Grading of IDH-mutant astrocytoma using diffusion, susceptibility and perfusion-weighted imaging

Background The accurate grading of IDH-mutant astrocytoma is essential to make therapeutic strategies and assess the prognosis of patients. The purpose of this study was to investigate the usefulness of DWI, SWI and DSC-PWI in grading IDH-mutant astrocytoma. Methods One hundred and seven patients with IDH-mutant astrocytoma who underwent DWI, SWI and DSC-PWI were retrospectively reviewed. Minimum apparent diffusion coefficient (ADCmin), intratumoral susceptibility signal intensity(ITSS) and maximum relative cerebral blood volume (rCBVmax) values were assessed. ADCmin, ITSS and rCBVmax values were compared between grade 2 vs. grade 3, grade 3 vs. grade 4 and grade 2 + 3 vs. grade 4 tumors. Logistic regression, tenfold cross-validation,and receiver operating characteristic (ROC) curve analyses were used to assess their diagnostic performances. Results Grade 4 IDH-mutant astrocytomas showed significantly lower ADCmin and higher rCBVmax as compared to grade 3 tumors (adjusted P < 0.001). IDH-mutant grade 3 astrocytomas showed significantly lower ITSS levels as compared with grade 4 tumors (adjusted P < 0.001). ITSS levels between IDH-mutant grade 2 and grade 3 astrocytomas were significantly different (adjusted P = 0.002). Combined the ADCmin, ITSS and rCBVmax resulted in the highest AUC for differentiation grade 2 and grade 3 tumors from grade 4 tumors. Conclusion ADCmin, rCBVmax and ITSS can be used for grading the IDH-mutant astrocytomas. The combination of ADCmin, ITSS and rCBVmax could improve the diagnostic performance in grading of IDH-mutant astrocytoma.


Background
According to the 2021 World Health Organization (WHO) central nervous system (CNS) classification system, IDH-mutant astrocytomas were divided into three categories: grade 2, IDH-mutant astrocytoma, grade 3, IDH-mutant astrocytoma and grade 4, IDH-mutant astrocytoma [1]. Previous studies revealed that prognosis of IDH-mutant astrocytomas significantly varied according to their grade [2,3]. The prognosis of grade 4, IDH-mutant astrocytoma, compared with its lower grade counterparts, showed a dismal outcome. Thus, a preoperative grading of these entities could be helpful in patients' therapeutics and prognostics, especially in patients with contraindication to operation or in cases with unresectable lesions [4].
Diffusion-weighted imaging (DWI) could noninvasively assess the Brownian movement of water molecules in vivo, and indirectly reflect the cellularity of cerebral Open Access *Correspondence: dairongcao@163.com 1 Department of Radiology, First Affiliated Hospital of Fujian Medical University, 20 Cha-Zhong Road, Fuzhou 350005, Fujian, People's Republic of China Full list of author information is available at the end of the article tumors by means of apparent diffusion coefficient (ADC) values [5]. Dynamic susceptibility contrast perfusionweighted imaging (DSC-PWI) could provide hemodynamic parameters with tumor angiogenesis and may be useful for glioma grading [6]. As another advanced technique, susceptibility-weighted imaging (SWI) can semi-quantitatively assess tumor micro-hemorrhages and vasculature through intratumoral susceptibility signal intensity (ITSS) [7]. All these methods have been reported differentiation between various grades of gliomas [8][9][10][11].
However, few studies focus on grading of IDH-mutant astrocytic gliomas. Therefore, we determined to focus on our study about IDH-mutant astrocytic gliomas only, according to the updated WHO classification. We hypothesized that advanced imaging techniques including DWI, SWI, and DSC-PWI might be useful for grading of IDH-mutant astrocytic gliomas, and the combination of DWI, DSC-PWI, and SWI may contribute to improve the accuracy of diagnosis in IDH-mutant astrocytic gliomas.

Patients
This retrospective study was approved by the ethics committee of our hospital and patients' informed consent was waived. Patients from July 2014 to March 2020 were reviewed at our hospital. The inclusion criteria were the following: (1) IDH-mutant grade 2-4 astrocytomas confirmed by histopathology according to the 2021 WHO classification criteria; (2) a pretreatment MRI scans included conventional MRI(cMRI), DWI, SWI and DSC-PWI. The exclusion criteria were as follows: (1) motion artifacts; (2) poor images quality. A total of 107 patients, including 56 patients with grade 2 IDH mutant astrocytoma, 29 patients with grade 3 IDH mutant astrocytoma, and 22 patients with grade 4 IDH mutant astrocytoma, were enrolled in this study. Of these patients, 63 were male and 44 were female. The age of patients ranged from 18 to 68 years, with a mean age of 40.39 years.

MRI Protocol
Imaging was performed with 3.0 T MR systems (Magnetom Skyra or Verio; Siemens, Erlangen, Germany). The cMRI sequences included T2WI, FLAIR, T1WI, and contrast-enhanced T1WI. All cMRI sequences were performed with a field of view (FOV) = 220 mm × 220 mm, slices = 20, slice thickness = 5 mm and interslice gap = 1 mm . The total scan time of T2WI, FLAIR, T1WI,  and contrast-enhanced T1WI were 72 s, 66 s,136 s, and  66 s, respectively. DWI was performed with a single-shot echo planar imaging sequence. The b-values were 0 and 1000 s/ mm2 with diffusion gradients encoded in the 3 orthogonal directions to generate 3 sets of diffusion-weighted images. The ADC map was generated automatically by the MR imaging system. The imaging parameters were as follows: FOV = 220 mm × 220 mm, repetition time/echo time (TR/TE) = 8200/102 ms; NEX = 2.0; slices = 20; slice thickness = 5 mm; interslice gap = 1 mm. The total scan time was 1 min 32 s. SWI was performed by 3D fully flow-compensated gradient echo sequence, and a reconstruction by combining the magnitude and phase images. SWI map was generated automatically on the MR imaging scanner. The parameters were as follows: FOV = 220 mm × 220 mm;TR/ TE = 27/20 ms; flip angle(FA) = 15°; NEX = 1; slices = 60; slice thickness = 2.0 mm; intersection gap = 0.4 mm. The total scan time was 2 min 19 s.
DSC-PWI was obtained with a gradientrecalled T2*-weighted echo-planar imaging sequence. The imaging parameters were as follows: FOV = 220 mm × 220 mm; TR/TE = 1000-1250/54 ms; FA = 35°; NEX = 1.0; slices = 20; slice thickness/intersection gap = 5 mm/1 mm. When the scan was to the fourth phase of DSC-PWI, 0.1 mmol/kg body weight of gadopentetate dimeglumine was injected with an MR-compatible power injector at a flow rate of 5 ml/s through an intravenous catheter, followed by a 20 ml continuous saline flush. The series of 20 sections, 60 phases, and 1200 images were acquired in 1 min 36 s. T2* weighted signal was converted to ∆R2* for CBV processing. The whole-brain CBV maps were obtained by using a singlecompartment model, T1 extravasation correction, and an arterial input function. The anterior or medial cerebral artery was identified manually for the arterial input function.

Image analysis
For the analysis of DWI data, five non-overlapping small round regions of interest (ROIs) were carefully placed in the tumor of the ADC maps to obtained minimum ADC values (ADC min ) with visual inspection [12]. The ROIs (20-40mm 2 ) were placed in the solid part of the lesion (defined on T2WI, SWI and contrast-enhanced T1WI), avoiding necrotic, cystic, hemorrhagic, or apparent blood vessel regions that might affect the measurements of ADC values.
For the evaluation of DSC-PWI data, five non-overlapping round ROIs were placed in the tumor on the CBV map to obtained maximum CBV (CBV max ). When the rCBV ROIs were drawn, an attempt was made to place them in regions where the rCBV appeared highest. To minimize variances, the relative CBV max (rCBV max ) was measured by dividing the tumor CBV max by the mean CBV of the contralateral normal appearing white matter. Measurements of rCBVmax values were performed with the same ROIs as those used for ADC measurements. The ROIs for the ADC and rCBV measurements were not identical and were not from the same region of the tumor in each patient. ADC min and rCBV max values were obtained by another neuroradiologist (with 7 years of experience) who blinded to tumors histopathologic data..

Statistical analysis
Initially, Kolmogorov-smirnov test was used for normality of ADCmin and rCBV values. Kruskall-Wallis one way analysis of variance (ANOVA) was used to seek the difference of ADC min and rCBV max values between the three grades of the tumor. Comparisons of ADC min and rCBV values among different grade IDH-mutant astrocytomas were made with student's t-test. Interobserver variabilities of ITSSs levels were analyzed by kappa statistics. Fisher's exact test was used to assess ITSS scores among different grade IDH-mutant astrocytomas. Bonferroni correction was performed for multiple comparisons of ADC min , ITSS, and rCBV, and the corrected p values should be less than 0.0125.A simple linear regression analysis was performed to calculate the Spearman's correlation coefficient between ADC min , ITSS, rCBV max and tumor grades.
Logistic regression analysis was used to assess the diagnostic effect of total parameters from the DWI,SWI and DSC-PWI model. The receiver operating characteristic (ROC) analysis curves with tenfold cross-validation were obtained to determine the optimal cutoff value for differentiating grade 2,3 and 4 IDH-mutant astrocytoma. For each logistic regression model, cross-validation was performed with a tenfold procedure. The optimal cutoff value was determined as the point in the upper left-hand corner that maximized the sum of the sensitivity and specificity. In addition, comparisons of area under the ROC curve (AUC) for different quantitative variables were made with Z test proposed by DeLong et al. [13].

Results
The interobserver agreement of ITSS was excellent (Kappa coefficients = 0.858). The ADC min , ITSS and rCBV max values of tumors with various grades were summarized in Table 1-2 and Fig. 1. Figures 2, 3 and 4 showed the cases of IDH-mutant astrocytomas with different grades. ADC min and rCBV max values were significant different between grade 2 + 3 and grade 4 IDHmutant astrocytomas, and between grade 3 and grade 4 IDH-mutant astrocytomas (both adjusted P < 0.001). while ITSS scores were significant differentbetween grade 2 and grade 3 IDH-mutant astrocytomas, and between grade 2 + 3 and grade 4 IDH-mutant astrocytomas (adjusted P = 0.002, < 0.001, respectively).

Spearman's analysis
Spearman's test demonstrated significant relationships between the ITSS scores and the grades of tumor (r = 0.615, P < 0.001). Both ADC min and rCBV max have significant correlations with grades of IDH-mutant  astrocytomas (rCBV max , r = 0.662 and P < 0.001; ADC min , r = 0.638 and P < 0.001).

ROC curve analysis
The results of ROC curve analysis are shown in Table 3

Discussion
In the current study, we demonstrated the usefulness of ADC min , ITSS and rCBV max vlaues for grading IDHmutant astrocytomas. The combination of ADC min , ITSS and rCBV max vlaues showed a better trend to distinguish the grade 2 and/or grade 3 IDH-mutant astrocytomas from grade 4 IDH-mutant astrocytomas. The diffusion of water molecules could be limited by tumor cells and subcellular structures. DWI could noninvasively reflect the restricted diffusion of water molecules and evaluate tumor cellularity by ADC values [5]. Previous studies revealed that the ADC values decreased significantly in high-grade astrocytic tumors [14]. Our findings showed a significant difference of ADC values between grade 2 + 3 and grade 4, which were consistent with previous studies. In addition, our study demonstrated that the ADC min value of grade 4 IDH-mutant astrocytomas was comparable with study by Tan et al. [15]. Prior studies reported the differences of ADC values among grade 2, 3 and 4 astrocytoma without considering IDH mutational status [16]. However, the current study assessed the values of ADC min in different grades of IDH-mutant astrocytomas. Moreover, we also evaluated the difference of grade 2 + 3 and grade 4 IDH-mutant astrocytomas since some researchers have reported similar prognosis betweem grade 2 and grade 3 IDH-mutant astrocytomas [17][18][19]. Consequently, we found significant difference of ADC min values between grade2 + 3 and grade 4 tumors, but no significant difference of ADCmin was abserved between grade 2 and grade 3 tumors. It has been reported that higg ADC value was correlated with favorable prognosis [20], our results will be conducive DSC-PWI could provide the vascularity and angiogenesis information of astrocytoma by rCBV values [21]. Higher grade gliomas tend to have higher rCBV values [6,22,23]. Several studies have reported that rCBV was useful in grading astrocytomas [24][25][26]. Saini et al. [10] found that CBV could be used to differentiate the grade II gliomas and grade III gliomas with high sensitivity and specificity. In contrast, our study found no significant difference of rCBV max between grade 2 and grade 3. This discrepancy may be due to the enrolled patients were different and previous investigations did not consider the IDH mutational status of gliomas. It has been shown that IDH mutation leads to 2-HG accumulation, which may decreased hypoxiainducible-factor-1 activation and inhibited angiogenesis-related signaling [27,28]. Some previous studies reported that rCBV values were significantly correlated with time to progression in patients with gliomas [20,29,30]. This may also be used to explain why grade 2 and grade 3 IDH-mutant astrocytomas have few difference in survival. We found that the rCBV max values of grade 2 and grade 3 were lower than those of grade 4, Previous studies have demonstrated that SWI was useful for grading astrocytomas [9,10]. In our study, most of patients with grade 4 IDH-mutant astrocytomas demonstrated an ITSS score of 3 and majority of patients with grade 2 showed no ITSS. Astrocytomas contained high levels of deoxyhemoglobin, due to angiogenesis and the increase of blood supply [24]. Additionally, deoxyhemoglobin generates susceptibility effect, which is related to the signal loss on SWI. The present study showed that ITSS was not detected in 13 of 29 patients with grade 3 IDH-mutant astrocytomas. However, Saini et al. [10] reported that all of the anaplastic astrocytomas could be seen with ITSS. A possible reason was the relatively small sample of anaplastic astrocytoma in their study. Our findings revealed that ITSS scores were significant difference between grade 2 + 3 and grade 4 IDHmutant astrocytomas, suggesting that ITSS may serve as a potential biomarker for grading IDH-mutant astrocytic gliomas. In the current study, ITSS socres were significant different between grade 2 and grade 3 IDH-mutant  astrocytomas while rCBV max were not. We speculated that ITSS reflected not only blood vessels but also microbleeding and necrosis of tumor, which may lead to the above results. We found a positive correlation between rCBVmax and tumor grade, which is consistent with previous study [9]. Our study found that ITSS was correlated with tumor grade; this finding was in accordance with the results from previous studies [9,31]. Our results showed that there was a negative correlation between ADC min and tumor grade,suggesting that tumor grade is closely related to the proliferation of tumor cells [32].
We found that the area under ROC for combined ADC min , rCBV max and ITSS was significantly larger than ITSS used alone for grade 3 and grade 4 IDH-mutant astrocytomas. For grade 2 + 3 and grade 4 IDH-mutant astrocytomas, AUC for combined DWI, SWI and DSC-PWI was larger than DWI, SWIand DSC-PWIalone. Our results suggested that the combination of those advanced MRI techniques could improve the diagnostic efficacy in grading IDH-mutant astrocytic gliomas.
This study has some limitations. The retrospective design and single-center sampling may lead to selection bias. Wetzel et al. [33]assessed inter-and intraobserver reproducibility for different techniques of measuring rCBV in patients with intracranial mass lesions, and they found the method that we used to obtain rCBV provides a high interobserver and intraobserver reproducibility. For this retrospective analysis,the DSC-PWI acquisition in our study is suboptimal and was not obtained according to the national consensus recommendation [34]. A multi-centered prospective investigation with standard  DSC-PWI acquisition is warranted to verify these results and ensure the reproducibility. Additionally, other genetic alterations such as CDKN2A/B and CDK4 were not assessed because of relevant molecular informations were not avaliable in our study. Advanced MRI imaging techniques combined with larger datasets on other genetic alterations should be performed in future studies.

Conclusions
DWI, SWI and DSC-PWI are useful for assessing the tumor grade in this cohort of patients with IDH-mutant astrocytomas. A combination of ADC min , ITSS and rCBV max may improve the diagnostic accuracy of IDHmutant astrocytomas grading.