gms | German Medical Science

Objective: Diffusion MRI with tractography-based estimation of fiber-connectivity is applied for surgical planning with the aim of maximal safe tumor resection, but preservation of functionally relevant white-matter (WM) tracts. Free-water (FW) partial volume effects however limit tracking algorithms at the boarder of cerebrospinal fluid or in the presence of vasogenic edema, which obscures WM estimates especially in peritumoral areas, endangering delineation of fiber displacement and tract identification. We compared a novel method for free-water correction of diffusion weighted imaging (DWI) to previous approaches on single- and multi-shell synthetic data and healthy adults, and applied it to a brain tumor patient database.

Methods: 28 patients with cerebral glioma (mean age 51+/-17 yrs, 20 high-grade) were preoperatively examined using anatomical MRI and DWI. To determine the free-water volume fraction (FWVF), an artificial neural network (ANN) was trained on synthetic data generated from patient-specific DWI data. Voxels with known tissue properties (e.g. corpus callosum, ventricles, gray matter) were extracted, and by superposition of different voxels and varying tissue proportions, a synthetic dataset with known diffusion microstructure was composed. On this dataset, the ANN was trained to predict the FWVF, and then used to infer the FWVF of the whole brain. Finally, the diffusion signal uncontaminated by FW was determined.

Results: Evaluation of the exactness of the proposed method on the synthetic dataset showed similar results as compared to FW elimination for two-shell data, but better results as compared to other approaches for one- and three-shell data. Applying our novel FW correction algorithm on patient data, the predicted water compartment affected by edema ranged up to 30-40%. While FW correction resulted in corrected fractional anisotropy (FA) values within peritumoral edema, normal appearing white matter values were not altered. Peritumoral fiber-tracking improved and allowed delineation of fiber-tracts which were not depicted without FW correction (see Figure 1 [Fig. 1], Figure 2 [Fig. 2]).

Conclusion: We present a novel method for free-water correction in diffusion MRI, which is fast, independent of diffusion shells (b-values), and at no cost of additional acquisition time for the patient. Most importantly, it improves WM estimates and preoperative peritumoral fiber-tracking, which might be critical for surgical planning and postoperative functional outcome.