Synopsis

Keywords: Tumors, Diffusion/other diffusion imaging techniques

To investigate the utility of time-dependent DWI for differentiating between glioblastoma and brain metastasis, 65 patients with glioblastoma and 27 patients with brain metastasis were examined. ADC was not significantly different between the two tumor types neither at a short (7.1ms) nor at a long (44.5ms) diffusion time, whereas the ADC difference (ΔADCmean, ΔADC5, ΔADC95) and ADC change ratio (rADCmean, rADC5, rADC95) were significantly higher in brain metastasis than in glioblastoma. The ΔADCmean showed the best diagnostic performance. The ΔADCmean and rADCmean showed a significant negative correlation with ADC_44.5msmean, but not with extracellular extravascular space volume fraction derived from DCE-MRI.

INTRODUCTION

Preoperative differentiation between brain metastasis and glioblastoma is clinically important, but is often problematic as they can show similar characteristics on conventional MR images. There have been controversies regarding the abilities of conventional ADC values to differentiate brain metastases from glioblastomas.^1,2 Complementary to conventional pulsed gradient spin-echo sequences (PGSE), oscillating gradient spin-echo (OGSE) sequences³ allow for DWI with a short diffusion time, enabling time-dependent diffusion analysis which could provide specific information regarding restricted diffusion. Our purposes were to investigate the utility of time-dependent DWI parameters in distinguishing brain metastases from glioblastomas, and to examine their correlation with extravascular extracellular space (v_e).

METHODS

A retrospective study was performed including 27 patients with brain metastasis (mean age, 68 ± 10 years) and 65 patients with glioblastoma (69 ± 13 years). All patients underwent preoperative MR imaging using a 3T system (MAGNETOM Prisma; Siemens Healthcare, Erlangen, Germany) with a 20-channel head/neck coil. Time-dependent DW images were acquired using research DWI sequences with OGSE¹ using sine-modulated trapezoidal waveforms (effective diffusion time = 7.1 ms) and PGSE (44.5 ms) with b values of 0 and 1500 s/mm². Maps of the difference in ADC values between OGSE and PGSE sequences were generated: ΔADC = ADC_7.1ms - ADC_44.5ms. In addition, maps of the ratio of ADC change between OGSE and PGSE sequences were generated: rADC = (ADC_7.1ms - ADC_44.5ms)/ADC_44.5ms × 100 (%). Moreover, maps of the volume of extravascular extracellular space per unit volume of tissue (v_e) were obtained using DCE-MRI. ROI analysis was performed by two independent radiologists to measure the ADCs at the two diffusion times (ADC_7.1ms and ADC_44.5ms), ΔADC, rADC, and v_e (Fig. 1). The interobserver agreement regarding parametric measurements by the two observers was analyzed by calculating the intraclass correlation coefficient (ICC). Measurements by the two observers for each patient were averaged for further analysis. The mean, fifth percentile, and 95th percentile values of each parameter were compared between brain metastases and glioblastomas using the Mann-Whitney U test. In addition, the diagnostic performances of the parameters were evaluated using ROC curve analysis. Furthermore, we examined the correlation of the time-dependent parameters with ADCs and v_e using Pearson’s correlation coefficient.

RESULTS

The agreement of the two observers was excellent for all measures (ICCs ranged from 0.825 to 0.981). The results of the parametric comparisons are shown in Figure 2. No significant difference was noted between glioblastomas and brain metastases in any of the three indices of ADC_44.5ms and ADC_7.1ms. ΔADCmean (P<0.01), ΔADC5 (P<0.05), ΔADC95 (P<0.01), rADCmean (P<0.01), rADC5 (P<0.01), and rADC95 (P<0.01) were significantly higher in brain metastases than in glioblastomas, whereas the v_e5 (P<0.05) values were significantly lower for brain metastases than for glioblastomas. The ROC curve analysis showed significance for ΔADCmean (AUC=0.877, P<0.01), ΔADC95 (0.865, <0.01), rADCmean (0.819, <0.01), rADC5 (0.652, 0.02), rADC95 (0.796, <0.01), and v_e5 values (0.630, 0.04). Considering ADC_44.5ms, ADC_7.1ms, ΔADC, rADC, and v_e values, the highest AUC was obtained for ADC_44.5ms5, ADC_7.1ms95, ΔADCmean, rADCmean, and v_e5. Pairwise comparisons revealed that AUCs of ΔADCmean and rADCmean were significantly greater than that of ADC_44.5ms5 (each P<0.001). The ROC curves for the ADC_44.5ms5, ADC_7.1ms95, ΔADCmean, and rADCmean are shown in Figure 3. The ΔADCmean (r=-0.309, P<0.01) and rADC mean (-0.618, <0.01) showed a significant negative correlation with ADC_44.5msmean, but not with v_e5 (r=-0.071, -0.006, P=0.50, 0.95).

DISCUSSION

Our results suggest that time-dependent DWI parameters, especially ΔADCmean, are useful imaging markers for distinguishing brain metastases from glioblastomas, while ADC itself is not. As the majority of brain metastases are composed of epithelial cells, the cell gaps are very narrow due to the cell junction system, while glioblastomas are composed of heterogeneous cells with extracellular matrix and fine hemorrhage and necrosis. Theoretically, narrower extracellular space could explain stronger diffusion time-dependence of ADC in brain metastasis. However, we did not find significant correlation between the time-dependent DWI parameters and v_e derived from DCE-MRI. Further studies are needed to elucidate the pathological factors that account for our findings.

CONCLUSION

The time-dependent DWI parameters, especially ΔADCmean may be useful in distinguishing brain metastases from glioblastomas.

Acknowledgements

No acknowledgement found.