Assessment of diffusion tensor imaging metrics in differentiating low-grade from high-grade gliomas

Abstract

Aim

The aim of this article is to assess diffusion tensor imaging (DTI) metrics in differentiating low-grade from high-grade gliomas.

Patients and methods

A prospective study was conducted on 35 patients with gliomas who underwent DTI. Gliomas were classified into low-grade and high-grade gliomas. The fractional anisotropy (FA), mean diffusivity (MD), linear coefficient (CL), planar coefficient (CP) and spherical coefficient (CS) of the solid tumoral part and peri-tumoral regions were calculated.

Results

There was significant difference (p = 0.001) in MD of the solid tumoral part of low-grade (1.78 ± 0.33 × 10⁻³mm²/s) and high-grade (1.16 ± 0.22 × 10⁻³mm²/s) gliomas. The selection of 1.42 × 10⁻³mm²/s as a cutoff value of MD of the tumoral part was used to differentiate low-grade and high-grade gliomas; the best results were obtained with area under the curve (AUC) of 0.957 and accuracy of 91.4%. There was a significant difference in FA, MD, CP and CS of peri-tumoral regions of both groups with p values of 0.006, 0.042, 0.030 and 0.037, respectively. The cutoff values of MD, FA, CS and CP of the peri-tumoral region used to differentiate low-grade from high-grade gliomas were 1.24, 0.315, 0.726 and 0.321 with AUC of 0.694, 0.773, 0.734 and 0.724 and accuracy of 68.6%, 80.0%, 74.3% and 74.3%, respectively. The combined MD of the solid tumoral part and FA of the peri-tumoral region used to differentiate low-grade from high-grade gliomas revealed AUC of 0.974 and accuracy of 88.6%.

Conclusion

We conclude that the combination of MD of the solid tumoral part and FA of the peri-tumoral region is a noninvasive method to differentiate low-grade from high-grade gliomas.

Introduction

Gliomas are the most common primary brain tumors, representing 80% of primary brain tumors. According to World Health Organization (WHO) classification of brain tumors, high-grade gliomas constitute grade III and grade IV glial pathologies, and high-grade astrocytomas are the most common primary brain malignancies in adults. On the other hand, low-grade gliomas consist of a heterogeneous group of glial tumors, representing 15% brain tumors in adults. Grading of gliomas is crucial for the appropriate therapeutic strategies, clinical outcome^1–5 and in assessment of prognosis as survival rates differ significantly between the two groups.^3,6 Histopathology is the gold standard for glioma grading, but it carries a sampling error because of a limited number of biopsy samples.^7,8 Routine and advanced magnetic resonance (MR) imaging sequences are used for grading of gliomas, but their results are overlapping. Several studies have shown that approximately 20% of low-grade gliomas may show regions of contrast enhancement and that a third of non-enhancing gliomas are pathologically found to be high grade. There is inconsistency in perfusion imaging parameters due to different MR imaging protocols, pharmacokinetic models, or perfusion-analysis software. The integrity of metabolites measured by proton MR spectroscopy may be affected by heterogeneity of the tumor and the technical parameters of MR devices.^8–15

Diffusion tensor imaging (DTI) takes advantage of the diffusion of water in brain tissue within three main directions, which is decreased perpendicularly to the myelin sheaths and cell membranes of white-matter axons.^3–8 There are different metrics of DTI used as a quantitate analysis. The most common metrics of DTI used are mean diffusivity (MD) and fractional anisotropy (FA). MD reveals the rate of water molecules’ diffusional motion as the tumor cellularity is the main target of histologic tumor classification with DTI. There is an inverse relationship between the cellularity and the MD value of gliomas.^14–17 The FA expresses the orientation of the tissue microstructure and is related to the structural orientation of different tissues.^14–18 The other three basic metrics that expresses the shape of the DT are: linear anisotropy coefficient (CL), in which diffusion is mainly along the direction corresponding to the largest eigenvalue; planar anisotropy coefficient (CP), in which diffusion is restricted to the plane spanned by the two eigenvectors with the two largest eigenvalues; and spherical anisotropy coefficient (CS), which indicates the isotropic diffusion. Each anisotropy coefficient shows a unique measure in the different areas of the white matter. These differences are due to the impact of the linear, planar, and spherical shape components of the DTI.^19–26 Few studies discuss the role of diffusion tensor MR imaging in grading of gliomas.^27–31 The novelty and uniqueness of this study is assessing the combined DT indices of tumoral and peri-tumoral regions in differentiating low-grade from high-grade gliomas.

Material and methods

Patients

This prospective study was approved by the local ethics committee and informed consents were obtained from all patients prior to the examination. The study included 38 patients provisionally diagnosed to have untreated gliomas based on conventional MR imaging. We excluded three patients from the study because two patients proved to be metastasic and one patient had lymphoma. The final patients included in this study were 35 patients (22 male and 13 female). Their ages ranged from 52 to 72 years (mean age, 63 years). The final diagnosis was based on histopathological examinations. Gliomas were classified into low-grade gliomas (WHO grade I and II) in 16 patients and high-grade gliomas (WHO grade III and IV) in 19 patients. The histopathological subtypes of WHO grade I were diffuse astrocytoma (n = 5) and oligoastrocytoma (n = 1), grade II were astrocytomas (n = 8) and oligodendrogliomas (n = 2), grade III was anaplastic astrocytomas (n = 6) and grade IV was glioblastoma multiforms (n = 13).

MR imaging

The MR imaging of the brain was performed using a 1.5 Tesla scanner (Ingenia, Philips) using Stream Head Neck 20 channel coil. T1-weighted (repetition time (TR)/echo time (TE) = 600/25 ms), T2-weighted (TR/TE = 6000/90 ms) and fluid-attenuated inversion recovery (FLAIR) (TR/TE/inversion time (TI) = 10,000/115/2700 ms) sequences were obtained. The scanning parameters were matrix of 80 × 80, field of view (FOV) of 250 × 170 mm² and slice thickness of 5 mm. Post-contrast T1-weighted images were obtained after intravenous administration of gadoterate meglumine, 0.5 ml/kg (0.1 mmol/kg) body weight with maximum dose of 10 ml using a 20–22 G venous cannula with a flow rate of 2 ml/second.

DTI

DTI was obtained using a single-shot echo-planar imaging sequence (TR/TE 3200/90 ms) with parallel imaging (SENSitivity Encoding (SENSE) reduction factor P 2). Diffusion gradients were applied along 32 axes, using a b value of 0 and 1000 s/mm². FOV was 250 × 170 mm², data matrix of 92 × 88 with voxel dimensions of 2.43 × 2.54 × 2.5 mm³. Forty-eight slices were obtained, with a thickness of 2.5 mm, with no gap and the total scan duration was seven to eight minutes.

Post-processing

Image analysis was performed by one radiologist (EA) expert in neuroradiology for 10 years who was blinded to the histopathological results. The digital imaging and communications in medicine (DICOM) images were transferred to a workstation (extended MR Workspace 2.6.3.5, Philips Medical Systems Nederland B.V). The images were loaded to DTI software provided by the vendor. Automated registration of the DTI data was performed to eliminate eddy current artifacts. Co-registration of the FA maps to contrast T1-weighted images was performed for accurate placement of regions of interest (ROIs). A well-defined circular ROI (0.4 cm²) was placed in the solid tumoral part and another ROI was placed in the peri-tumoral region (Figure 1). The ROI was placed in the tumoral part at the most enhanced solid region seen at contrast T1-weighted images, avoiding the cystic or necrotic region. The peri-tumoral ROI was located in the immediate peri-tumoral region that appears as a bright T2 hyperintensity with a fingerlike projection and of low signal on T1-weighted images. If a peri-tumoral high T2 signal was absent, the ROI was placed just immediately adjacent to the tumor margin. Placement of ROIs within the normal gray matter structures was avoided to avoid the inherent DTI differences between white matter and gray matter The eigenvalues (primary, secondary and tertiary) of these regions were estimated. FA, MD, CL, CP and CS were calculated using the following equations.^23,24

$$\text{MD}=(\lambda 1+\lambda 2+\lambda 3)/3$$

$$\text{FA}=32(\lambda 1-\overline{\lambda })2+(\lambda 2-\overline{\lambda })2+(\lambda 3-\overline{\lambda })2\lambda 12+\lambda 22+\lambda 32$$

Figure 1.

Grade II astrocytoma. (a) Axial T2-weighted image shows an intra-axial space-occupying lesion of abnormal signal intensity seen in the right thalamic region with hyperintensity in the peri-tumoral region. (b) Axial contrast T1-weighted image shows multiple enhanced nodular areas within the lesion in T2 images. (c) Color FA map co-registered with contrast T1-weighted image shows ROI (1) located within the enhanced solid tumoral part and ROI (2) located in the peri-tumoral region. The MD of the tumoral part is 1.88. The MD, FA, CL, CS and CP of the peri-tumoral part are 1.21, 0.37, 0.41 and 0.78, respectively. (d) MD map shows relative iso-signal of the tumor parts. (e) FA map showing low signal of the solid tumor part. FA: fractional anisotropy; ROI: region of interest; MD: mean diffusivity; CL: linear coefficient; CS: spherical coefficient; CP: planar coefficient.

CL = (λ₁ – λ₂)/(λ₁ + λ₂ + λ₃)

CP = 2 (λ₂ – λ₃)/(λ₁ + λ₂ + λ₃)

CS = 3 λ₃/(λ₁ + λ₂ + λ₃) where $\overline{\lambda }$ denotes the mean of the three eigenvalues. The CL, CP, and CS values lie in the range from 0 to 1 and the sum of these three metrics is equal to 1.²²

Statistical analysis

Statistical analyses were carried out using Statistical Package for Social Sciences version 20 (SPSS, Chicago, IL, USA). Quantitative data were presented as mean ± standard deviation (SD), median and range while qualitative data were presented as number (n) and percentage (%). Shapiro–Wilk test results, histograms and boxplots were used to determine the normality of the parameters (FA, MD, CL, CP and CS). Normally distributed data were compared between the two major groups using independent samples t test and between each WHO group. Data that violated the normality assumptions were compared using the Mann–Whitney test. Probability (p) values <0.05 were considered statistically significant. The receiver operating characteristic (ROC) curves of different matrices of the tumoral and peri-tumoral regions were obtained by calculating the area under the curve (AUC). The optimum cutoff values of different matrices of the tumoral and peri-tumoral regions with highest accuracy selected to differentiate low-grade from high-grade gliomas were calculation of sensitivity and specificity. The multivariate linear regression analysis was conducted for variables that reached a p value of 0.05 for best combination of DTI metrics in the tumoral and peri-tumoral parts used to differentiate low-grade from high-grade gliomas.

Results

Table 1 shows the mean, standard deviations, minimum and maximum DTI metrics of the tumoral part and peri-tumoral region in low-grade and high-grade gliomas. Table 2 shows the cutoff values of DT metrics in the tumoral and peri-tumoral regions of gliomas with area under the curve, accuracy, sensitivity, specificity, specificity, positive predictive value and negative predictive values.

Table 1.

Mean, standard deviations, minimum and maximum of DTI metrics in the solid tumoral part and peri-tumoral region of low-grade and high-grade gliomas.

	Mean ± SD (min/max)
Metric	Low-grade gliomas (n = 16)	High-grade gliomas (N = 19)	p value
Tumor part
MD	1.78 ± 0.33 (1.28–2.54)	1.16 ± 0.22 (0.05–1.47)	0.001
FA	0.16 ± 0.12 (0.05–0.49)	0.12 ± 0.10 (0.61–0.49)	0.272
CL	0.07 ± 0.07 (0.01–0.30)	0.04 ± 0.04 (0.01–0.15)	0.142
CS	0.82 ± 0.96 (0.62–0.95)	0.87 ± 0.08 (0.59–0.98)	0.133
CP	0.25 ± 0.16 (0.06–0.67)	0.17 ± 0.12 (0.03–0.55)	0.188
Peri-tumoral region
MD	1.11 ± 0.36 (0.62–2.06)	1.39 ± 0.40 (0.74–2.06)	0.042
FA	0.35 ± 0.17 (0.10–0.71)	0.21 ± 0.10 (0.03–0.45)	0.006
CL	0.13 ± 0.11 (0.02–0.47)	0.07 ± 0.25 (0.01–0.19)	0.056
CS	0.69 ± 0.12 (0.47–0.92)	0.78 ± 0.11 (0.42–0.93)	0.037
CP	0.43 ± 0.22 (0.12–0.96)	0.28 ± 0.16 (0.08–0.74)	0.030

MD: mean diffusivity; FA: fractional anisotropy; CL: linear coefficient; CS: spherical coefficient; CP: planar coefficient.

Table 2.

The cutoff values of diffusion tensor (DT) matrices in the tumoral part and peri-tumoral region of gliomas with area under the curve, sensitivity, specificity, accuracy, positive predictive value (PPV) and negative predictive value (NPV).

Parameter	Cutoff	AUC	Sensitivity	Specificity	Accuracy	PPV	NPV
Tumor MD	1.42	0.957	94.7	87.5	91.4	90	93.3
Peri-tumoral MD	1.24	0.694	57.9	81.3	68.6	78.6	61.9
Peri-tumoral FA	0.31	0.773	89.5	68.8	80.0	77.3	84.6
Peri-tumoral CP	0.32	0.724	68.4	81.3	74.3	81.3	68.4
Peri-tumoral CS	0.72	0.734	78.9	68.8	74.3	75.0	73.3

AUC: area under the curve; MD: mean diffusivity; FA: fractional anisotropy; CL: linear coefficient; CP: planar coefficient; CS: spherical coefficient.

The mean MD (10⁻³mm²/s) values of solid tumoral part of high-grade gliomas was 1.16 ± 0.22 (Figure 1) and of low-grade gliomas was 1.78 ± 0.33 (Figure 2) with significant difference (p = 0.001). At ROC curve, the AUC of MD of the tumoral region used to differentiate low-grade from high-grade gliomas was 0.957. The selection of 1.42 as a cutoff value of MD of the tumoral region to differentiate between the low-grade and high-grade gliomas revealed accuracy of 91.4%, sensitivity of 94.7% and specificity of 87.5% (Figure 3). There was insignificant difference of other parameters in the solid tumor parts between low-grade and high-grade gliomas (Table 1).

Figure 2.

Glioblastoma multiformis. (a) Axial T2-weighted image shows an intra-axial space-occupying lesion is seen in the genu of the corpus callosum crossing the midline with hyperintensity in the peri-tumoral region. (b) Axial contrast T1-weighted image shows heterogeneous enhancement of the tumor. (c) Color FA map co-registered with contrast T1-weighted image shows ROI (1) located in the enhanced solid tumoral part and ROI (2) located in the peri-tumoral region. The MD of the tumoral part is 1.22. The MD, FA, CL, CS and CP of the peri-tumoral region are 1.72, 0.23, 0.12, and 0.38, respectively. (d) MD map shows an iso-signal with relative low-signal linear areas of the solid tumoral parts. (e) FA map shows heterogeneous mixed mainly low signal of the solid tumoral part. FA: fractional anisotropy; ROI: region of interest; MD: mean diffusivity; CL: linear coefficient; CS: spherical coefficient; CP: planar coefficient.

Figure 3.

ROC curve for different metrics of the solid tumoral part. The selection of 1.42 as a cutoff value of MD of the solid tumoral part was used to differentiate low-grade from high-grade gliomas; the best results were obtained with AUC of 0.957, accuracy of 91.4%, sensitivity of 94.7% and specificity of 87.5%. ROC: receiver operating characteristic; MD: mean diffusivity; AUC: area under the curve.

There was significant difference (p = 0.042) in MD of the peri-tumoral region between low-grade gliomas (1.11 ± 0.36) and high-grade gliomas (1.39 ± 0.40). There was significant difference (p = 0.006) in FA of the peri-tumoral region between low-grade gliomas (0.35 ± 0.17) and high-grade gliomas (0.21 ± 0.10). There was a significant difference (p = 0.037) in the CS of the peri-tumoral region between low-grade gliomas (0.6 ± 0.12) and high-grade gliomas (0.78 ± 0.11). There was an insignificant difference (p = 0.056) of CL of the peri-tumoral region between low-grade and high-grade gliomas. There was a significant difference (p = 0.030) in the CP of the peri-tumoral region between low-grade gliomas (0.43 ± 0.22) and high-grade gliomas (0.28 ± 0.16). At ROC curve, the AUC of MD, FA, CS and CP of the peri-tumoral region used to differentiate low-grade from high-grade gliomas were 0.694, 0.773, 0.734 and 0.724 respectively. The cutoff of MD, FA, CS and CP of the peri-tumoral region used to differentiate low-grade from high-grade gliomas were 1.24, 0.315, 0.726 and 0.321 with accuracy of 68.6%, 80.0%, 74.3% and 74.3% respectively (Table 2) (Figure 4). Combination of the tumoral MD and the peri-tumoral FA for differentiation of low-grade from high-grade gliomas revealed AUC of 0.974, accuracy of 88.6%, sensitivity of 89.5% and specificity of 87.5%.

Figure 4.

ROC curve for different metrics of the peri-tumoral region. The cutoff value of MD, FA, CS and CP of the peri-tumoral region used to differentiate low-grade gliomas from high-grade gliomas are 1.24, 0.315, 0.726 and 0.321 with areas under the curve of 0.694, 0.773, 0.734 and 0.724 and accuracy of 68.6%, 80.0%, 74.3% and 74.3%, respectively. ROC: receiver operating characteristic; MD: mean diffusivity; FA: fractional anisotropy; CL: linear coefficient; CS: spherical coefficient; CP: planar coefficient .

Discussion

In this study, there is significant difference between low-grade and high-grade gliomas using MD values in solid tumoral parts. The MD is equal to the apparent diffusion coefficient (ADC), which is a sensitive metric of the cellularity within solid tumor parts, so it shows significant decrease in high-grade gliomas.^8,10,25,26 Another study reported there is significant difference with high CP value in the solid tumoral parts of high-grade gliomas and they explain that difference with higher cellularity; however, the CP values also are seen increasing in crossing fibers and in regions of edema.³⁰ Another study added that higher anisotropic values within the solid parts of high-grade compared to low-grade gliomas may be attributed to the high cellularity of high-grade gliomas.³ Another study added that ADC, RD, and AD are useful DTI metrics for the differentiation between low-grade and high-grade gliomas with a diagnostic accuracy of more than 90%, and can be used as noninvasive reliable biomarkers in the grading of gliomas.⁸

The peri-tumoral region is a crucial site of differentiation between tumor types as regards to DTI metrics because of structural changes and edema grade differences between the low and high glioma grades. In peri-tumoral regions, the MD and anisotropic metrics show significant difference between low-grade and high-grade gliomas; the immediate peri-tumoral region of low-grade gliomas usually shows no or mild vasogenic edema compared to evident edema in the peri-tumoral region of the high-grade gliomas mixed with tumoral infiltrations that indicate increase cellularity. In peri-tumoral regions, there was as significant reduction of the FA values of high-grade gliomas compared to low-grade gliomas that may be attributed to tubular anisotropy changes.^3,32 Another study added that FA is a useful diagnostic marker that could be used in distinguishing the low-grade from high-grade gliomas.¹⁷

In this study, CL didn’t show statistically a significant difference in peri-tumoral regions, indicating similar levels of single-orientated fibers. One study reported that there is significantly higher CL in the peri-tumoral region of high-grade gliomas. This is attributed to highly invasive tumor cells within high-grade glioma margins leading to decreased linearity in nearby space.³³ Another study added that there is significant difference in the CS values between low-grade and high-grade gliomas in the peri-tumoral region.²¹

In this study, the combination of MD of the solid tumoral part and FA of the peri-tumoral region of gliomas revealed the highest accuracy in differentiating low-grade from high-grade gliomas. This is attributed to MD of the solid tumoral part representing the degree of cellularity of the gliomas with high cellularity in high-grade gliomas and FA of the peri-tumoral region expressing the orientation and infiltration of the peri-tumoral regions with tumor cells and associated edema.

There are a few limitations of this study. First, the small number of patients limits the statistical results of the cases. We recommend further studies including a large number of gliomas. Second, this study was performed using 1.5 Tesla. Further studies at higher 3 Telsa³⁴ with application of other methods of fiber tracking are recommended to improve the results. Third, this study used DTI metrics. Further studies with application of advanced post-processing such as diffusion kurtosis, intravoxel incoherent motion diffusion MR imaging sequence with several b values and multi-parametric imaging with dynamic susceptibly perfusion-weighted MR imaging and MR spectroscopy^35–41 will improve grading of gliomas and correlation with prognostic parameters.

Conclusion

We conclude that the combination of MD of the solid tumoral part and FA of the peri-tumoral region is a noninvasive method in differentiating between low-grade and high-grade gliomas.

Funding

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Conflict of interest

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.