Time-dependent diffusion for assessment of treatment response in GL261 murine glioma model
Olivier Reynaud1, Jin Zhang1, Kerryanne Veronica Winters1, Dmitry Novikov1, and Sungheon Gene Kim1

1Center for Biomedical Imaging, Department of Radiology, NYU School of Medicine, New York, NY, United States


In this study, we investigate the feasibility of using the recently proposed diffusion MRI method, POMACE (Pulsed and Oscillating gradient MRI Assessment of Cell size and Extracellular space) for characterization of tissue changes induced by chemotherapy. Surface-to-volume ratio (S/V), cell size (Rcell) and extracellular space volume fraction (ECS) are estimated prior to and after 5FU and bevacizumab treatment in GL261 murine glioma model. Preliminary results indicate an early ECS decrease two days after 5FU/bevacizumab administration. By combining OGSE and PGSE, POMACE provides non-invasive and quantitative metrics regarding tumor microstructure that can be applied to evaluate treatment response.


Diffusion weighted imaging (DWI) is a promising tool to measure cellular structural characteristics non-invasively. However, cellular restriction of water diffusion in cancer cells is typically expected in 5-10 μm range which cannot be fully assessed by using either pulsed gradient spin echo (PGSE) or oscillating gradient spin echo (OGSE) diffusion experiment alone. Recent studies [1,2] showed that combining PGSE and OGSE DWI experiment allowed for characterization of the tumor microstructure using quantitative parameters, such as surface-to-volume ratio (S/V), cell size (Rcell) or extracellular space fraction (ECS). The purpose of this study was to investigate the utility of such techniques to characterize the early treatment response in a mouse glioblastoma model. The in vivo imaging findings will be compared with histopathological assessment.


GL261 murine GBM cells (106 in 5μL) were injected into the subcortex of 6-8 week old C57BL/6 mice (n=9, female) under anesthesia (air +3% isoflurane). The MR protocol consisted of 10 OGSE and 4 PGSE measurements at 7T (Bruker Biospec Avance2, Etlingen), probing the diffusion coefficient D(t) with diffusion times = 6/8/16/31 ms for PGSE and frequencies in the range of 65-225 Hz for the fast ramp cos-OGSE (TR/TE=3000/70 ms, BW=300 kHz, 1 readout segment, NA=20, NR=2, res. 250x250x1500 μm, FoV 20x20 mm, b=[0,200,400] s/mm2, dir. (1,1,1), TA=6 min, total time 84 min). Mice were imaged prior and after tumor treatment: pre-treatment MRI followed by bevacizumab (10 mg/kg) on the first day; 5FU (80 mg/kg) on the second day; post-treatment MRI on the third day. Mice were then sacrificed for ex vivo imaging, histology and light microscopy. S/V and free diffusivity (D0) are estimated using the linear dependence of D(t) with ω-1/2 in the very short time regime, as in [1]. The extracellular fraction (ECS), cell size (Rcell) and intra and extracellular diffusivities (Dintra/Decs) are estimated using a two-compartment geometrical model for the tumor environment. Cells were assumed impermeable for diffusion times t < 30 ms. Diffusion in the ECS was considered Gaussian and in the tortuosity limit, as a first approximation [2]. This approach of using the diffusion time range suitable for cancer cells is also shown in Figure 1 and referred to as POMACE (Pulsed and Oscillating gradient MRI Assessment of Cell size and Extracellular space). The fit stability was improved by setting Decs=2.2 μm2/ms. Parameter estimation was performed in Matlab (The Natick, MA) using non-linear fitting at voxel level to derive parametric maps of ECS, Rcell, S/V, D0, Dintra and Decs in the tumor. The POMACE parameters were assessed at tumor level for statistical significance between pre- and post-treatment imaging sessions in treated and control groups.


D(t) was highly dependent on diffusion time / oscillation frequency inside the tumor (+80% D(t) between t=31ms and ω=225Hz, Fig. 2). S/V, ECS and Rcell maps revealed strong microstructure heterogeneity within the tumor (Fig. 3). Regions of low S/V and high Rcell correlated well with the areas of high contrast enhancement (arrows in Fig. 3). As expected for spheres, the total S/V correlated very well with 6*(1-ECS)/Rcell inside the GBM (R2=0.87 after exclusion of rim voxels). A 17% decrease in S/V was found compared to POMACE-based estimations. At tumor level, the ECS ratio post/pre-treatment was found significantly lower than unity for the treated group (ΔECS=-5%, N=4, P<0.02, Fig. 4). No significant difference was found between groups for S/V, Rcell, D0, or Dintra. Changes in tumor growth (pre/post-treatment) were not found significantly different between control and treated groups (+49% vs. +65%, P>0.5).

Discussion and conclusion

Compared to contrast enhanced MRI and conventional DWI, time-dependent diffusion allowed to disentangle microstructural regional variations observed in post contrast images into objective quantifiable geometrical metrics (S/V, Rcell, ECS) that can be assessed with different modalities. Using POMACE, a novel way to combine OGSE and PGSE data, S/V changes can be decomposed into cell fraction (ECS) and cell size variations using the D(t) time-dependence at short frequencies / long times. Comparison of S/V, ECS and Rcell measurements suggest (a) partial cellular membrane deterioration, and/or (b) decrease of the apparent restrictive surface due to increase cellular packing. In addition, changes in tumor structure , such as ECS, prior to changes in tumor growth were detected in the treated group, although additional analysis including ex vivo MRI and histopathological assessment are required in order to validate these changes and establish the time-dependent diffusion parameters a set of objective markers of treatment response.


This work was supported by NIH R01 CA160620.


[1] Reynaud et al. Magn. Reson Med (2015) doi:10.1002/mrm.25865; [2] Reynaud et al. Proc. ISMRM 2015 (program number 344)


POMACE processing design. DWI OGSE high-frequency data is first fitted in the very short time regime (linear dependence of D(ω) with ω-1/2) to extract D0 and S/V [1]. Low-frequency OGSE and PGSE data is fitted to a geometrical model of impermeable spheres [2]. D0 is used here as a constraint to increase fit robustness.

Time dependent diffusion in tumor after treatment (N=4) and fit in the short and long-time regime (red/blue dashed lines).

DCE enhancement maps and parametric maps. The color scale for S/V, ECS and Rcell map is respectively [0.2-0.8] μm-1, [20-60] % and [2-6] μm. The red and white arrows point to high enhancement and low S/V areas.

Influence of tumor treatment on surface-to-volume ratio (S/V), extracellular space fraction (ECS) and cell radius (Rcell). * indicates P<0.05. ECS is significantly lower after treatment with 5FU+bevacizumab.

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