Evaluation of gliomas peritumoral diffusion and prediction of IDH1 mutation by IVIM-DWI

Patients characteristics

A total of 78 patients (males: 46, female: 32, age range: 13-76, average age of males: 49±12, average age of females: 50±15) were enrolled in this research (Figure 1). Totally, 26 patients were diagnosed as WHO grade I and grade II (males:16, females: 10, IDH1+: 16, IDH1-: 10, age range: 14-65), 52 cases were diagnosed as WHO grade III and grade IV (males:30, females: 22, IDH1+: 11, IDH1-: 41, age range: 13-76).

Interobserver agreement

Good interobserver agreement in terms of D* (ICC: 0.978, 95% CI: 0.916-0.948), D (ICC: 0.892, 95% CI: 0.826-0.922), f (ICC: 0.934, 95% CI: 0.864-0.896), and CBF (ICC: 0.884, 95% CI: 0.916-0.935) were observed. Meanwhile, the coefficients of variations of IVIM-derived parameters ranged from 6.5-15.8%.

Comparison of parameter values between group HGG and LGG

Remarkable differences were observed in some parameters of IVIM perfusion between group HGG and LGG (Table 1 and Figure 2). The values of CBF1 (141.69±72.77 vs 75.93±66.81 ml/min/100g), CBF3 (53.13±10.05 vs 42.94±11.78 ml/min/100g), D*1 (57.69±35.49 vs 49.9±30.38 ×10^-3mm²/sec), and age (53.79±13.2 vs 41.92±13.93 years) in group HHG were significantly higher than that of in group LGG. The D1 value related with diffusion was remarkably lower in group HGG than group LGG (0.506±0.34 vs 0.715±0.325 ×10^-3mm²/sec). The values of CBF1 and CBF3 were normalized to the value of CBF2 representing the mirror side normal areas, then the values of rCBF1-2 and rCBF3-2 were achieved. The values of rCBF1-2 and rCBF3-2 in group HGG were markedly higher than group LGG (Table 1). Therefore, remarkable difference could be observed these parameters such as CBF1, CBF3, rCBF1-2, rCBF3-2, D1, and D*1 between group HGG and LGG using IVIM-DWI. No significant difference was found in terms of gender between group HGG and LGG.

In addition, the differences of parameters derived using IVIM in tumor parenchyma, peritumoral, and mirror side normal areas were also analyzed (Table 1). In the group HGG, significant higher values of CBF and D in tumor parenchyma areas were found compared with peritumoral, and mirror side normal areas. Meanwhile, remarkable difference in terms of D* was observed between tumor areas and other regions. Similar findings were observed in the group LGG (Table 1). These results indicate that the difference of gliomas grade and tumor location could be effectively achieved using IVIM-DWI.

Comparison of parameter values between group IDH1^mut and IDH1^wt

The gene mutation status of IDH1 in all patients were measured using immunohistochemical method firstly. Then, IVIM derived parameters between group IDH1+ and IDH1- were analyzed (Table 2). The values of CBF1 (100.127±53.092 vs 150.172±35.020 ml/min/100g), CBF3 (46.197±13.134 vs 51.603±10.408 ml/min/100g), rCBF1-2 (2.049±1.879 vs 3.116±1.785 ml/min/100g), rCBF3-2 (1.038±0.567 vs 1.579±0.566 ml/min/100g), D*1 (50.9±16 vs 59.65±13 ×10^-3mm²/sec), and age (45.22±10.297 vs 52.27±15.837 years) in group IDH1^mut were significantly lower than that of in group IDH1^wt (Table 2). However, the levels of D1 (0.635±0.206 vs 0.635±0.206 ×10^-3mm²/sec) and f1 (0.449±0.206 vs 0.356±0.156) were remarkably higher in the group IDH1^mut than group IDH1^wt. No significant difference was found in terms of gender between group IDH1^mut and IDH1^wt.

Meanwhile, the differences of parameters in tumor parenchyma, peritumoral, and mirror side normal areas were also analyzed in both group IDH1^mut and IDH1^wt. Significant differences in values of CBF, D, and D* could be found between tumor areas and other regions (peritumoral and mirror side normal areas). Therefore, IVIM-DWI might be a promising method to predict IDH1 gene mutation and evaluation of gliomas peritumoral diffusion.

In addition, the cases distribution in the group HGG, LGG, IDH1+, IDH1^wt were analyzed (Table 3). Totally, 41 cases IDH1^wt patients were found in the group HGG, only 10 cases IDH1^wt patients in the group LGG. Meanwhile, 11 and 16 cases IDH1^mut patients were found in the group HGG and LGG, respectively (Table 3). The distribution difference was significant indicating that IDH1 wide type gliomas patients were more likely to be diagnosed as HGG.

ROC analysis of IVIM-DWI-derived parameters in IDH1 mutation and gliomas grade

The AUC values of parameters derived from IVIM-DWI were calculated between group HGG and LGG (Table 4 and Figure 3). The AUC value of rCBF1-2 was the highest, 0.861 (P<0.001) with specificity 94.1% and sensitivity 92.3%. Relative high AUC values were also observed in CBF1 (0.818), CBF3 (0.774), rCBF3-2 (0.773), and D1 (0.816). High specificity and sensitivity were observed in CBF1 (specificity: 82.4%, sensitivity: 88.5%) and D1 (specificity: 86.9%, sensitivity: 81.2%) (Table 4 and Figure 3).

The diagnosing efficiency using IDH1^mut and IDH1^wt was also calculated by analyzing parameters derived from IVIM-DWI (Table 5 and Figure 4). The CBF1 presented the highest ADU value, 0.892 (P=0.003), with specificity 82% and sensitivity 82.4%. Relative high ADU values for D1 (0.873, P<0.001) and CBF3 (0.831, P=0.002) were also found. The D1 (specificity: 85.5%, sensitivity: 84.4%) and CBF3 (specificity: 83.9%, sensitivity: 81.1%) showed good diagnosing efficiency (Table 5 and Figure 4).

Correlation analysis of IVIM-DWI-derived parameters and CBF1

The correlation relationship between CBF1 other IVIM-DWI-derived parameters were analyzed using Pearson method (Figure 5). rCBF1-2 (Figure 5A) had a remarkably positive correlation with CBF1 (r=0.852, p<0.001), and f1 (Figure 5B) showed a markedly negative correlation with CBF1 (r= -0.306, p=0.007). However, other parameters including CBF3 (Figure 5C) and D1 (Figure 5D) did not present remarkable correlation with CBF1.

Patients

A total of 98 glioma patients diagnosed by MRI were enrolled from Oct 2018 to Oct 2020 in our hospital. The exclusion and inclusion criteria were listed in the Figure 1. After checking the examining materials, 20 patients were excluded due to incomplete pathological or imaging details. Finally, 78 gliomas patients confirmed by pathology data and IDH immunohistochemical results were included. Among them, 26 (16 males and 10 females) low grade gliomas (LGG, WHO grade I and grade II) cases were enrolled in this study. The age ranged from 14 to 65 years old (median age: 42 years). 10 cases were identified with IDH1 wild type and 16 cases were mutation type. 52 (30 males and 22 females) high grade gliomas (HGG, WHO grade III and grade IV) patients, aged 13-76 years (median age 55 years) were enrolled in this research. Among them, 41 cases were IDH1 wild type and 11 cases were mutant type.

All protocols in this study was approved by the Ethical Committee of The Second Affiliated Hospital of Nanchang University (Approval number: 2018-08-002), and was in accordance with the Helsinki declaration. Written informed consent was obtained from patients.

Conventional MR imaging

Conventional MR imaging was conducted using a 3.0T MR scanner (Discovery MR 750; GE, Milwaukee, USA). The parameters were set as follows: T1-weighted (T1W) imaging, TI: 800 ms, TE: 25 ms, TR: 1900 ms. T2-weighted (T2W) imaging, TE: 100 ms, TR: 6600 ms, scanning layers: 24, layer spacing: 1 mm, layer thickness: 3 mm, matrix: 298×244, field of view: 240×240 mm.

3D pCASL

Firstly, the contrast agents were injected, then 3D pCASL was conducted. 3D FSE imaging sequences, background inhibition, and pseudo-continuous labeling were used in 3D pCASL. The parameters were set as follows: acquisition time: 5 min 16 s, NEX: 4, slices number: 40, thickness: 6 mm, bandwidth: ±60 kHz, field of view: 240×240 mm, post label delay: 1480 ms, TR/TE: 4686/8.

IVIM-DWI

Firstly, the contrast agents were injected, then IVIM-DWI was conducted. A single-shot echo-planar sequence in the axial plane was applied for application of IVIM-DWI. The parameters were set as follows: matrix: 180×180, field of view: 240×240 mm, gap: 1 mm, thickness: 4 mm, TR/TE: 4500/6. 12 b values (0, 20, 40, 60, 90, 110, 150, 300, 600, 1200, 1800, and 2800 s/mm²) were applied during three orthogonal directions. The total acquisition time was 5 min 48 s.

Imaging analysis

Images were moved to workstation (Advantage workstation, GE, Milwaukee, USA). Then, perfusion fraction (f), true diffusion coefficient (D), and perfusion-related diffusion coefficient (D*) were calculated according to the equation 1.

${\text{S}}_{(\text{b})}{\text{/S}}_{(0)}=\text{f}\cdot {\text{exp}}^{(-\text{b}\times \text{D}*)}+\left(1-\text{f}\right){\text{ exp}}^{(-\text{b}\times \text{D})}$ (1)

Fast moving compartment was represented by D*. The volume fraction of random microcirculation of blood over the total incoherent signal in each voxel was represented by f. The slow-moving compartment was represented by D. S_(b) and S₍₀₎ represent the signal intensity in the presence and absence of diffusion sensitization, respectively. ADC was calculated with b values according to the equation 2. CBF maps were calculated as described [20].

${\text{S}}_{(\text{b})}{\text{/S}}_{(0)}={\text{exp}}^{(-\text{b}\times \text{D})}$ (2)

Regions of interests (ROIs)

ROIs were conducted by the 3 experienced neuroradiologists. The maximum CBF values (CBF) of the tumor parenchyma areas (CBF1), peritumoral areas (CBF3) and mirror side normal areas (CBF2) were calculated. Then, relative CBF value (rCBF) was calculated as follows: rCBF = tumor side CBF value/mirror side CBF value. rCBF1-2= CBF1/CBF2, rCBF3-2= CBF3/CBF2.

Contrast-enhanced T2-FLAIR and T1WI was used to select the most obvious enhancement level or the largest level of tumor for analysis. The minimum D, maximum D * value, and maximum f value in tumor, peritumoral and normal areas were obtained on IVIM pseudo color image. D1, D*1, f1 were the IVIM parameters in the tumor parenchyma areas. D2, D*2, f2 were the IVIM parameters in mirror side normal areas. D3, D*3, f3 were the IVIM parameters in the peritumoral areas. Three ROIs (range 40-60 mm²) were drawn manually. Cerebrospinal fluid-filled, calcification, hemorrhagic, necrotic, and cystic areas should be avoided. The measurement was repeated twice with an interval of 1-2 weeks.

Data analysis

Data analysis was performed using 21.0 SPSS (IBM, USA). Data was presented as mean ± standard deviation. The data between group HGG and LGG, group IDH1+ and IDH1- were analyzed using Student's t-test. Receiver operating characteristic curve (ROC) was performed to get the area under the receiver operating characteristic curve (AUC). The IVIM-DWI-derived parameters were analyzed using specificity, sensitivity, and predicted cutoff value from ROC. Intraclass coefficient (ICC) with the 95% interval confidence was set in this study. p<0.05 was believed to be statistical significance.

Data availability

The data in the current study are available from the corresponding author on reasonable request.