Gliomas are malignant brain tumors whose potential to infiltrate healthy white matter (WM) necessitates a careful approach during surgical treatment. While tractography is clinically widely used in a qualitative manner to visualise WM bundles during surgery planning, quantifying the integrity of such bundles in glioma patients is a difficult challenge. Quantitative measures currently used in group-level studies are biased or heavily dependent on tractography parameters. In most populations, Anatomically Constrained Tractography (ACT) in combination with Spherical-deconvolution Informed Filtering of Tractograms (SIFT) successfully estimates WM connectivity in whole-brain tractograms. However, a critical segmentation step used in ACT typically overestimates the presence of gray matter (GM) in glioma patients due to the properties of glioma tissue on T1-weighted MRI. This leads to an incorrect derivation of the interface between GM and WM, which is used to define a seed and target region of interest (ROI) in ACT. While workarounds exist, we argue that these may bias SIFT by either under- or overestimating fiber reconstructions in and around the tumor. In this work, we present an alternative GM-WM segmentation method based on diffusion-weighted MRI using multi-tissue Constrained Spherical Deconvolution (CSD) and derive GM-WM interfaces from these novel segmentations. We demonstrate that this method results in a more accurate GM-WM interface in and around tumors in three glioma patients, and that the resulting quantitative tractograms better identify WM disconnections induced by surgery. This method represents a step towards quantitative tractography in glioma patients, and towards improved surgical decision-making in clinical practice.