The value of intravoxel incoherent motion imaging in predicting the survival of patients with astrocytoma

Abstract

Background

The evaluation of the prognosis of gliomas may have great value in individualized treatment.

Purpose

To evaluate the value of intravoxel incoherent motion (IVIM) in predicting the survival of patients with astrocytoma and comparing it to apparent diffusion coefficients (ADC).

Material and Methods

Sixty patients with pathologically confirmed cerebral astrocytomas underwent IVIM scans before any treatment was performed. Patients were divided into death group and survival group according to a two-year follow-up. ADC and quantitative parameters of IVIM including D, D*, and f were measured. Independent sample t test was used to compare the two groups of parameters. The accuracy of each parameter for two-year survival rate was analyzed by receiver operating characteristic (ROC) curve and Kaplan–Meier survival curves. The correlation between quantitative parameters and survival days was analyzed by Pearson correlation analysis.

Results

The ADC, D*, and f values were statistically significant different between the death and the survival groups (P < 0.05). The AUC of the ADC, D*, and f were 0.811, 0.858, and 0.892, respectively. The ADC cut-off value of 0.668 × 10⁻³ mm²/s corresponded to 82.6% sensitivity and 73% specificity. The D* cut-off value of 3.913 × 10⁻³ mm²/s corresponded to 78.4% sensitivity and 87% specificity. The f cut-off value of 0.487 corresponded to 83.8% sensitivity and 87% specificity. Significant log rank test was performed for each parameter to predict overall survival (P < 0.05). There was a correlation between ADC (r = 0.625, P = 0.023), D* (r = −0.655, P = 0.012), f (r = −0.725, P = 0.000) and survival days.

Conclusion

The D* and f values demonstrated great potential in predicting the two-year survival rate for patients with astrocytoma.

Keywords

Intravoxel incoherent motion imaging cerebral astrocytoma survival

Introduction

Astrocytoma is the most common neurogliocytoma, which is characterized by invasive growth with a high rate of mortality and recurrence (1,2). Due to the continuous progress in the treatment of gliomas, the prognosis of gliomas, especially high-grade gliomas, is still poor, which with an average survival period of 50 months in patients with grade III gliomas. The average survival time of patients with grade IV glioma is only 10 months (3,4). Early prediction of survival in patients with astrocytoma is of great clinical significance for guiding the treatment strategy and evaluating the efficacy of medicine and prognosis of patients. The current consensus among domestic and foreign scholars is the impact of pathological grading on the prognosis (5,6). Much literature has shown that IVIM can be used to classify gliomas and its quantitative parameters are correlated with glioma grades (7 –9). In theory, we can directly measure the survival of gliomas using the quantitative parameters of IVIM without the need for classification. Currently, many studies on prognosis survival are performed for each grade of glioma, but gliomas are only divided into four grades, and the prognosis is quite different even though patients may have the same pathological grade (10). The quantitative values of intravoxel incoherent motion imaging (IVIM) are more consistent and accurate. From this perspective, it is more appropriate to use IVIM to directly predict survival.

IVIM simplifies the movements of water molecules in vivo into two types: slow diffusion motion, including intracellular and extracellular diffusion movement and transmembrane movement; and fast diffusion movements, including movements related to microcirculatory vascular perfusion. The formula of IVIM also called bi-exponential model is: $\frac{S_{b}}{S_{0}} = (1 - f) e^{- b D} + f e^{- b D^{*}}$ , where S_b represents the signal intensity when b ≠ 0, and S₀ represents the signal intensity when b = 0. The D represents the slow diffusion coefficient, i.e. the simple diffusion of water molecules. D* represents the pseudo-diffusion coefficient, i.e. the diffusion coefficient related to capillary perfusion. The f is the perfusion fraction, presenting the proportion of microcirculatory perfusion to diffusion in total voxel diffusion. The IVIM model based on multi-b-value diffusion imaging can better capture the diffusion and perfusion information in complex in vivo tissues.

The aim of the present study was to evaluate the value of the parameters of IVIM when assessing the survival of patients with astrocytoma.

Material and Methods

General information

The present study was approved by the Ethics Committee at the First Hospital of Shanxi Medical University and all patients provided informed consent. Seventy-eight patients with astrocytoma confirmed histologically at the First Hospital of Shanxi Medical University from September 2014 to August 2016 were continuously collected. All patients underwent magnetic resonance imaging (MRI) scans in the First Hospital of Shanxi Medical University within one month before surgery. Inclusion criteria were as follows: patients with astrocytoma were confirmed by histopathology after the surgery; the clinical data and imaging data of patients who received MRI scans in the First Hospital of Shanxi Medical University were available; telephone follow-up was conducted every three months after the surgery and the follow-up was not interrupted until the end of August 2018 (≥2 years of follow-up); and the patients’ treatment plan meets the guidelines for diagnosis and treatment of central nervous system gliomas in PR China. The guideline is as follows: temozolomide (TMZ) was synchronized with radiotherapy and combined with TMZ adjuvant chemotherapy for at least six cycles after surgery. Specifically, all patients started radiotherapy 2–4 weeks after surgery. The radiation dose of high-grade astrocytoma was 60 Gy and the low-grade astrocytoma was 45–54 Gy. The dose of TMZ during synchronous chemotherapy was 75 mg/m²/d for 42 days and the dose of TMZ during adjuvant chemotherapy was 150–200 mg/m²/days for six cycles every 28 days. Exclusion criteria were as follows: patients who received preoperative treatment such as punctures, gamma knife, and radiotherapy chemotherapy; and patients with incomplete tumor removal.

Data collection

The images were acquired on a 3.0-T GE Signa HDxt MRI scanner (GE, Milwaukee, WI, USA) equipped with an eight-channel head and neck combined coil. All patients underwent T1-weighted (T1W) imaging, T2-weighted (T2W) imaging, T2W fluid attenuated inversion recovery (FLAIR), IVIM and T1W imaging enhancement scans before surgery. Parameters of T1W imaging sequences were: repetition time (TR) = 220 ms; echo time (TE) = 2.4 ms; field of view (FOV) = 24 × 24 cm²; and layer thickness = 6.0 mm. Parameters of T2W imaging sequence were: TR = 3570 ms; TE = 175 ms; and FOV = 22 × 22 cm². Parameters of T2W-FLAIR sequences were: TR = 8024 ms; TE = 126.8 ms; FOV = 22 × 22 cm²; layer thickness = 6.0 mm.

IVIM sequence with a single-shot echo planar imaging was performed with the following parameters: TR = 3000 s; TE = 115.5 s; FOV = 24 × 24 cm²; matrix = 128 × 128; slice thickness = 6 mm; slice spacing = 1.0 mm; and 13 b-values (0, 20, 40, 80, 100, 150, 200, 400, 800, 1200, 1800, 2500, and 3000 s/mm²) in three orthogonal directions with number of excitations 2, 2, 2, 2, 2, 2, 2, 2, 2, 3, 3, 4, and 4, respectively.

Data postprocessing

The original IVIM images were sent to GE Advanced Workstation 4.6. MADC-DWI software in GE Functool 9.4.05a was used for IVIM analysis. The cut-off of the low values was set as 200 s/mm². The slow diffusion coefficient (D) was calculated by fitting the higher b value (>200) with a mono-exponential model, and the fast diffusion coefficient (D*) and perfusion fraction (f) were generated using the lower b value (<200) with a mono-exponential model. The apparent diffusion coefficient (ADC) was obtained using a single exponential model (b = 0, 1000). Regions of interest (ROI) were placed in the tumor parenchyma (corresponding to the tumor enhancement area), avoiding cystic degeneration, bleeding, necrosis, macrovascular, edema, calcification, and other regions. The mean size of the ROIs was approximately 20–40 mm². Three ROIs were selected in the tumor parenchyma and the average value was taken as the final result. Measurements were completed by two experienced radiologists (Fig. 1).

Fig. 1.

(a–e) Astrocytoma (WHO III) in a 56-year-old woman who survived for 668 days. (a) Contrast-enhanced T1W image showed an irregular rim-enhancing mass. (b) Standard diffusion coefficient map (ADC = 0.687 × 10⁻³ mm²/s). (c) Fast diffusion coefficient map showed linear hyperperfusion in the inside and the margin of the tumor (D* = 4.34 × 10⁻³ mm²/s). (d) Perfusion-related fraction map clearly highlighted the area with high perfusion (f = 0.585). (e) The pathological pictures showed that the density of tumor cells was increased, and the nuclear heteromorphism was obvious. The cell hyperplasia was active. Occasionally the mitotic image was seen. (f–j) Astrocytoma (WHO II) in a 62-year-old man who survived for 523 days. (f) Contrast-enhanced T1W image also shows a rim-enhancing mass. (g) Standard diffusion coefficient map (ADC = 0.701 × 10⁻³ mm²/s). (h) Fast diffusion coefficient map shows linear slightly high perfusion at the margin of the tumor (D* = 4.05 × 10⁻³ mm²/s). (i) Perfusion-related fraction map (f = 0.489). (j) The pathological pictures showed that the tumor cells were active in growth, with certain heterogeneity, and no obvious nuclear division was seen. ADC, apparent diffusion coefficient; T1W, T1-weighted; WHO, World Health Organization.

Statistical analysis

ADC, D, D*, and f values of each group conformed to normal distribution. All data were analyzed by SPSS 17.0 and MedCalc 19.0.7. The results were expressed as mean ± SD ( $\bar{x} \pm s$ ). Independent samples t test was used to compare the ADC, D, D*, and f values between the two-year death group and the survival group. Receiver operating characteristic (ROC) curve was used to test the sensitivity and specificity corresponding to cut-off point of each parameter for predicting the two-year survival rate. Areas under the ROC curve (AUC) were calculated. The statistical difference between each parameter was quantified by Z-test. Overall survival was analyzed by the Kaplan–Meier method, using log-rank statistics. Pearson correlation analysis was used to analyze the correlation between the parameters and patient’s survival time. The survival time refers to the period from the initial MRI scan to death, or from the MRI scan to the last documented survival day. P < 0.05 was considered statistically significant.

Result

Comparison of IVIM-derived parameters between the death group and the survival groups over the two-year follow-up

Of the 78 patients with astrocytoma, 10 cases were lost, four cases underwent surgery alone, two cases received radiotherapy only, and two were not completely resected. Finally, 60 patients were included (34 men, 26 women; mean age = 62.23 ± 16.32 years) with 33 high-grade (World Health Organization [WHO] III–IV), 27 low-grade (WHO I–II), 18 cases of frontal lobes, 17 cases of temporal lobes, 13 cases of occipital lobes, and 12 cases of parietal lobes. The mean survival days was 837.5 ± 581.7. Patients were divided into the death group (37 cases) and survival group (23 cases) according to the results of the two-year follow-up. Details are shown in Table 1.

Table 1.

Comparison of quantitative parameters between the death and survival groups in the biennium.

	Death group (37 cases)	Survival group (23 cases)	t	P
Age (years)	62.6 ± 10.3	60.9 ± 14.2	1.246	0.069
Male/Female	21/16	13/10	–	0.598
High-grade/low-grade tumors	31/6	2/21	–	0.000*
ADC (10^–3 mm²/s)	0.628 ± 0.073	0.739 ± 0.123	–2.436	0.026*
D（10⁻³ mm²/s）	0.518 ± 0.152	0.603 ± 0.164	–0.085	0.094
D*（10⁻³ mm²/s）	4.463 ± 1.615	3.078 ± 1.031	3.524	0.012*
f	0.429 ± 0.183	0.340 ± 0.127	5.526	0.000*

Values are given as mean ± SD.

*P < 0.05 demonstrates that the difference was statistically significant.

ADC, apparent diffusion coefficient; D, slow diffusion coefficient; D*, fast diffusion coefficient; f, fractional perfusion-related volume.

The value of the parameters in the death and survival groups were expressed as mean ± SD (x ± s) (Table 1). Independent samples t-test showed that apart from the D value (P = 0.094), the ADC value, D* value, and f value in the death and survival groups were statistically significant (P < 0.05). The ADC value of the death group was lower than that of the survival group, and the D* and f values were higher than those of the survival group. The intraclass correlation coefficient (ICC) was used to evaluate the consistency of measurements of parameters by different observers. In this experiment, the ICC values were as follows: ICC_ADC = 0.936; ICC_D = 0.912; ICC_D* = 0.939; and ICC_f = 0.942.

Evaluating the accuracy of the two-year survival rate using different parameters

In Fig. 2, the ROC showed the AUC of the ADC, D*, and f values, which were 0.811, 0.858, and 0.892, respectively. To evaluate the two-year survival rate, the cut-off value of the ADC was 0.668 × 10⁻³ mm²/s, with a sensitivity of 82.6% and specificity of 73.0%, respectively. The cut-off value of D* was 3.913 × 10⁻³ mm²/s, with a sensitivity of 78.4% and specificity of 87.0%. The cut-off value of f value was 0.487, with a sensitivity of 83.8% and a specificity 87.0%. Relative to D* and ADC values, the AUC of f is the largest. However, there was no significant difference between the AUCs of each parameter in predicting two-year survival rate (P > 0.05) (Table 2). The Kaplan–Meier survival curves using the thresholds were statistically significant (Fig. 3), with a log-rank test of P_ADC = 0.004, P_D* = 0.000, and P_f = 0.000, respectively.

Fig. 2.

ROC of IVIM-derived parameters in evaluating two-year survival of patients. ADC, apparent diffusion coeffieient; D*, fast diffusion coefficient; f, fractional perfusion-related volume; IVIM, intravoxel incoherent motion imaging; ROC, receiver operating characteristic curve.

Table 2.

Pairwise comparison of AUC.

	Z	P
AUC_f∼ADC	0.799	0.424
AUC_f∼D*	0.511	0.610
AUC_D*∼ADC	0.569	0.570

P < 0.05 demonstrates that the difference was statistically significant.

ADC, apparent diffusion coefficient; AUC, area under the curve; D*, fast diffusion coefficient; f, fractional perfusion-related volume.

Fig. 3.

Kaplan–Meier survival curves compared overall survival according to cut-off values of ADC (a), D* (b), and f (c). ADC, apparent diffusion coefficient; D*, fast diffusion coefficient; f, fractional perfusion-related volume.

Correlation between IVIM-derived parameters and patient survival time

Pearson correlation analysis showed a positive correlation between ADC value and survival days (P = 0.023, r = 0.625). D* and f values were negatively correlated with survival days (P = 0.012, r = –0.655; P = 0.000, r = –0.725) (Fig. 4). The correlation between f and survival days was higher than D* and ADC.

Fig. 4.

The scatter plot between ADC, D*, f, and survival days. ADC, apparent diffusion coefficient; D*, fast diffusion coefficient; f, fractional perfusion-related volume; r, correlation coefficient.

Discussion

Astrocytoma has characteristics of obvious infiltrative growth, resulting in blurry boundaries from normal brain tissue, so it cannot be completely resected through surgical treatment. Despite receiving postoperative chemoradiotherapy treatment, the survival time of > 50% of patients was still < 24 months. In the present study, the two-year mortality rate of patients with astrocytoma was 61.7%, which was consistent with the literature (11). If we can understand the biological characteristics of the tumors and predict the patient’s clinical prognosis before treatment, the clinicians can develop personalized treatment plans and guide postoperative medication. It is widely accepted that low-grade tumors grow more slowly and have better prognosis than high-grade tumors. However, since some of the low-grade tumors have the anaplastic characteristics of developing into malignant tumors, the prognosis of these patients with low-grade tumors is poor (12). In our study, 6 (16.2%) patients with low-grade astrocytoma received the excision of tumor as thoroughly as possible and regular radiotherapy or chemotherapy, but the survival time was <2 years. Because tumors are usually heterogeneous, the pathological grade is mainly determined by the substantial portion of the higher malignant degree. A tumor biopsy from an inappropriate site may underestimate the biological behavior and the clinical treatment of the low-grade tumors, which can lead to bad prognosis (13). Therefore, there are some limitations of the WHO pathological grade in predicting the prognosis and the survival of patients with glioma (14).

IVIM can assess the true diffusion (D value) and perfusion information (D* and f values) from a microscopic perspective, which is expected to accurately assess the patient’s prognosis as early as possible (15). Currently, there are few reports about using IVIM to evaluate the prognosis of patients with astrocytoma. In the present study, we evaluated the two-year survival of patients using IVIM, and found that the ADC value of the death group was lower than that of survival group, and the D* value and f value were higher than those in the survival group. According to previous literature, ADC can reflect the tumor cell density. The lower the ADC value is, the higher the tumor cell density is, and D* and f represent microangiogenesis; the greater the D* and f values, the more microvessels. The proliferation of tumor cell density and microvessels may predict the poor prognosis of the patients (16,17). Therefore, the ADC value of the death group was lower than that of the survival group, and the D* and f values were higher. According to the calculation of the AUC, the ADC, D*, and f values can all effectively predict the survival time in two years (AUC_f = 0.892, AUC_D* = 0.858, AUC_ADC = 0.811, respectively). However, the results of each parameter between groups were not statistically significant (P > 0.05). Compared with the D* and ADC values, the f value had higher sensitivity and specificity in predicting the two-year survival rate. The reason may be that the parameter f obtained from the bi-exponential model can distinguish diffusion from perfusion, making the result more accurate. Second, it may be that the multi-component perfusion had a great effect on D* (18 –21). Specifically, the bi-exponential model assumes that there is only one type of blood vessel within the voxel, namely the capillaries (22). In biological tissues, each voxel contains many types of blood vessels, such as arteries, veins, arterioles, venules, etc. There are significant differences in blood flow velocity in different types of blood vessels (23) and its perfusion parameter is bound to fail to accurately reflect the perfusion information of the tissue, thus influencing the judgment of prognosis.

So far, little has been reported on the correlation between IVIM parameters and survival rates. Zulfiqar et al. (24) found that low ADC value was associated with decreased survival. Our results also showed a positive correlation between ADC values and survival (r = 0.625). The f and D* values were negatively correlated with survival, and f value (r = –0.725) was significantly higher than the D* value (r = –0.655). Federau et al. (25) showed that the D* value might be related to the degree of formation of mature blood vessels, mainly in the non-enhanced regions. The enhancement of tumor parenchyma is more common in immature microvessels. The f value reflects not only the blood flow in the microvessels, but also the microvascular perfusion. Numerous studies have also shown that the formation of microvessels is associated with the survival of patients (26,27), which may explain why the correlation between f and survival was slightly stronger.

There are still some limitations in this study. One is that the total number of samples is relatively small and more patients are needed to confirm the results. Second, factors such as age, therapeutic consistency, and molecular markers were not taken into account, and more experimental data were needed in the future.

In conclusion, IVIM has great potential in evaluating the prognosis of patients with astrocytoma, especially for low-grade astrocytomas with malignant potential, which may provide valuable information for clinical assessment of the biological behavior of the lesions and the formulation of treatment plans.

Footnotes

Acknowledgements

We espacially thank Dr Kaiyu Wang from GE Healthcare for her technique support of our research.

Declaration of conflicting interests

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

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Jiang Wu

References

Zhao

Guo

Huang

, et al. Quantitative analysis of permeability for glioma grading using dynamic contrast-enhanced magnetic resonance imaging. Oncol Lett 2017; 14:5418–5426.

Hilario

Sepulveda

Perez-Nuñez

, et al. A prognostic model based on preoperative MRI predicts overall survival in patients with diffuse gliomas. AJNR Am J Neuroradiol 2014; 35:1096–1102.

Mangla

Ginat

Kamalian

, et al. Correlation between progression free survival and dynamic susceptibility contrast MRI perfusion in WHO grade III glioma subtypes. J Neurooncol 2014; 116:325–331.

Weller

Felsberg

Hartmann

, et al. Molecular predictors of progression-free and overall survival in patients with newly diagnosed glioblastoma: a prospective translational study of the German Glioma Network. J Clin Oncol 2009; 27:5743–5750.

Wang

Niu

Gao

, et al. Prognostic Factors for the Survival Outcome of High-grade Multicentric Glioma. World Neurosurg 2018; 112:e269–e277.

Chen

, et al. A restricted cell population propagates glioblastoma growth after chemotherapy. Nature 2012; 488:522–526.

Togao

Hiwatashi

Yamashita

, et al. Differentiation of high-grade and low-grade diffuse gliomas by intravoxel incoherent motion MR imaging. Neuro Oncol 2016; 18:132–141.

Cao

Suo

Han

, et al. Application of a simplified method for estimating perfusion derived from diffusion-weighted MR imaging in glioma grading. Front Aging Neurosci 2018; 9:432.

Niu

Shakir

, et al. An evidence-based approach to assess the accuracy of intravoxel incoherent motion imaging for the grading of brain tumors. Medicine (Baltimore) 2018; 97:e13217.

10.

Pedeutour-Braccini

Burel-Vandenbos

Gozé

, et al. Microfoci of malignant progression in diffuse low-grade gliomas: towards the creation of an intermediate grade in glioma classification. Virchows Arch 2015; 466:433–444.

11.

Federau

Cerny

Roux

, et al. IVIM perfusion fraction is prognostic or survival in brain glioma. Clin Neuroradiol 2017; 27:485–492.

12.

Villanueva-Meyer

Wood

Choi

, et al. MRI features and IDH mutational status of Grade II diffuse gliomas: impact on diagnosis and prognosis. AJR Am J Roentgenol 2018; 210:621–628.

13.

Szopa

Burley

Kramer-Marek

, et al. Diagnostic and therapeutic biomarkers in glioblastoma: current status and future perspectives. Biomed Res Int 2017; 2017:8013575.

14.

Bobek-Billewicz

Stasik-Pres

Hebda

, et al. Anaplastic transformation of low-grade gliomas (WHO II) on magnetic resonance imaging. Folia Neuropathol 2014; 52:128–140.

15.

Raja

Sinha

Saini

, et al. Assessment of tissue heterogeneity using diffusion tensor and diffusion kurtosis imaging for grading gliomas. Neuroradiology 2016; 58:1217–1231.

16.

Chen

Hou

Lou

, et al. The correlation between MR diffusion-weighted imaging and pathological grades on glioma. Eur Rev Med Pharmacol Sci 2014; 18:1904–1909.

17.

Deike

Wiestler

Graf

, et al. Prognostic value of combined visualization of MR diffusion and perfusion maps in glioblastoma. J Neurooncol 2016; 126:463–472.

18.

Wetscherek

Stieltjes

Laun

FB.

Flow-compensated intravoxel incherent motion diffusion imaging. Magn Reson Med. 2015; 74:410–419.

19.

Wurnig

Kenkel

Filli

, et al. A standardized parameter-free algorithm for combined intravoxel incoherent motion and diffusion kurtosis analysis of diffusion imaging data. Invest Radiol 2016; 51:203–210.

20.

Wurnig

Donati

Ulbrich

, et al. Systematic analysis of the intravoxel incoherent motion threshold separating perfusion and diffusion effects: Proposal of a standardized algorithm. Magn Reson Med 2015; 74:1414–1422.

21.

Bisdas

Koh

Roder

, et al. Intravoxel incoherent motion diffusion-weighted MR imaging of gliomas: feasibility of the method and initial results. Neuroradiology 2013; 55:1189–1196.

22.

Surov

Meyer

Wienke

Associations between apparent diffusion coefficient (ADC) and KI 67 in different tumors: a meta-analysis. Oncotarget 2018; 9:8675–8680.

23.

Shen

Zhao

Jiang

, et al. Intravoxel incoherent motion diffusion-weighted imaging analysis of diffusion and microperfusion in grading gliomas and comparison with arterial spin labeling for evaluation of tumor perfusion. J Magn Reson Imaging 2016; 44: 620–632.

24.

Zulfiqar

Yousem

Lai

ADC values and prognosis of malignant astrocytomas: does lower ADC predict a worse prognosis independent of grade of tumor?–a meta-analysis. AJR Am J Roentgenol 2013; 200:624–629.

25.

Federau

Meuli

O’Brien

, et al. Perfusion measurement in brain gliomas with intravoxel incoherent motion MRI. AJNR Am J Neuroradiol 2014; 35:256–262.

26.

Iima

Reynaud

Tsurugizawa

, et al. Characterization of glioma microcirculation and tissue features using intravoxel incoherent motion magnetic resonance imaging in a rat brain model. Invest Radiol 2014; 49:485–490.

27.

Chen

Lin

, et al. Classification of microvascular patterns via cluster analysis reveals their prognostic significance in glioblastoma. Hum Pathol 2015; 46:120–128.