Purpose: Restricted water diffusion in microstructural tissue depends upon measurement (diffusion) time. A random walk with barriers model (RWBM) could analyze these effects on diffusion to reveal microstructural properties from macroscopic imaging characteristics within heterogeneous brain tumors. Gross tumor volume in gadolinium enhanced T1-weighted images (GTV-Gd) usually contains tumor mixtures, which is difficult to be analyzed by a single diffusion measurement. This research aims to combine clustering analysis with the RWBM to demonstrate microstructural diffusion properties in brain tumors.
Methods: Ten patients with brain tumors (eight with brain metastases, two with glioblastoma) had diffusion weighted (DW) images with three different diffusion times using pulsed diffusion gradients (PG) and sine-modulated oscillating gradients at f=30Hz (OG30) and f=50Hz (OG50) of a prototype sequence on a 3T scanner (MAGNETOM Skyra, Siemens Healthcare). DW images were acquired in 6 diffusion directions with 8 b-values (0 to 2600 s/mm²) for PG, 6 b-values (0 to 1300 s/mm²) for OG30, and 5 b-values (0 to 500 s/mm²) for OG50. The delineated GTV-Gd of each tumor, if > 1 cm³, was classified to smaller volumes using DW image intensities and fuzzy c-means. The RWBM that characterizes time-dependent diffusion was applied to DW images in the GTV-Gd to yield five parameters: cell diameter, unrestricted diffusion coefficient (D₀) at a short time limit, bulk diffusion coefficient (D(inf)) at a long time limit, membrane permeability and effective restriction.
Results: Of 23 brain tumors, the mean tumor cell diameter of GTV_Gds was (16.5±4.20) μm. D₀ was (2.20±0.70) um²/ms. D(inf) decreased to (0.60±0.14) μm²/ms due to microstructure restrictions with the mean effective restriction of 2.8±1.2 and the mean membrane permeability of (0.053±0.009) μm/ms.
Conclusion: The RWBM yields parameters beyond analyzing DW images with a single diffusion time, and has potential to reveal microstructure of brain tumors with further investigation.
Funding Support, Disclosures, and Conflict of Interest: This work is supported in part by NIH/NCI 1U01CA183848.