion Glioma types and Outcome assessed Major findings method + details on analysis grades included Sudre et al. (2020) [25] 333 Mean age 48.9 (range 20–81) 198/135 DSC WHO grade II: 101 WHO grade & IDH mutation Shape, distribution and texture No details provided on imag- WHO grade III: 74 status features showed significant dif- ing protocol and whether or WHO grade IV: 158 ferences across mutation status. not a prebolus was adminis- IDHmut: 151 WHO grade II-III differentiation tered IDHwt: 182 was mostly driven by shape fea- Mean rCBV values* tures, while texture and intensity BSW-model based leakage feature were more relevant for correction the III-IV separation. Increased number of features became significant when differentiating grades further apart from one another. Gliomas were correctly stratified by mutation status in 71% and by grade in 53% of the cases Wang et al. (2020) [31] 30 IDHmut: mean age 42.8 (range 17/13 DCE WHO grade II: 22 IDH mutation status Compared to IDHmut LGGs, (X) 22–67) Dynamic gradient echo WHO grade III: 8 IDHwt LGGs exhibited signifi- IDHwt: sequence IDHmut: 18 cantly higher perfusion metrics mean age 47.9 (range 19–78) Pre-bolus of 0.1 mmol/kg IDHwt: 12 (p < 0.05) gadopentetate dimeglumine at a rate of 4 mL/s Mean Ktrans, Vp, Ve Model based leakage correc- tion Wu et al. (2020) [26] 44 63.8 ± 7.4 27/17 DSC WHO grade III: 19 IDH mutation status Compared with IDHwt, IDHmut Gradient echo sequence WHO grade IV: 25 had significantly decreased No pre-bolus administration** IDHmut: 19 rCBV at the high-angiogenic Mean rCBV values IDHwt: 25 enhancing tumor habitats and Gamma-variate curve fitting 1p/19q codel: 7 low-angiogenic enhancing leakage correction 1p/19q non-codel: 3 tumor habitats Xing et al. (2017) [27] 42 IDHmut: mean age 35.8 ± 9.1 26/16 DSC WHO grade II: 24 IDH mutation status The threshold value of < 2.35 (X) IDHwt: mean age Gradient echo sequence WHO grade III: 18 for relative maximum CBV in 46.0 ± 18.4 Pre-bolus of 0.1 mmol/kg IDHmut: 17 the prediction of IDH mutation gadobenate dimeglumine IDHwt: 25 status provided a sensitivity, Mean rCBVmax values* specificity, positive predictive BSW-model based leakage value, and negative predictive correction value of 100.0%, 60.9%, 85.6%, and 100.0%, respectively van Santwijk et al. Insights into Imaging (2022) 13:102 Page 8 of 12 Table 1 (continued) Authors (year) N Age (years) M/F MRI perfusion Glioma types and Outcome assessed Major findings method + details on analysis grades included Xing et al. (2019) [28] 75 IDHmut: mean age 52.2 ± 12.7 41/34 DSC WHO grade IV: 75 IDH mutation status Both rCBVmax-t and rCBVmax-p IDHwt: mean age 40.7 ± 10.8 Gradient echo sequence IDHmut: 10 showed significant differences Pre-bolus of 0.1 mmol/kg IDHwt: 65 between IDHmut and IDHwt. gadobenate dimeglumine The optimal cutoff values in Mean rCBVmax values* prediction of IDH-m. < 7.27 for BSW-model based leakage rCBVmax-tumor, and < 0.97 for correction rCBVmax-peri-enhancing region Zhang et al. (2020) [29] 43 47.0 ± 13.0 20/23 DSC WHO grade II: 14 IDH mutation status Ve (AUC = 0.816, sensitiv- (X) Gradient echo sequence WHO grade III: 14 ity = 0.84, specificity = 0.79) and (X) DSC imaging followed DCE WHO grade IV: 15 Kep (AUC = 0.818, sensitiv- imaging; no separate pre-bolus IDHmut: 20 ity = 0.76, specificity = 0.78) was administered IDHwt: 23 provided the highest differential Mean rCBVmax values* efficiency for IDH mutation BSW-model based leakage status prediction correction DCE Dynamic gradient echo sequence Pre-bolus of 0.1 mmol/kg gadodiamide Mean Ktrans, Vp, Ve* Model based leakage correc- tion Marked in italics are the publications included in the meta-analysis on the use of DSC-value; Marked in bold are the publications included in the meta-analysis on the use of the DCE-values AUC, area under the curve; DCE, dynamic contrast enhancement magnetic resonance perfusion imaging; DSC, dynamic susceptibility contrast magnetic resonance perfusion imaging; F, females; HGG, high-grade glioma; IDH, isocitrate dehydrogenase; IDHmut, mutation of the isocitrate dehydrogenase gene(s); IDHwt, wild-type isocitrate dehydrogenase gene(s); Kep, rate constant between the extravascular extracellular space and blood plasma; ktrans, volume transfer coefficient; LGG, low grade glioma; M, males; MRI, magnetic resonance imaging; nCBV, normalized cerebral blood volume; rCBV, relative cerebral blood volume; rCBVmax-t, maximum relative cerebral blood volume in the tumor-enhancing region; rCBVmax-p, maximum relative cerebral blood volume in the peri-enhancing region; Ve, fractional volume of the extravascular extracellular space; Vp, fractional blood plasma volume; WHO, World Health Organization; 95%-CI, 95%-confidence interval *Study provides a variety of perfusion statistics (either DSC or DCE metrics; values included mean, standard deviation and a variety of percentiles) **Lack of pre-bolus administration was compensated by use of a flip-angle of 60° which reduced T1 effects [44] v an Santwijk et al. Insights into Imaging (2022) 13:102 Page 9 of 12 Table 2 Combined effect size for the different DCE/DSC When using DCE perfusion imaging, IDHmut HGG parameters (either WHO 2021 Astroctytoma grade 3 or 4 or WHO 2021 Oligodendroglioma grade 3) showed significantly Ktrans Ve Vp CBV lower ktrans values as compared to IDHwt HGG (WHO Eec ff t size 0.813 0.844 0.777 0.832 2021 Astrocytoma grade 4) [22]. In oligodendroglial Standard error 0.02 0.03 0.03 0.03 tumors (WHO 2021 Oligodendroglioma, IDHmut, 95%-CI lower limit 0.726 0.766 0.683 0.749 1p/19q-codeleted; Grade 2 or 3), however, Lee et  al. 95%-CI upper limit 0.900 0.921 0.871 0.914 found that ktrans, Kep and Ve showed tendencies toward ktrans, volume transfer coefficient; rCBV, relative cerebral blood volume; Ve, higher values as compared to astrocytic tumors [23]. Ve fractional volume of the extravascular extracellular space; Vp, fractional blood and Vp values were found to be significantly lower in plasma volume; 95%-CI, 95%-confidence interval IDHmut glioma (WHO 2021 Astrocytoma and WHO 2021 Oligodendroglioma) as compared to IDHwt glioma, regardless of WHO II-IV grading [22, 29]. Based on Ve could be subdivided into WHO grade II (n = 326); WHO and Kep values, a sensitivity/specificity of 84%/79% and grade III (n = 410) and WHO grade IV (n = 599). Regard- 76%/78% was observed with regard to differentiate IDH ing the IDH genotype, 701 gliomas were IDH-mutated mutation status [29]. The study of Hilario et al. also sug - and 603 tumors expressed an IDH wildtype genotype. gested that ktrans, Vp and Ve could be used to differenti - 1p/19q codeletion (WHO 2021 Oligodendroglioma ate between LGG and HGG non-invasively [22]. WHO grade 2 or 3) was observed in 132 tumors; non- Studies using artificial intelligence showed promising codeletion of 1p/19q chromosome arms was observed in results with regard to prediction of IDH mutation sta- 429 tumors. All included papers used histopathological/ tus. Choi et  al. showed that a convolutional long short- molecular assessment by a trained neuropathologist who term memory model with an attention mechanism had adhered to the WHO 2016 glioma classification as the an accuracy, sensitivity, and specificity of 92.8%, 92.6%, gold standard. and 93.1%, respectively, in the validation set (AUC: 0.98; Eight papers used DSC perfusion MRI [20, 21, 24–28, 95%-CI 0.969–0.991) with regard to IDH genotype pre- 30]; three papers used both DCE and DSC perfusion MRI diction by use of DSC perfusion MRI. In the test set, an [22, 23, 29]; one paper used DCE perfusion MRI only accuracy, sensitivity, and specificity of 91.7%, 92.1%, and [26]. Two papers used artificial intelligence methods to 91.5% were observed, respectively. The AUC value of the assess different perfusion metrics between various sub - IDH genotype prediction demonstrated to be 0.95 with types of gliomas [20, 25], whereas the other publications a 95% CI ranging between 0.898 and 0.982. Subsequent used more traditional statistics. analysis of the signal intensity curves of DSC imaging As assessed by DSC perfusion MRI, IDHmut glioma elucidated high attention on the combination of the end displayed significantly lower rCBV values as compared to of the pre-contrast baseline, the up/downslopes of signal IDHwt glioma [21, 26–28, 30]. When using a retrospec- drops, and/or post-bolus plateaus for the curves used to tively determined rCBVmax threshold value of < 2.35, the predict IDH genotype [20]. Another study showed that authors described a sensitivity/specificity of 100%/61% when using a random forest algorithm, shape, distribu- and AUC of 0.82 (95%-CI: 0.66–0.93) when differentiat - tion and rCBV-extracted features elucidated significant ing IDHmut (either WHO 2021 Astroctytoma grade 2, differences across mutation status. WHO grade II-III dif - 3 or 4 or WHO 2021 Oligodendroglioma grade 2 or 3) ferentiation was mostly driven by shape features, while and IDHwt gliomas [27]. By use of the skewness of nor- texture and intensity feature were more relevant for the malized CBV (nCBV) values (normalized by use of the distinguishing of III and IV. Based on this random forest CBV value of the normal-appearing contralateral cen- algorithm, gliomas were correctly stratified by mutation trum semiovale), IDHmut, 1p/19q non-codeleted glioma status in 71% and by grade in 53% of the cases [25]. (WHO 2021 Astrocytoma grade 2, 3 or 4) could be dis- tinguished from IDHwt glioma and IDHmut, 1p/19q Meta‑analysis codeleted glioma (WHO 2021 Oligodendroglioma) with Meta-analysis of the data (n = 237 patients) showed that a sensitivity/specificity of 84%/59% (AUC-value of 0.690 CBV values have an accuracy of correctly predicting IDH and 95%-CI 0.573–0.807). IDHmut, 1p/19q non-code- genotype with an AUC of 0.832 and a standard error of leted gliomas (WHO 2021 Astrocytoma grade 2, 3 or 4) 0.03 (95%-CI 0.75–0.91). When using DCE parameters showed significant lower nCBV values compared to the (n = 122), an AUC of 0.81, 0.84 and 0.78 is observed for IDHmut, 1p/19q codeleted gliomas (WHO 2021 Oligo- ktrans, Ve and Vp, respectively. Standard errors (and dendroglioma grade 2 or 3) and the IDH wildtype glio- 95%-CI) for ktrans, Ve and Vp were found to be 0.02 mas [24]. van Santwijk et al. Insights into Imaging (2022) 13:102 Page 10 of 12 (95%-CI 0.73–0.90), 0.03 (95%-CI 0.77–0.92) and 0.03 angiopoietin 2 in WHO grade II glioma was confirmed by (95%-CI 0.68–0.87), respectively (Table  2). The corre - staining of human tumor tissue microarrays. More spe- sponding Forest-Plots of the different perfusion metrics cifically, IDHwt gliomas displayed a specific angiogenic are provided in Fig.  2. I analysis showed that included gene expression signature (i.e., upregulation of Angiopoi- DCE-MRI studies were homogeneous (I < 1%). In the etin 2 and serpin family H) which resulted in enhanced individual analyses of ktrans, Ve and Vp, studies were endothelial cell migration and matrix remodeling. In found to be non-significantly heterogeneous (p = 0.834; the same study, transcription factor analysis indicated p = 0.548; p = 0.519, respectively). The meta-analysis of increased transforming growth factor beta and hypoxia DSC-MRI studies showed to have moderate heterogene- signaling in IDHwt gliomas [39]. Based on these studies, ity (I = 35%; p = 0.215). The role of perfusion MRI met - we can conclude that gliomas with different IDH geno - rics in predicting the 1p/19q-codeletion status could not type have distinct molecular vascularization. In addition, be meta-analyzed using the acquired data. the blood vessels in LGG displayed alterations in gene expression which partially overlapped with changes pre- Discussion viously identified in HGG vessels [39]. As IDHwt glioma This systematic review and meta-analysis shows that vessels have a distinct vascular gene expression pattern perfusion MRI can be used to effectively predict IDH associated with vascular remodeling, these microstruc- genotype non-invasively following the WHO 2016/2021 tural changes can be used to explain why IDHwt glioma glioma classification. Different DSC or DCE perfusion show significantly higher perfusion metrics compared parameters were found to have an equal performance to IDHmut glioma. These insights in genotype and phe - regarding the non-invasive prediction of IDH genotype. notype justify the use of perfusion MRI to predict IDH Prediction of the 1p/19q-codeletion status could not be genotype. The role of 1p/19q codeletion status on angio - meta-analyzed using the acquired data. genesis and vascular growth, however, remains partially The role of perfusion MRI in non-invasive glioma clas - elusive. Previous research demonstrated that 1p/19q sification can be significant and can be explained by the codeletion was associated with higher CBV values com- different glioma vasculature fingerprints which provide pared with glioma with intact alleles [40]. Another paper a specialized microenvironment for glioma cells [32]. reported specific genotypic differences in oligodendro - Within HGG, blood vessels are abnormal and display a glioma by use of DSC perfusion MRI with significantly distinct gene expression signature which differs from the higher rCBVmax values in LGG with 1p/19q codeletion genotype of blood vessels in normal brain tissue [33–35]. [41]. It is believed that 1p/19q codeleted LGG show an These genotypic differences result in high expression of increased metabolism and angiogenesis and have an certain angiogenic factors, including vascular endothelial extensive internal vascular network. This is supported by growth factor, transforming growth factor β2, and pleio- the study of Kapoor et al. in which a significantly higher trophin [35–38]. In LGG, on the other hand, potential rCBVmax was observed in 1p/19q codeleted LGG (WHO molecular alterations regarding angiogenesis have been 2021 Oligodendroglioma WHO grade 2). Additionally, an investigated less extensively. In 2018, Zhang et  al. dem- increased vascular endothelial growth factor expression, onstrated that WHO grade II glioma expressed an inter- CD31, and CD105, was observed as compared with gli- mediate stage of vascular abnormality, less severe than oma with intact alleles [42]. that of glioblastoma vessels but distinct from normal ves- The clinical usability of MRI perfusion imaging to sels. Enhanced expression of laminin subunit alpha 4 and predict IDH genotype remains partially elusive as the Fig. 2 Forest-plot of the area under the curve (AUC) of the receiver operator curve (ROC) of the different perfusion metrics in predicting IDH mutation status. IDH, isocitrate dehydrogenase, ktrans, volume transfer coefficient; rCBV, relative cerebral blood volume; Ve, fractional volume of the extravascular extracellular space; Vp, fractional blood plasma volume; 95%-CI, 95%-confidence interval v an Santwijk et al. Insights into Imaging (2022) 13:102 Page 11 of 12 differences were based on aggregated results. Although Supplementary Information several papers provided specific threshold values [27, The online version contains supplementary material available at https:// doi. org/ 10. 1186/ s13244- 022- 01230-7. 28], no clinically useful threshold values are available. The current review did not include studies arterial spin Additional file 1: Search strategies. labeling (ASL) as a perfusion MR imaging method as only sparse literature with regard to ASL was found in explor- Author contributions atory literature searches. A recent paper by Wang et  al. L.vS. and V.K. systematically searched and reviewed the available literature (2019) reported that only a mild correlation was found under direct supervision of D.H. D.H. helped to solve disagreements when- ever needed. L.vS. and V.K., under direct supervision of D.H., carried out the between the IDH1 genotypes and ASL derived glioma meta-analysis. L.vS. and V.K. wrote the first draft of the manuscript, which was perfusion parameters. There was no significant associa - reviewed and corrected by D.H., F.M. and M.S. D.H. wrote subsequent versions tion between 1p/19q codeletion and perfusion in grade II in close collaboration with L.vS. and V.K. These versions were corrected by F.M. and M.S. All authors read and approved the final manuscript. and III gliomas [43]. Funding Strength and limitations No funding was received for conducting this research. By adhering to the 2016 WHO glioma classification to be Availability of data and materials included, some valuable papers needed to be excluded, Data analyzed in this study will be made available upon reasonable request by though also resulted in rather homogeneous dataset to be contacting the corresponding author. meta-analyzed. One of the strengths of this reviews con- cerns the relative homogeneous imaging protocols which Declarations were meta-analyzed. For example, all DSC-imaging pro- Ethics approval and consent to participate tocols were imaged after administering a pre-bolus injec- No ethical approval was necessary to conduct this review. tion of a gadolinium-based contrast agent. Also, for the Consent for publication included studies which investigated the diagnostic accuracy All authors consent publication of this manuscript. of DCE-imaging, mean perfusion values (i.e., Ve, Vp and ktrans). However, different studies used different values Competing interests The authors declare no competing interests. of perfusion parameters (mean rCBV vs. mean rCBV max values), which partially limits the generalizability of results Author details [15]. The homogeneity of the meta-analyzed patients and Department of Medical Imaging, Radboud University Medical Center, Geert Grooteplein Zuid 10, 6525 EZ Nijmegen, The Netherlands. Depar tment histopathological outcomes (i.e., IDH genotype and 1p/19q of Radiology and Nuclear Medicine, Erasmus MC, University Medical Center codeletion status) strengthen the here described find Rotterdam, Rotterdam, The Netherlands. ings. Another limitation of the here applied methodology Received: 14 March 2022 Accepted: 20 April 2022 concerns the fact that this systematic review was executed without registration in an international database. Conclusion References This review and meta-analysis showed that accuracy of 1. Louis DN, Perry A, Reifenberger G et al (2016) The 2016 World Health DSC parameters was not different from the accuracy of Organization classification of tumors of the central nervous system: a summary. Acta Neuropathol 131(6):803–820 DCE parameters to non-invasive predict the IDH geno- 2. Lapointe S, Perry A, Butowski NA (2018) Primary brain tumours in adults. type in glioma patients. The use of perfusion MRI with Lancet 392(10145):432–446 regard to predicting 1p/19q codeletion status could not 3. Narang J, Jain R, Scarpace L et al (2011) Tumor vascular leakiness and blood volume estimates in oligodendrogliomas using perfusion CT: an be determined using these data. analysis of perfusion parameters helping further characterize genetic subtypes as well as differentiate from astroglial tumors. J Neurooncol 102(2):287–293 Abbreviations 4. Saito T, Yamasaki F, Kajiwara Y et al (2012) Role of perfusion-weighted 95%-CI: 95%-Confidence interval; AUC : Area under the curve; DCE MRI: imaging at 3 T in the histopathological differentiation between astrocytic Dynamic contrast enhancement magnetic resonance perfusion imaging; DSC and oligodendroglial tumors. Eur J Radiol 81(8):1863–1869 MRI: Dynamic susceptibility contrast magnetic resonance perfusion imaging; 5. Lecavalier-Barsoum M, Quon H, Abdulkarim (2014) Adjuvant treatment HGG: High-grade glioma; IDH: Isocitrate dehydrogenase; IDHmut: Mutation of of anaplastic oligodendrogliomas and oligoastrocytomas. Cochrane the isocitrate dehydrogenase gene(s); IDHwt: Wild-type isocitrate dehydroge- Database Syst Rev (5) nase gene(s); Kep: Rate constant of gadolinium efflux between the extravas- 6. Louis DN, Perry A, Wesseling P et al (2021) The 2021 WHO classifica- cular extracellular space and blood plasma; ktrans: Volume transfer constant; tion of tumors of the central nervous system: a summary. Neuro Oncol LGG: Low grade glioma; MRI: Magnetic resonance imaging; nCBV: Normalized 23(8):1231–1251 cerebral blood volume; rCBV: Relative cerebral blood volume; rCBVmax-p: 7. Weller M, van den Bent M, Preusser M et al (2021) EANO guidelines on Maximum relative cerebral blood volume in the peri-enhancing region; the diagnosis and treatment of diffuse gliomas of adulthood. Nat Rev Clin rCBVmax-t: Maximum relative cerebral blood volume in the tumor-enhancing Oncol 18(3):170–186 region; Ve: Fractional volume of the extravascular extracellular space; Vp: Fractional blood plasma volume; WHO: World Health Organization. van Santwijk et al. Insights into Imaging (2022) 13:102 Page 12 of 12 8. van Kempen E, Post M, Mannil M et al (2021) Accuracy of machine learn- with the prognosis of different molecular glioma subtypes. Neurol Sci ing algorithms for the classification of molecular features of gliomas on 41(12):3625–3632 MRI: a systematic literature review and meta-analysis. Cancers 13:2606 31. Wang X, Cao M, Chen H, Ge J, Suo S, Zhou Y (2020) Simplified perfusion 9. Jain R, Johnson DR, Patel SH et al (2020) “Real world” use of a highly reli- fraction from diffusion-weighted imaging in preoperative prediction of able imaging sign: “T2-FLAIR mismatch” for identification of IDH mutant IDH1 mutation in WHO grade II–III gliomas: comparison with dynamic astrocytomas. Neuro Oncol 22(7):936–943 contrast-enhanced and intravoxel incoherent motion MRI. Radiol Oncol 10. Zhou J, Li N, Yang G, Zhu Y (2011) Vascular patterns of brain tumors. Int J 54(3):301–310. https:// doi. org/ 10. 2478/ raon- 2020- 0037. Surg Pathol 19(6):709–717 32. Takano S (2012) Glioblastoma angiogenesis: VEGF resistance solutions 11. Brendle C, Hempel J-M, Schittenhelm J et al (2018) Glioma grading and and new strategies based on molecular mechanisms of tumor vessel determination of IDH mutation status and ATRX loss by DCE and ASL formation. Brain Tumor Pathol 29(2):73–86 perfusion. Clin Neuroradiol 28(3):421–428 33. Madden SL, Cook BP, Nacht M et al (2004) Vascular gene expression in 12. Ma H, Wang Z, Xu K et al (2017) Three-dimensional arterial spin labeling nonneoplastic and malignant brain. Am J Pathol 165(2):601–608 imaging and dynamic susceptibility contrast perfusion-weighted imag- 34. Pen A, Moreno MJ, Martin J, Stanimirovic DB (2007) Molecular markers of ing value in diagnosing glioma grade prior to surgery. Exp Ther Med extracellular matrix remodeling in glioblastoma vessels: microarray study 13(6):2691–2698 of laser-captured glioblastoma vessels. Glia 55(6):559–572 13. Essig M, Shiroishi MS, Nguyen TB et al (2013) Perfusion MRI: the five most 35. Dieterich LC, Mellberg S, Langenkamp E et al (2012) Transcriptional profil- frequently asked technical questions. AJR Am J Roentgenol 200(1):24–34 ing of human glioblastoma vessels indicates a key role of VEGF-A and TGF 14. Petrella JR, Provenzale JM (2000) MR perfusion imaging of the brain: beta 2 in vascular abnormalization. J Pathol 228(3):378–390 techniques and applications. AJR Am J Roentgenol 175(1):207–219 36. Zhang L, Dimberg A (2016) Pleiotrophin is a driver of vascular abnormali- 15. Boxerman JL, Quarles CC, Hu LS et al (2020) Consensus recommendations zation in glioblastoma. Mol Cell Oncol 3(6):e1141087 for a dynamic susceptibility contrast MRI protocol for use in high-grade 37. Zhang L, Laaniste L, Jiang Y, Alafuzoff I, Uhrbom L, Dimberg A (2016) gliomas. Neuro Oncol 22(9):1262–1275 Pleiotrophin enhances PDGFB-induced gliomagenesis through increased 16. Moher D, Liberati A, Tetzlaff J, Altman DG (2010) Preferred reporting items proliferation of neural progenitor cells. Oncotarget 7(49):80382–80390 for systematic reviews and meta-analyses: the PRISMA statement. Int J 38. Zhang L, Kundu S, Feenstra T et al (2015) Pleiotrophin promotes vascular Surg 8(5):336–341 abnormalization in gliomas and correlates with poor survival in patients 17. Eriksen MB, Frandsen TF (2018) The impact of patient, intervention, with astrocytomas. Sci Signal 8(406):ra125 comparison, outcome (PICO) as a search strategy tool on literature search 39. Zhang L, He L, Lugano R et al (2018) IDH mutation status is associated quality: a systematic review. J Med Libr Assoc 106(4):420–431 with distinct vascular gene expression signatures in lower-grade gliomas. 18. Higgins JPT, G.S.e (2011) Cochrane handbook for systematic reviews Neuro Oncol 20(11):1505–1516 of interventions version 5.1.0 [updated March 2011]. The Cochrane 40. Chawla S, Krejza J, Vossough A et al (2013) Differentiation between Collaboration oligodendroglioma genotypes using dynamic susceptibility contrast 19. Wallace BC, Lajeunesse MJ, Dietz G et al (2017) OpenMEE: intuitive, open- perfusion-weighted imaging and proton MR spectroscopy. AJNR Am J source software for meta-analysis in ecology and evolutionary biology. Neuroradiol 34(8):1542–1549 Methods Ecol Evol 8(8):941–947 41. Whitmore RG, Krejza J, Kapoor GS et al (2007) Prediction of oligoden- 20. Choi KS, Choi SH, Jeong B (2019) Prediction of IDH genotype in gliomas droglial tumor subtype and grade using perfusion weighted magnetic with dynamic susceptibility contrast perfusion MR imaging using an resonance imaging. J Neurosurg 107(3):600–609 explainable recurrent neural network. Neuro Oncol 21(9):1197–1209 42. Kapoor GS, Gocke TA, Chawla S et al (2009) Magnetic resonance perfu- 21. Hempel J-M, Schittenhelm J, Klose U et al (2019) In vivo molecular profil- sion-weighted imaging defines angiogenic subtypes of oligodendro - ing of human glioma cross-sectional observational study using dynamic glioma according to 1p19q and EGFR status. J Neurooncol 92(3):373–386 susceptibility contrast magnetic resonance perfusion imaging. Clin 43. Wang N, Xie S, Liu H, Chen G, Zhang W (2019) Arterial spin labeling Neuroradiol 29(3):479–491 for glioma grade discrimination: correlations with IDH1 genotype and 22. Hilario A, Hernandez-Lain A, Sepulveda JM et al (2019) Perfusion MRI 1p/19q status. Transl Oncol 12(5):749–756 grading diffuse gliomas: impact of permeability parameters on molecular 44. Welker K, Boxerman J, Kalnin A, Kaufmann T, Shiroishi M, Wintermark M biomarkers and survival. Neurocirugia 30(1):11–18 (2015) ASFNR recommendations for clinical performance of MR dynamic 23. Lee JY, Ahn KJ, Lee YS et al (2018) Differentiation of grade II and III oligo - susceptibility contrast perfusion imaging of the brain. AJNR Am J Neuro- dendrogliomas from grade II and III astrocytomas: a histogram analysis radiol 36(6):E41-51 of perfusion parameters derived from dynamic contrast-enhanced (DCE) and dynamic susceptibility contrast (DSC) MRI. Acta Radiol 59(6):723–731 Publisher’s Note 24. Lee MK, Park JE, Jo Y et al (2020) Advanced imaging parameters improve Springer Nature remains neutral with regard to jurisdictional claims in pub- the prediction of diffuse lower-grade gliomas subtype, IDH mutant with lished maps and institutional affiliations. no 1p19q codeletion: added value to the T2/FLAIR mismatch sign. Eur Radiol 30(2):844–854 25. Sudre CH, Panovska-Griffiths J, Sanverdi E et al (2020) Machine learning assisted DSC-MRI radiomics as a tool for glioma classification by grade and mutation status. BMC Med Inform Decis Mak 20(1):1–14 26. Wu H, Tong H, Du X et al (2020) Vascular habitat analysis based on dynamic susceptibility contrast perfusion MRI predicts IDH mutation status and prognosis in high-grade gliomas. Eur Radiol 30(6):3254–3265 27. Xing Z, Yang X, She D, Lin Y, Zhang Y, Cao D (2017) Noninvasive assess- ment of IDH mutational status in World Health Organization grade II and III astrocytomas using DWI and DSC-PWI combined with conventional MR imaging. AJNR Am J Neuroradiol 38(6):1138–1144 28. Xing Z, Zhang H, She D et al (2019) IDH genotypes differentiation in glio - blastomas using DWI and DSC-PWI in the enhancing and peri-enhancing region. Acta Radiol 60(12):1663–1672 29. Zhang H, Lyu G, He W et al (2020) DSC and DCE histogram analyses of glioma biomarkers, including IDH, MGMT, and TERT, on differentiation and survival. Acad Radiol 27(12):e263–e271 30. Brendle C, Klose U, Hempel J-M et al (2020) Association of dynamic susceptibility magnetic resonance imaging at initial tumor diagnosis http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Insights into Imaging Springer Journals

A systematic review and meta-analysis on the differentiation of glioma grade and mutational status by use of perfusion-based magnetic resonance imaging

Download PDF
12 pages

Loading next page...
 
/lp/springer-journals/a-systematic-review-and-meta-analysis-on-the-differentiation-of-glioma-EiVJPvxbxu

References (51)

Publisher
Springer Journals
Copyright
Copyright © The Author(s) 2022
eISSN
1869-4101
DOI
10.1186/s13244-022-01230-7
Publisher site
See Article on Publisher Site

Abstract

Background: Molecular characterization plays a crucial role in glioma classification which impacts treatment strategy and patient outcome. Dynamic susceptibility contrast (DSC) and dynamic contrast enhanced (DCE) perfusion imaging have been suggested as methods to help characterize glioma in a non-invasive fashion. This study set out to review and meta-analyze the evidence on the accuracy of DSC and/or DCE perfusion MRI in predicting IDH genotype and 1p/19q integrity status. Methods: After systematic literature search on Medline, EMBASE, Web of Science and the Cochrane Library, a qualita- tive meta-synthesis and quantitative meta-analysis were conducted. Meta-analysis was carried out on aggregated AUC data for different perfusion metrics. Results: Of 680 papers, twelve were included for the qualitative meta-synthesis, totaling 1384 patients. It was observed that CBV, ktrans, Ve and Vp values were, in general, significantly higher in IDH wildtype compared to IDH mutated glioma. Meta-analysis comprising of five papers (totaling 316 patients) showed that the AUC of CBV, ktrans, Ve and Vp were 0.85 (95%-CI 0.75–0.93), 0.81 (95%-CI 0.74–0.89), 0.84 (95%-CI 0.71–0.97) and 0.76 (95%-CI 0.61–0.90), respectively. No conclusive data on the prediction of 1p/19q integrity was available from these studies. Conclusions: Future research should aim to predict 1p/19q integrity based on perfusion MRI data. Additionally, correlations with other clinically relevant outcomes should be further investigated, including patient stratification for treatment and overall survival. Keywords: Dynamic contrast enhancement magnetic resonance perfusion imaging, Dynamic susceptibility contrast magnetic resonance perfusion imaging, Glioma, Molecular classification Key points • Perfusion MR imaging shows a promising method to characterize glioma non-invasively. • Significant higher perfusion metrics are observed in IDH-wildtype glioma. • The effects of 1p/19q mutations on perfusion metrics *Correspondence: [email protected] are understudied and remain unelucidated. Department of Medical Imaging, Radboud University Medical Center, Geert Grooteplein Zuid 10, 6525 EZ Nijmegen, The Netherlands Full list of author information is available at the end of the article © The Author(s) 2022. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/. van Santwijk et al. Insights into Imaging (2022) 13:102 Page 2 of 12 To this end, artificial intelligence applied to con - Introduction ventional MRI sequences (i.e., pre- and post-contrast Following the 2016 World Health Organization (WHO) T1-weighted, T2-weighted and T2-weighted FLAIR classification system of tumors of the central nervous sys - images) to predict mutational status has provided prom- tem, the high-grade glioma group can be divided in two ising results in recent years (for a review, see [8]). In addi- subgroups. One subgroup comprises the anaplastic oli- tion, various signs have been identified which can help godendroglioma IDH mutant and 1p/19q codeleted, and the radiologist to predict the molecular status of glioma the anaplastic oligodendroglioma not otherwise speci- in the daily clinical setting. For example, the T2-FLAIR fied. The second subgroup comprises the IDH mutant mismatch sign has been found to be a reliable non-inva- glioblastoma, the IDH wildtype glioblastoma, and the sive marker for identification of IDH mutant astrocyto - glioblastoma not otherwise specified [1]. Knowledge on mas [9]. the exact mutational status of glioma is not only impor- Bearing in mind the pathophysiological differences tant for classification, it also has significant impact on between various glioma subtypes and the related changes prognosis [2] and treatment strategy [3–5]. With regard in the gliomas vasculature, perfusion-based imaging to low grade gliomas, two groups of gliomas can be dis- could increase the diagnostic accuracy of non-invasive tinguished. The first groups consist of oligodendroglial characterization of glioma subtypes. For example, oligo- tumors which are isocitrate dehydrogenase (IDH) mutant dendroglial tumors are characterized by a branching pat- and 1p/19q codeleted The second groups consist of astro - tern of vascularization, whereas astrocytic glioma shows cytic tumors. It is comprised of (1) IDH mutated, 1p/19q a distinctively different vascularization [10]. Therefore, non-codeleted diffuse astrocytoma, (2) the IDH wildtype perfusion based MR imaging (either dynamic suscepti- astrocytoma, and (3) the diffuse astrocytoma not other - bility contrast (DSC) or dynamic contrast enhancement wise specified [1]. (DCE) perfusion MR imaging) has been the subject of However, the recently published WHO 2021 classi- research to non-invasively identify molecular character- fication system has placed even more emphasis on the istics [11, 12]. molecular characteristics of glioma subtypes. The group DSC-perfusion MR imaging relies on the susceptibility of diffuse astrocytic and oligodendroglial gliomas can be induced signal loss on T2*-weighted sequences, resulting subdivided based on the IDH mutations. IDH wildtype from a bolus of gadolinium-based contrast agent pass- tumors are classified as high-grade gliomas, without ing through the capillaries. The most commonly used exception. In order to be classified as glioblastoma (IDH DSC perfusion parameter is Cerebral Blood Volume wildtype; grade 4), nuclear ATRX loss has to be present. (CBV). CBV can be estimated by use of the area under Additionally, IDH wild-type diffuse astrocytic tumors in the curve (AUC) of the signal intensity-time curve [13, adults without the histological features of glioblastoma, 14]. However, more recent studies compute CBV maps but with one or more of three genetic parameters (tel- by integrating the transverse relaxivity changes which omerase reverse transcriptase gene [TERT] promoter occur dynamically over a first-pass injection followed by mutation, epidermal growth factor receptor [EGFR] gene leakage correction due to the leaky blood–brain barrier amplification, or combined gain of entire chromosome 7 in most tumors (for a recent overview and recommenda- and loss of entire chromosome 10 [+ 7/ − 10]) are now tions, see [15]). DCE-perfusion MR imaging relies on the also classified as glioblastoma. In the 2021 classification, evaluation of T1 shortening induced by a gadolinium- all IDH-mutant diffuse astrocytic tumors with intact based contrast agent bolus leaking from the blood vessels 1p/19q chromosomes are considered a single type called into the tissue. Pharmacokinetic modeling can be used to astrocytoma, IDH-mutant with WHO grades ranging derive various values including, Ve and Vp. ktrans repre- from 2 to 4.  Grading of these tumors takes into account sents the capillary permeability; Ve represents the frac- molecular findings such as the homozygous deletion tional volume of the gadolinium-based contrast agent in of CDKN2A/B, which is associated with a worse prog- the extravascular-extracellular space; Vp represents the nosis. IDH-mutant astrocytomas with these molecular fractional volume of the of the gadolinium-based con- alterations will be classified as WHO grade of 4, even trast agent in the plasma space [13]. if microvascular proliferation or necrosis is absent [6]. Although various studies with different methodologies Additionally, IDH mutant oligodendroglial gliomas with and outcomes have been published since the release of codeleted 1p/19q chromosomes are considered oligoden- the WHO 2016 classification system of glioma, a com - drogliomas. While the establishment of the sophisticated prehensive overview of the accuracy of perfusion based molecular markers to classify gliomas is an important MR imaging to predict the molecular characteristics of advance in glioma diagnosis, all of the literature which glioma is still lacking. In addition, a systematic overview is covered within this review is based on the 2016 WHO of the literature on this topic could help to shape future classification of central nervous system tumors [6, 7]. v an Santwijk et al. Insights into Imaging (2022) 13:102 Page 3 of 12 research and daily clinical practice to focus on the most were cross-checked afterward, and discrepancies were promising technique (either DSC- or DCE-perfusion resolved in consensus. MRI). The aim of the current paper was therefore to pro - vide an overview of the relevant literature with regard to Qualitative meta‑synthesis and quantitative meta‑analysis the use of DSC and DCE perfusion imaging used to dif- Eligible literature was synthesized qualitatively follow- ferentiate glioma grade and mutational status. ing the PICO-strategy as proposed by Eriksen et  al. [17]. Also, quality of primary diagnostic accuracy stud- Materials and methods ies was assessed using the QUADAS-2. Meta-analysis Search strategy and inclusion/exclusion methodology was conducted on the AUC and the 95% Confidence This systematic review and meta-analysis was conducted Interval (95%-CI) using a random effects model. From following the Preferred Reporting Items for System- the included studies, perfusion metrics and the afore- atic Reviews and Meta-Analyses (PRISMA) statement mentioned statistics were extracted. If one of these vari- [16]. Databases searched for literature were: Medline ables was missing, the researchers aimed to re-calculate (accessed through PubMed), EMBASE, Web of Science, the value when possible [18]. In addition, corresponding and the Cochrane Library. The full search strategies for authors were contacted to provide missing details, with each database are made available in the Additional file  1. up to two reminders send by e-mail. When not all neces- Cross-referencing was used to add relevant literature to sary data could be acquired, studies could not be meta- the database. Searches were conducted between May 1, analyzed. Meta-analyses were conducted on different 2020 and January 1, 2021. Inclusion criteria were: (1) the subgroups of target conditions. Meta-analysis was per- use of either DSC or DCE perfusion MRI; (2) the inclu- formed with the use of OpenMetaAnalyst (MetaAnalyst, sion of patients suffering from glioma; (3) glioma grading Tufts Medical Center) [19] and/or SPSS (version 25; IBM and classification by use of the WHO 2016 classification Corp., Armonk, NY) and results were displayed in forest system [1]; and (4) the aim of the study needed to com- plots. The Higgins test was used to test for heterogene - prise the non-invasive classification of histopathological ity between included studies. Low heterogeneity between features and/or molecular characteristics (WHO grade, groups is marked with an I < 40%, whereas considerable IDH genotype and/or 1p/19q codeletion status). Besides, heterogeneity is indicated by I > 75% [18]. papers needed to report results as quantitative measures (e.g., sensitivity, specificity, mean accuracy and/or mean Results area under the receiver operator curve (AUC)). Papers A total of 552 studies were identified after systematic were excluded if they were based on animals or non- searching. Duplicates were removed and 379 papers were human samples or a pediatric population. Letters, pre- systematically screened on title and abstract resulting in prints, case reports, congress proceedings, and narrative the inclusion of 34 papers for full-text analysis. Reasons reviews were excluded as well. for exclusion of the 345 papers are provided in Fig.  1. All papers were independently assessed by two After full-text analysis, the investigators met to discuss researchers in three steps. First, screening on title and the identified non-consensus papers to resolve disagree - abstract was carried out. Second, full-text analysis was ments and to reach consensus. Of the 34 papers, 12 could employed to assess whether the papers met the inclu- be included in the qualitative meta-synthesis. Twenty- sion- and/or exclusion criteria. Finally, information was two papers were therefore excluded (details provided in extracted from the included papers. Researchers met Fig.  1). No discrepancies between the judgement of the periodically to discuss their findings and resolve dis - two researchers remained after discussion, resulting in crepancies. Standardized tables were used to acquire the the final inclusion of 12 papers for the qualitative meta- information of interest from the included articles by two synthesis [20–31] (Fig. 1). Five papers provided sufficient researchers (LvS and DH) independently. Data extracted data to be included in the quantitative meta-analysis [22, from each study were (a) first author and year of publi - 24, 26, 27, 29] (Table 1). cation, (b) number of patients included, (c) mean age Using the QUADAS2 (QUality Assessment tool for of the included participants, (d) gender of the included Diagnostic Accuracy Studies), the most current version participants, (e) use of DSC and/or DCE, (f ) which his- of the QUADAS tool of the QUADAS task force, the topathological/molecular outcome was assessed, (g) risk of bias was considered low in all included studies perfusion based MR imaging metrics and (h) accompa- (Table 2). nying statistics (e.g., AUC value, standard deviation, 95% confidence interval (CI) and/or standard error). Perfor - Qualitative meta‑synthesis mance was expressed in accuracy, AUC and/or sensi- The twelve included studies [20–31] totaled 1384 patients tivity and specificity for each outcome. Extracted data (792 males; 592 females) suffering from glioma. Gliomas van Santwijk et al. Insights into Imaging (2022) 13:102 Page 4 of 12 Fig. 1 PRISMA flow diagram v an Santwijk et al. Insights into Imaging (2022) 13:102 Page 5 of 12 Table 1 Overview of the included studies Authors (year) N Age (years) M/F MRI perfusion Glioma types and Outcome assessed Major findings method + details on analysis grades included Brendle et al. (2020) [30] 56 Mean age 48.0 ± 16.0 33/23 DSC WHO grade II: 29 IDH mutation status & 1p/19q The mean rCBV was significantly Gradient echo sequence WHO grade III: 20 codeletion status different between the astrocytic Pre-bolus of contrast agent WHO grade IV: 7 tumors, oligodendrogliomas and was applied (0.025 mmol/kg IDHmut: 32 IDHmut astrocytic tumors and gadobutrol) IDHwt: 24 oligodendrogliomas and IDHwt Mean rCBV values astrocytic tumors BSW-model based leakage correction Choi et al. (2019) [20] 463 Mean age 52.2 ± 14.8 272/191 DSC WHO grade II: 32 IDH mutation status The IDH mutation status predic- Gradient echo sequence WHO grade III: 142 tions had an accuracy, sensitivity, Pre-bolus of 0.1 mmol/kg WHO grade IV: 289 and specificity of 92.8%, 92.6%, gadobutrol IDHmut: 328 and 93.1%, respectively, in the Mean rCBV values IDHwt: 125 validation set with an AUC of 0.9 No information on post- 1p/19q codel: 56 (95%-CI 0.969–0.991). In the test processing with regard to 1p/19q non-codel: 407 set, the IDH genotype prediction leakage-correction had an accuracy, sensitivity, and specificity of 91.7%, 92.1%, and 91.5%, respectively, with an AUC of 0.95 (95%-CI 0.898–0.982) Hempel et al. (2019) [21] 100 Mean age 51.4 ± 14.7 55/45 DSC WHO grade II: 40 IDH mutation status rCBV was significantly lower in Gradient echo sequence WHO grade III: 30 patients with IDHmut than in Pre-bolus of 0.1 mmol/kg WHO grade IV: 30 those with the IDHwt. Mean gadobutrol IDHmut: 31 rCBV values showed a sensitivity/ Mean rCBV values* IDHwt: 46 specificity of 52/91 for the pre - BSW-model based leakage 1p/19q codel: 23 diction of IDH mutation status correction with an AUC of 0.780 Hilario et al. (2019) [22] 49 Range 28/21 DSC LGG: 8 IDH mutation status Significant differences in the val- (X) 16–78 Gradient echo sequence HGG: 41 ues of leakage (p = 0.01), Ktrans Pre-bolus of 0.1 mmol/kg IDHmut: 10 (p = 0.002), Vp (p = 0.032) and Ve gadobutrol IDHwt: 31 (p < 0.001) between high-grade No rCBV values provided and low-grade diffuse gliomas BSW-model based leakage were observed correction The highest AUC was demon- strated by the DCE permeability parameters Ktrans (AUC = 0.838, CI95% 0.710–0.967, p = 0.003) and Ve (AUC = 0.878, CI95% 0.768–0.988, p = 0.001). Among IDHmut and IDHwt highgrade gliomas, there were significant differences in leakage (p = 0.004) and Ktrans values (p = 0.028) showing lower leakage and Ktrans values van Santwijk et al. Insights into Imaging (2022) 13:102 Page 6 of 12 Table 1 (continued) Authors (year) N Age (years) M/F MRI perfusion Glioma types and Outcome assessed Major findings method + details on analysis grades included DCE Dynamic gradient echo sequence 5 mL of gadobutrol at a rate of 3 mL/s Mean Ktrans, Vp, Ve and Kep values Model based leakage correc- tion Lee et al. (2018) [23] 39 Mean age 43.6 (range 21–82) 19/20 DSC WHO grade II: 19 WHO grade Ktrans, Kep, and Ve showed Gradient echo sequence WHO grade III: 20 tendencies toward higher values No prebolus administration in oligodendroglial tumors than described astrocytic tumors Mean rCBV values* BSW-model based leakage correction DCE Dynamic gradient echo sequence 0.1 mmol/kg gadobutrol at a rate of 4 mL/s Mean Ktrans Model based leakage correc- tion Lee et al. (2020) [24] 110 IDHmut.- 1p/19q noncodel: 56/54 DSC WHO grade II: 45 IDH mutation status & 1p/19q When using nCBV skewness, (X) mean age 40.7 ± 12.8; Gradient echo sequence WHO grade III: 65 codeletion status the AUC was found to be 0.690 IDHwt: mean age 51.2 ± 14.0; Pre-bolus of 0.1 mmol/kg IDHmut: 65 (95%-CI: 0.573, 0.807) with a IDHmut-1p/19q codel: mean gadoterate meglumine IDHwt: 45 sensitivity of 84.2 and specificity age 46.5 ± 11.7 Mean normalized rCBV values* 1p/19q codel: 46 of 59.3 to distinguish IDHmut- BSW-model based leakage 1p/19q non-codel: 19 1p/19q noncodel from the other correction two groups v an Santwijk et al. Insights into Imaging (2022) 13:102 Page 7 of 12 Table 1 (continued) Authors (year) N Age (years) M/F MRI perfusion Glioma types and Outcome assessed Major findings method + details on analysis grades included Sudre et al. (2020) [25] 333 Mean age 48.9 (range 20–81) 198/135 DSC WHO grade II: 101 WHO grade & IDH mutation Shape, distribution and texture No details provided on imag- WHO grade III: 74 status features showed significant dif- ing protocol and whether or WHO grade IV: 158 ferences across mutation status. not a prebolus was adminis- IDHmut: 151 WHO grade II-III differentiation tered IDHwt: 182 was mostly driven by shape fea- Mean rCBV values* tures, while texture and intensity BSW-model based leakage feature were more relevant for correction the III-IV separation. Increased number of features became significant when differentiating grades further apart from one another. Gliomas were correctly stratified by mutation status in 71% and by grade in 53% of the cases Wang et al. (2020) [31] 30 IDHmut: mean age 42.8 (range 17/13 DCE WHO grade II: 22 IDH mutation status Compared to IDHmut LGGs, (X) 22–67) Dynamic gradient echo WHO grade III: 8 IDHwt LGGs exhibited signifi- IDHwt: sequence IDHmut: 18 cantly higher perfusion metrics mean age 47.9 (range 19–78) Pre-bolus of 0.1 mmol/kg IDHwt: 12 (p < 0.05) gadopentetate dimeglumine at a rate of 4 mL/s Mean Ktrans, Vp, Ve Model based leakage correc- tion Wu et al. (2020) [26] 44 63.8 ± 7.4 27/17 DSC WHO grade III: 19 IDH mutation status Compared with IDHwt, IDHmut Gradient echo sequence WHO grade IV: 25 had significantly decreased No pre-bolus administration** IDHmut: 19 rCBV at the high-angiogenic Mean rCBV values IDHwt: 25 enhancing tumor habitats and Gamma-variate curve fitting 1p/19q codel: 7 low-angiogenic enhancing leakage correction 1p/19q non-codel: 3 tumor habitats Xing et al. (2017) [27] 42 IDHmut: mean age 35.8 ± 9.1 26/16 DSC WHO grade II: 24 IDH mutation status The threshold value of < 2.35 (X) IDHwt: mean age Gradient echo sequence WHO grade III: 18 for relative maximum CBV in 46.0 ± 18.4 Pre-bolus of 0.1 mmol/kg IDHmut: 17 the prediction of IDH mutation gadobenate dimeglumine IDHwt: 25 status provided a sensitivity, Mean rCBVmax values* specificity, positive predictive BSW-model based leakage value, and negative predictive correction value of 100.0%, 60.9%, 85.6%, and 100.0%, respectively van Santwijk et al. Insights into Imaging (2022) 13:102 Page 8 of 12 Table 1 (continued) Authors (year) N Age (years) M/F MRI perfusion Glioma types and Outcome assessed Major findings method + details on analysis grades included Xing et al. (2019) [28] 75 IDHmut: mean age 52.2 ± 12.7 41/34 DSC WHO grade IV: 75 IDH mutation status Both rCBVmax-t and rCBVmax-p IDHwt: mean age 40.7 ± 10.8 Gradient echo sequence IDHmut: 10 showed significant differences Pre-bolus of 0.1 mmol/kg IDHwt: 65 between IDHmut and IDHwt. gadobenate dimeglumine The optimal cutoff values in Mean rCBVmax values* prediction of IDH-m. < 7.27 for BSW-model based leakage rCBVmax-tumor, and < 0.97 for correction rCBVmax-peri-enhancing region Zhang et al. (2020) [29] 43 47.0 ± 13.0 20/23 DSC WHO grade II: 14 IDH mutation status Ve (AUC = 0.816, sensitiv- (X) Gradient echo sequence WHO grade III: 14 ity = 0.84, specificity = 0.79) and (X) DSC imaging followed DCE WHO grade IV: 15 Kep (AUC = 0.818, sensitiv- imaging; no separate pre-bolus IDHmut: 20 ity = 0.76, specificity = 0.78) was administered IDHwt: 23 provided the highest differential Mean rCBVmax values* efficiency for IDH mutation BSW-model based leakage status prediction correction DCE Dynamic gradient echo sequence Pre-bolus of 0.1 mmol/kg gadodiamide Mean Ktrans, Vp, Ve* Model based leakage correc- tion Marked in italics are the publications included in the meta-analysis on the use of DSC-value; Marked in bold are the publications included in the meta-analysis on the use of the DCE-values AUC, area under the curve; DCE, dynamic contrast enhancement magnetic resonance perfusion imaging; DSC, dynamic susceptibility contrast magnetic resonance perfusion imaging; F, females; HGG, high-grade glioma; IDH, isocitrate dehydrogenase; IDHmut, mutation of the isocitrate dehydrogenase gene(s); IDHwt, wild-type isocitrate dehydrogenase gene(s); Kep, rate constant between the extravascular extracellular space and blood plasma; ktrans, volume transfer coefficient; LGG, low grade glioma; M, males; MRI, magnetic resonance imaging; nCBV, normalized cerebral blood volume; rCBV, relative cerebral blood volume; rCBVmax-t, maximum relative cerebral blood volume in the tumor-enhancing region; rCBVmax-p, maximum relative cerebral blood volume in the peri-enhancing region; Ve, fractional volume of the extravascular extracellular space; Vp, fractional blood plasma volume; WHO, World Health Organization; 95%-CI, 95%-confidence interval *Study provides a variety of perfusion statistics (either DSC or DCE metrics; values included mean, standard deviation and a variety of percentiles) **Lack of pre-bolus administration was compensated by use of a flip-angle of 60° which reduced T1 effects [44] v an Santwijk et al. Insights into Imaging (2022) 13:102 Page 9 of 12 Table 2 Combined effect size for the different DCE/DSC When using DCE perfusion imaging, IDHmut HGG parameters (either WHO 2021 Astroctytoma grade 3 or 4 or WHO 2021 Oligodendroglioma grade 3) showed significantly Ktrans Ve Vp CBV lower ktrans values as compared to IDHwt HGG (WHO Eec ff t size 0.813 0.844 0.777 0.832 2021 Astrocytoma grade 4) [22]. In oligodendroglial Standard error 0.02 0.03 0.03 0.03 tumors (WHO 2021 Oligodendroglioma, IDHmut, 95%-CI lower limit 0.726 0.766 0.683 0.749 1p/19q-codeleted; Grade 2 or 3), however, Lee et  al. 95%-CI upper limit 0.900 0.921 0.871 0.914 found that ktrans, Kep and Ve showed tendencies toward ktrans, volume transfer coefficient; rCBV, relative cerebral blood volume; Ve, higher values as compared to astrocytic tumors [23]. Ve fractional volume of the extravascular extracellular space; Vp, fractional blood and Vp values were found to be significantly lower in plasma volume; 95%-CI, 95%-confidence interval IDHmut glioma (WHO 2021 Astrocytoma and WHO 2021 Oligodendroglioma) as compared to IDHwt glioma, regardless of WHO II-IV grading [22, 29]. Based on Ve could be subdivided into WHO grade II (n = 326); WHO and Kep values, a sensitivity/specificity of 84%/79% and grade III (n = 410) and WHO grade IV (n = 599). Regard- 76%/78% was observed with regard to differentiate IDH ing the IDH genotype, 701 gliomas were IDH-mutated mutation status [29]. The study of Hilario et al. also sug - and 603 tumors expressed an IDH wildtype genotype. gested that ktrans, Vp and Ve could be used to differenti - 1p/19q codeletion (WHO 2021 Oligodendroglioma ate between LGG and HGG non-invasively [22]. WHO grade 2 or 3) was observed in 132 tumors; non- Studies using artificial intelligence showed promising codeletion of 1p/19q chromosome arms was observed in results with regard to prediction of IDH mutation sta- 429 tumors. All included papers used histopathological/ tus. Choi et  al. showed that a convolutional long short- molecular assessment by a trained neuropathologist who term memory model with an attention mechanism had adhered to the WHO 2016 glioma classification as the an accuracy, sensitivity, and specificity of 92.8%, 92.6%, gold standard. and 93.1%, respectively, in the validation set (AUC: 0.98; Eight papers used DSC perfusion MRI [20, 21, 24–28, 95%-CI 0.969–0.991) with regard to IDH genotype pre- 30]; three papers used both DCE and DSC perfusion MRI diction by use of DSC perfusion MRI. In the test set, an [22, 23, 29]; one paper used DCE perfusion MRI only accuracy, sensitivity, and specificity of 91.7%, 92.1%, and [26]. Two papers used artificial intelligence methods to 91.5% were observed, respectively. The AUC value of the assess different perfusion metrics between various sub - IDH genotype prediction demonstrated to be 0.95 with types of gliomas [20, 25], whereas the other publications a 95% CI ranging between 0.898 and 0.982. Subsequent used more traditional statistics. analysis of the signal intensity curves of DSC imaging As assessed by DSC perfusion MRI, IDHmut glioma elucidated high attention on the combination of the end displayed significantly lower rCBV values as compared to of the pre-contrast baseline, the up/downslopes of signal IDHwt glioma [21, 26–28, 30]. When using a retrospec- drops, and/or post-bolus plateaus for the curves used to tively determined rCBVmax threshold value of < 2.35, the predict IDH genotype [20]. Another study showed that authors described a sensitivity/specificity of 100%/61% when using a random forest algorithm, shape, distribu- and AUC of 0.82 (95%-CI: 0.66–0.93) when differentiat - tion and rCBV-extracted features elucidated significant ing IDHmut (either WHO 2021 Astroctytoma grade 2, differences across mutation status. WHO grade II-III dif - 3 or 4 or WHO 2021 Oligodendroglioma grade 2 or 3) ferentiation was mostly driven by shape features, while and IDHwt gliomas [27]. By use of the skewness of nor- texture and intensity feature were more relevant for the malized CBV (nCBV) values (normalized by use of the distinguishing of III and IV. Based on this random forest CBV value of the normal-appearing contralateral cen- algorithm, gliomas were correctly stratified by mutation trum semiovale), IDHmut, 1p/19q non-codeleted glioma status in 71% and by grade in 53% of the cases [25]. (WHO 2021 Astrocytoma grade 2, 3 or 4) could be dis- tinguished from IDHwt glioma and IDHmut, 1p/19q Meta‑analysis codeleted glioma (WHO 2021 Oligodendroglioma) with Meta-analysis of the data (n = 237 patients) showed that a sensitivity/specificity of 84%/59% (AUC-value of 0.690 CBV values have an accuracy of correctly predicting IDH and 95%-CI 0.573–0.807). IDHmut, 1p/19q non-code- genotype with an AUC of 0.832 and a standard error of leted gliomas (WHO 2021 Astrocytoma grade 2, 3 or 4) 0.03 (95%-CI 0.75–0.91). When using DCE parameters showed significant lower nCBV values compared to the (n = 122), an AUC of 0.81, 0.84 and 0.78 is observed for IDHmut, 1p/19q codeleted gliomas (WHO 2021 Oligo- ktrans, Ve and Vp, respectively. Standard errors (and dendroglioma grade 2 or 3) and the IDH wildtype glio- 95%-CI) for ktrans, Ve and Vp were found to be 0.02 mas [24]. van Santwijk et al. Insights into Imaging (2022) 13:102 Page 10 of 12 (95%-CI 0.73–0.90), 0.03 (95%-CI 0.77–0.92) and 0.03 angiopoietin 2 in WHO grade II glioma was confirmed by (95%-CI 0.68–0.87), respectively (Table  2). The corre - staining of human tumor tissue microarrays. More spe- sponding Forest-Plots of the different perfusion metrics cifically, IDHwt gliomas displayed a specific angiogenic are provided in Fig.  2. I analysis showed that included gene expression signature (i.e., upregulation of Angiopoi- DCE-MRI studies were homogeneous (I < 1%). In the etin 2 and serpin family H) which resulted in enhanced individual analyses of ktrans, Ve and Vp, studies were endothelial cell migration and matrix remodeling. In found to be non-significantly heterogeneous (p = 0.834; the same study, transcription factor analysis indicated p = 0.548; p = 0.519, respectively). The meta-analysis of increased transforming growth factor beta and hypoxia DSC-MRI studies showed to have moderate heterogene- signaling in IDHwt gliomas [39]. Based on these studies, ity (I = 35%; p = 0.215). The role of perfusion MRI met - we can conclude that gliomas with different IDH geno - rics in predicting the 1p/19q-codeletion status could not type have distinct molecular vascularization. In addition, be meta-analyzed using the acquired data. the blood vessels in LGG displayed alterations in gene expression which partially overlapped with changes pre- Discussion viously identified in HGG vessels [39]. As IDHwt glioma This systematic review and meta-analysis shows that vessels have a distinct vascular gene expression pattern perfusion MRI can be used to effectively predict IDH associated with vascular remodeling, these microstruc- genotype non-invasively following the WHO 2016/2021 tural changes can be used to explain why IDHwt glioma glioma classification. Different DSC or DCE perfusion show significantly higher perfusion metrics compared parameters were found to have an equal performance to IDHmut glioma. These insights in genotype and phe - regarding the non-invasive prediction of IDH genotype. notype justify the use of perfusion MRI to predict IDH Prediction of the 1p/19q-codeletion status could not be genotype. The role of 1p/19q codeletion status on angio - meta-analyzed using the acquired data. genesis and vascular growth, however, remains partially The role of perfusion MRI in non-invasive glioma clas - elusive. Previous research demonstrated that 1p/19q sification can be significant and can be explained by the codeletion was associated with higher CBV values com- different glioma vasculature fingerprints which provide pared with glioma with intact alleles [40]. Another paper a specialized microenvironment for glioma cells [32]. reported specific genotypic differences in oligodendro - Within HGG, blood vessels are abnormal and display a glioma by use of DSC perfusion MRI with significantly distinct gene expression signature which differs from the higher rCBVmax values in LGG with 1p/19q codeletion genotype of blood vessels in normal brain tissue [33–35]. [41]. It is believed that 1p/19q codeleted LGG show an These genotypic differences result in high expression of increased metabolism and angiogenesis and have an certain angiogenic factors, including vascular endothelial extensive internal vascular network. This is supported by growth factor, transforming growth factor β2, and pleio- the study of Kapoor et al. in which a significantly higher trophin [35–38]. In LGG, on the other hand, potential rCBVmax was observed in 1p/19q codeleted LGG (WHO molecular alterations regarding angiogenesis have been 2021 Oligodendroglioma WHO grade 2). Additionally, an investigated less extensively. In 2018, Zhang et  al. dem- increased vascular endothelial growth factor expression, onstrated that WHO grade II glioma expressed an inter- CD31, and CD105, was observed as compared with gli- mediate stage of vascular abnormality, less severe than oma with intact alleles [42]. that of glioblastoma vessels but distinct from normal ves- The clinical usability of MRI perfusion imaging to sels. Enhanced expression of laminin subunit alpha 4 and predict IDH genotype remains partially elusive as the Fig. 2 Forest-plot of the area under the curve (AUC) of the receiver operator curve (ROC) of the different perfusion metrics in predicting IDH mutation status. IDH, isocitrate dehydrogenase, ktrans, volume transfer coefficient; rCBV, relative cerebral blood volume; Ve, fractional volume of the extravascular extracellular space; Vp, fractional blood plasma volume; 95%-CI, 95%-confidence interval v an Santwijk et al. Insights into Imaging (2022) 13:102 Page 11 of 12 differences were based on aggregated results. Although Supplementary Information several papers provided specific threshold values [27, The online version contains supplementary material available at https:// doi. org/ 10. 1186/ s13244- 022- 01230-7. 28], no clinically useful threshold values are available. The current review did not include studies arterial spin Additional file 1: Search strategies. labeling (ASL) as a perfusion MR imaging method as only sparse literature with regard to ASL was found in explor- Author contributions atory literature searches. A recent paper by Wang et  al. L.vS. and V.K. systematically searched and reviewed the available literature (2019) reported that only a mild correlation was found under direct supervision of D.H. D.H. helped to solve disagreements when- ever needed. L.vS. and V.K., under direct supervision of D.H., carried out the between the IDH1 genotypes and ASL derived glioma meta-analysis. L.vS. and V.K. wrote the first draft of the manuscript, which was perfusion parameters. There was no significant associa - reviewed and corrected by D.H., F.M. and M.S. D.H. wrote subsequent versions tion between 1p/19q codeletion and perfusion in grade II in close collaboration with L.vS. and V.K. These versions were corrected by F.M. and M.S. All authors read and approved the final manuscript. and III gliomas [43]. Funding Strength and limitations No funding was received for conducting this research. By adhering to the 2016 WHO glioma classification to be Availability of data and materials included, some valuable papers needed to be excluded, Data analyzed in this study will be made available upon reasonable request by though also resulted in rather homogeneous dataset to be contacting the corresponding author. meta-analyzed. One of the strengths of this reviews con- cerns the relative homogeneous imaging protocols which Declarations were meta-analyzed. For example, all DSC-imaging pro- Ethics approval and consent to participate tocols were imaged after administering a pre-bolus injec- No ethical approval was necessary to conduct this review. tion of a gadolinium-based contrast agent. Also, for the Consent for publication included studies which investigated the diagnostic accuracy All authors consent publication of this manuscript. of DCE-imaging, mean perfusion values (i.e., Ve, Vp and ktrans). However, different studies used different values Competing interests The authors declare no competing interests. of perfusion parameters (mean rCBV vs. mean rCBV max values), which partially limits the generalizability of results Author details [15]. The homogeneity of the meta-analyzed patients and Department of Medical Imaging, Radboud University Medical Center, Geert Grooteplein Zuid 10, 6525 EZ Nijmegen, The Netherlands. Depar tment histopathological outcomes (i.e., IDH genotype and 1p/19q of Radiology and Nuclear Medicine, Erasmus MC, University Medical Center codeletion status) strengthen the here described find Rotterdam, Rotterdam, The Netherlands. ings. Another limitation of the here applied methodology Received: 14 March 2022 Accepted: 20 April 2022 concerns the fact that this systematic review was executed without registration in an international database. Conclusion References This review and meta-analysis showed that accuracy of 1. Louis DN, Perry A, Reifenberger G et al (2016) The 2016 World Health DSC parameters was not different from the accuracy of Organization classification of tumors of the central nervous system: a summary. Acta Neuropathol 131(6):803–820 DCE parameters to non-invasive predict the IDH geno- 2. Lapointe S, Perry A, Butowski NA (2018) Primary brain tumours in adults. type in glioma patients. The use of perfusion MRI with Lancet 392(10145):432–446 regard to predicting 1p/19q codeletion status could not 3. Narang J, Jain R, Scarpace L et al (2011) Tumor vascular leakiness and blood volume estimates in oligodendrogliomas using perfusion CT: an be determined using these data. analysis of perfusion parameters helping further characterize genetic subtypes as well as differentiate from astroglial tumors. J Neurooncol 102(2):287–293 Abbreviations 4. Saito T, Yamasaki F, Kajiwara Y et al (2012) Role of perfusion-weighted 95%-CI: 95%-Confidence interval; AUC : Area under the curve; DCE MRI: imaging at 3 T in the histopathological differentiation between astrocytic Dynamic contrast enhancement magnetic resonance perfusion imaging; DSC and oligodendroglial tumors. Eur J Radiol 81(8):1863–1869 MRI: Dynamic susceptibility contrast magnetic resonance perfusion imaging; 5. Lecavalier-Barsoum M, Quon H, Abdulkarim (2014) Adjuvant treatment HGG: High-grade glioma; IDH: Isocitrate dehydrogenase; IDHmut: Mutation of of anaplastic oligodendrogliomas and oligoastrocytomas. Cochrane the isocitrate dehydrogenase gene(s); IDHwt: Wild-type isocitrate dehydroge- Database Syst Rev (5) nase gene(s); Kep: Rate constant of gadolinium efflux between the extravas- 6. Louis DN, Perry A, Wesseling P et al (2021) The 2021 WHO classifica- cular extracellular space and blood plasma; ktrans: Volume transfer constant; tion of tumors of the central nervous system: a summary. Neuro Oncol LGG: Low grade glioma; MRI: Magnetic resonance imaging; nCBV: Normalized 23(8):1231–1251 cerebral blood volume; rCBV: Relative cerebral blood volume; rCBVmax-p: 7. Weller M, van den Bent M, Preusser M et al (2021) EANO guidelines on Maximum relative cerebral blood volume in the peri-enhancing region; the diagnosis and treatment of diffuse gliomas of adulthood. Nat Rev Clin rCBVmax-t: Maximum relative cerebral blood volume in the tumor-enhancing Oncol 18(3):170–186 region; Ve: Fractional volume of the extravascular extracellular space; Vp: Fractional blood plasma volume; WHO: World Health Organization. van Santwijk et al. Insights into Imaging (2022) 13:102 Page 12 of 12 8. van Kempen E, Post M, Mannil M et al (2021) Accuracy of machine learn- with the prognosis of different molecular glioma subtypes. Neurol Sci ing algorithms for the classification of molecular features of gliomas on 41(12):3625–3632 MRI: a systematic literature review and meta-analysis. Cancers 13:2606 31. Wang X, Cao M, Chen H, Ge J, Suo S, Zhou Y (2020) Simplified perfusion 9. Jain R, Johnson DR, Patel SH et al (2020) “Real world” use of a highly reli- fraction from diffusion-weighted imaging in preoperative prediction of able imaging sign: “T2-FLAIR mismatch” for identification of IDH mutant IDH1 mutation in WHO grade II–III gliomas: comparison with dynamic astrocytomas. Neuro Oncol 22(7):936–943 contrast-enhanced and intravoxel incoherent motion MRI. Radiol Oncol 10. Zhou J, Li N, Yang G, Zhu Y (2011) Vascular patterns of brain tumors. Int J 54(3):301–310. https:// doi. org/ 10. 2478/ raon- 2020- 0037. Surg Pathol 19(6):709–717 32. Takano S (2012) Glioblastoma angiogenesis: VEGF resistance solutions 11. Brendle C, Hempel J-M, Schittenhelm J et al (2018) Glioma grading and and new strategies based on molecular mechanisms of tumor vessel determination of IDH mutation status and ATRX loss by DCE and ASL formation. Brain Tumor Pathol 29(2):73–86 perfusion. Clin Neuroradiol 28(3):421–428 33. Madden SL, Cook BP, Nacht M et al (2004) Vascular gene expression in 12. Ma H, Wang Z, Xu K et al (2017) Three-dimensional arterial spin labeling nonneoplastic and malignant brain. Am J Pathol 165(2):601–608 imaging and dynamic susceptibility contrast perfusion-weighted imag- 34. Pen A, Moreno MJ, Martin J, Stanimirovic DB (2007) Molecular markers of ing value in diagnosing glioma grade prior to surgery. Exp Ther Med extracellular matrix remodeling in glioblastoma vessels: microarray study 13(6):2691–2698 of laser-captured glioblastoma vessels. Glia 55(6):559–572 13. Essig M, Shiroishi MS, Nguyen TB et al (2013) Perfusion MRI: the five most 35. Dieterich LC, Mellberg S, Langenkamp E et al (2012) Transcriptional profil- frequently asked technical questions. AJR Am J Roentgenol 200(1):24–34 ing of human glioblastoma vessels indicates a key role of VEGF-A and TGF 14. Petrella JR, Provenzale JM (2000) MR perfusion imaging of the brain: beta 2 in vascular abnormalization. J Pathol 228(3):378–390 techniques and applications. AJR Am J Roentgenol 175(1):207–219 36. Zhang L, Dimberg A (2016) Pleiotrophin is a driver of vascular abnormali- 15. Boxerman JL, Quarles CC, Hu LS et al (2020) Consensus recommendations zation in glioblastoma. Mol Cell Oncol 3(6):e1141087 for a dynamic susceptibility contrast MRI protocol for use in high-grade 37. Zhang L, Laaniste L, Jiang Y, Alafuzoff I, Uhrbom L, Dimberg A (2016) gliomas. Neuro Oncol 22(9):1262–1275 Pleiotrophin enhances PDGFB-induced gliomagenesis through increased 16. Moher D, Liberati A, Tetzlaff J, Altman DG (2010) Preferred reporting items proliferation of neural progenitor cells. Oncotarget 7(49):80382–80390 for systematic reviews and meta-analyses: the PRISMA statement. Int J 38. Zhang L, Kundu S, Feenstra T et al (2015) Pleiotrophin promotes vascular Surg 8(5):336–341 abnormalization in gliomas and correlates with poor survival in patients 17. Eriksen MB, Frandsen TF (2018) The impact of patient, intervention, with astrocytomas. Sci Signal 8(406):ra125 comparison, outcome (PICO) as a search strategy tool on literature search 39. Zhang L, He L, Lugano R et al (2018) IDH mutation status is associated quality: a systematic review. J Med Libr Assoc 106(4):420–431 with distinct vascular gene expression signatures in lower-grade gliomas. 18. Higgins JPT, G.S.e (2011) Cochrane handbook for systematic reviews Neuro Oncol 20(11):1505–1516 of interventions version 5.1.0 [updated March 2011]. The Cochrane 40. Chawla S, Krejza J, Vossough A et al (2013) Differentiation between Collaboration oligodendroglioma genotypes using dynamic susceptibility contrast 19. Wallace BC, Lajeunesse MJ, Dietz G et al (2017) OpenMEE: intuitive, open- perfusion-weighted imaging and proton MR spectroscopy. AJNR Am J source software for meta-analysis in ecology and evolutionary biology. Neuroradiol 34(8):1542–1549 Methods Ecol Evol 8(8):941–947 41. Whitmore RG, Krejza J, Kapoor GS et al (2007) Prediction of oligoden- 20. Choi KS, Choi SH, Jeong B (2019) Prediction of IDH genotype in gliomas droglial tumor subtype and grade using perfusion weighted magnetic with dynamic susceptibility contrast perfusion MR imaging using an resonance imaging. J Neurosurg 107(3):600–609 explainable recurrent neural network. Neuro Oncol 21(9):1197–1209 42. Kapoor GS, Gocke TA, Chawla S et al (2009) Magnetic resonance perfu- 21. Hempel J-M, Schittenhelm J, Klose U et al (2019) In vivo molecular profil- sion-weighted imaging defines angiogenic subtypes of oligodendro - ing of human glioma cross-sectional observational study using dynamic glioma according to 1p19q and EGFR status. J Neurooncol 92(3):373–386 susceptibility contrast magnetic resonance perfusion imaging. Clin 43. Wang N, Xie S, Liu H, Chen G, Zhang W (2019) Arterial spin labeling Neuroradiol 29(3):479–491 for glioma grade discrimination: correlations with IDH1 genotype and 22. Hilario A, Hernandez-Lain A, Sepulveda JM et al (2019) Perfusion MRI 1p/19q status. Transl Oncol 12(5):749–756 grading diffuse gliomas: impact of permeability parameters on molecular 44. Welker K, Boxerman J, Kalnin A, Kaufmann T, Shiroishi M, Wintermark M biomarkers and survival. Neurocirugia 30(1):11–18 (2015) ASFNR recommendations for clinical performance of MR dynamic 23. Lee JY, Ahn KJ, Lee YS et al (2018) Differentiation of grade II and III oligo - susceptibility contrast perfusion imaging of the brain. AJNR Am J Neuro- dendrogliomas from grade II and III astrocytomas: a histogram analysis radiol 36(6):E41-51 of perfusion parameters derived from dynamic contrast-enhanced (DCE) and dynamic susceptibility contrast (DSC) MRI. Acta Radiol 59(6):723–731 Publisher’s Note 24. Lee MK, Park JE, Jo Y et al (2020) Advanced imaging parameters improve Springer Nature remains neutral with regard to jurisdictional claims in pub- the prediction of diffuse lower-grade gliomas subtype, IDH mutant with lished maps and institutional affiliations. no 1p19q codeletion: added value to the T2/FLAIR mismatch sign. Eur Radiol 30(2):844–854 25. Sudre CH, Panovska-Griffiths J, Sanverdi E et al (2020) Machine learning assisted DSC-MRI radiomics as a tool for glioma classification by grade and mutation status. BMC Med Inform Decis Mak 20(1):1–14 26. Wu H, Tong H, Du X et al (2020) Vascular habitat analysis based on dynamic susceptibility contrast perfusion MRI predicts IDH mutation status and prognosis in high-grade gliomas. Eur Radiol 30(6):3254–3265 27. Xing Z, Yang X, She D, Lin Y, Zhang Y, Cao D (2017) Noninvasive assess- ment of IDH mutational status in World Health Organization grade II and III astrocytomas using DWI and DSC-PWI combined with conventional MR imaging. AJNR Am J Neuroradiol 38(6):1138–1144 28. Xing Z, Zhang H, She D et al (2019) IDH genotypes differentiation in glio - blastomas using DWI and DSC-PWI in the enhancing and peri-enhancing region. Acta Radiol 60(12):1663–1672 29. Zhang H, Lyu G, He W et al (2020) DSC and DCE histogram analyses of glioma biomarkers, including IDH, MGMT, and TERT, on differentiation and survival. Acad Radiol 27(12):e263–e271 30. Brendle C, Klose U, Hempel J-M et al (2020) Association of dynamic susceptibility magnetic resonance imaging at initial tumor diagnosis

Journal

Insights into Imaging – Springer Journals

Published: Jun 7, 2022

Keywords: Dynamic contrast enhancement magnetic resonance perfusion imaging; Dynamic susceptibility contrast magnetic resonance perfusion imaging; Glioma; Molecular classification

There are no references for this article.