Diffusion kurtosis imaging can efficiently assess the glioma grade, cellular proliferation and survival
Rifeng Jiang1, Jingjing Jiang1, Jingjing Shi1, Yihao Yao1, Nanxi Shen1, Changliang Su1, Ju Zhang1, and Wenzhen Zhu1

1Radiology, Tongji Hospital, Wuhan, China, People's Republic of

Synopsis

Compared with conventional diffusion metrics, the kurtosis metrics derived from DKI in the solid region of the tumors were better diagnostic factors in distinguishing HGGs from LGGs and identifying grade II, III and IV gliomas. The kurtosis metrics offered great potential to noninvasively predict the cellular proliferation of gliomas. DKI was also useful to evaluate the survival of glioma patients, and MK was a significant death risk in glioma patients.

Introduction

Conventional diffusion imaging techniques are not sufficiently accurate in evaluating glioma grade and cellular proliferation, which are critical for glioma treatment. Diffusion kurtosis imaging (DKI), an advanced non-Gaussian diffusion imaging technique, has shown potential in grading glioma (1-3); however, the sample size was limited in these studies, and no comparison of DKI metrics with apparent diffusion coefficient (ADC) was reported. Moreover, the relationships between DKI and the proliferative activity of glioma cells as well as the prognosis of glioma patients were not evaluated. Therefore, the roles of DKI in gliomas still have not been fully elucidated. In this study, we assessed and compared the value of the kurtosis metrics (MK, Ka and Kr) and conventional diffusion metrics (MD, FA and ADC) in grading gliomas, and also evaluated the correlations between these metrics and the Ki-67 labeling index (Ki-67 LI) as well as the survival of glioma patients.

Methods

Seventy-four patients with histopathologically confirmed cerebral glioma between July 2012 and June 2014 underwent routine MRI, DWI and DKI. According to WHO criteria, 3 had grade I glioma, 31 had grade II, 19 had grade III, and 21 had grade IV. Mean kurtosis (MK), axial kurtosis (Ka), radial kurtosis (Kr), mean diffusivity (MD), fractional anisotropy (FA) and ADC were semi-automatically obtained in the solid part of tumor and the contralateral normal-appearing white matter (NAWM). The values in NAWM were used as reference to normalize. Histology was applied to determine the grade of the tumor and the expression of Ki-67. Follow-ups were performed till August 2015. Statistical analysis included the independent-samples t-test, One-way ANOVA, receiver operating characteristic (ROC) curve, Pearson correlation analysis, Kaplan-Meier’s survival curve and Cox’s regression analysis.

Results

MK, Ka, Kr, MD and ADC in the solid part of tumor were significantly different between high-grade gliomas(HGG) and low-grade gliomas(LGG), between grade II and grade III gliomas, and between grade III and grade IV gliomas (P<0.001 for all). FA did not significantly differ between glioma grades (P>0.05 for all). The kurtosis metrics showed the highest AUCs and optimal sensitivity and specificity in all the differentiations. Significant correlations were found between Ki-67 LI and MK, Ka, Kr, MD or ADC (P<0.001 for all), but not between Ki-67 and FA (P>0.05). Among 18 dead patients that were followed-up, significant correlations were revealed between the survival time and MK or MD (P<0.05), but not between survival time and FA (P>0.05). Kaplan-Meier’s survival curves stratified by MK, MD and FA for the endpoint of all-cause mortality were analyzed. The Log Rank P values for MK and MD were 0.003 and 0.004 respectively, but >0.05 for FA. A multivariable prediction including MK, MD and FA of all-cause death was also performed using Cox’s regression model, and the results demonstrated that only MK was a significant predictors positively associated with the death in the glioma patients (Hazard Ratio=1.891; 95% CI=1.004-3.561; P=0.049).

Conclusion

The kurtosis metrics in the solid region of the tumors were better diagnostic factors in distinguishing HGGs from LGGs and identifying grade II, III and IV gliomas; the kurtosis metrics also offered great potential to noninvasively predict the cellular proliferation of gliomas. Compared with conventional diffusion metrics, the kurtosis metrics show greater potential as imaging markers for the accurate demonstration of the microstructural changes caused by increasing glioma grade and cellular proliferation, which might achieve more accurate diagnosis and optimal therapy for glioma patients. What is more, based on our current follow-up data analysis, DKI was useful to evaluate the survival of glioma patients, and MK was a significant death risk in glioma patients.

Acknowledgements

The authors thank Xingxing He for his advice on manuscript writing, thank Guoping Wang, Cong Liu, Sanpeng Xu, Yaobing Chen and Dong Kuang for their advice on Ki-67 detection, and also thank Ping Yin, Wenhua Liu and Chanchan Liu for their assistance with the statistical analyses.

References

1. Van Cauter S, Veraart J, Sijbers J, et al. Gliomas: diffusion kurtosis MR imaging in grading. Radiology. 2012;263(2):492-501.

2. Raab P, Hattingen E, Franz K, Zanella FE, Lanfermann H. Cerebral Gliomas: Diffusional Kurtosis Imaging Analysis of Microstructural Differences. Radiology. 2010;254(3):876-81.

3. Tietze A, Hansen MB, Ostergaard L, Jespersen SN, Sangill R, Lund TE, Geneser M, Hjelm M and Hansen B. Mean Diffusional Kurtosis in Patients with Glioma: Initial Results with a Fast Imaging Method in a Clinical Setting. AJNR American journal of neuroradiology. 2015; 36(8):1472-1478.

Figures

Correlation of diffusion kurtosis imaging with tumor grade and Ki-67. Rows 1-3 correspond to three patients with diffuse astrocytoma (WHO grade II) in the left temporal lobe, anaplastic astrocytoma (WHO grade III) in the left frontal lobe and glioblastoma (WHO grade IV) in the right fronto-temporal lobe, respectively. Columns a-e are contrast-enhanced T1-FLAIR, MK, MD, FA and Ki-67 images (400×), respectively.

ROC curves for all the metrics in differentiating tumor grades. ROC curves and AUCs for all the metrics in the solid region of the tumor for the differentiation (a) between HGGs and LGGs, (b) between grade II and III and (c) between grade III and IV gliomas.

Statistical values of all metrics for differentiating between HGGs and LGGs, grade II and grade III, and grade III and grade IV gliomas

Correlations between Ki-67 and each metric. Scatter diagrams demonstrating the correlations between Ki-67 labeling index and (a) MK, (b) Ka, (c) Kr, (d) MD, (e) FA or (f) ADC.

Kaplan-Meier’s survival curves (a) stratified by MK and (b) MD for the endpoint of all-cause mortality. The Log Rank P values for MK and MD were 0.003 and 0.004 respectively. Note the cut-off values of MK and MD in survival curves were derived from those of corresponding ROC curves for the differentiation between HGG and LGG.



Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
4192