Evaluation of Serial Diffusion Tensor Imaging to Reveal Subclinical Tumor Prior to Detection on Anatomical Imaging in Patients with Glioblastoma

Abstract

As the majority of patients with glioblastoma (GBM) experience tumor recurrence, early detection of recurrence can facilitate timely adaptation to treatment in order to improve outcomes. We hypothesize that the presence of tumor cells can result in microstructural disruptions in white matter (WM) integrity that can be visualized using DTI prior to tumor detection on conventional, anatomical MR imaging. This retrospective study aimed to evaluate whether changes in serial diffusion tensor imaging (DTI) parameters preceded and localized to the area of GBM tumor recurrence. Under an IRB approved protocol, patients with recurrent GBM following RT who had serial DTI at tumor recurrence and at least 2 prior imaging sessions were included. The serial images, acquired about 1-3 months apart, were gathered from 2014 to 2017. The following images were analyzed: post-contrast T1-weighted (T1w) MRI and DTI at tumor recurrence (t0), the immediate prior scan (t0-1), and the second prior scan (t0-2) to recurrence. Diffusion parameter maps, fractional anisotropy (FA) and mean diffusivity (MD), were extracted. The diffusion maps at recurrence were co-registered with those at t0-1 and t0-2. The T1w images at each time point were aligned with their corresponding diffusion maps. An experienced radiation oncologist manually segmented recurrent gross tumor volume (rGTV) for each patient on T1w images. The rGTV was co-registered to diffusion maps at each time point to form t0-1GTV and t0-2GTV. Mean FA and MD values were calculated for rGTV (t0), t0-1GTV, and t0-2GTV. Qualitative evaluation of abnormal regions on diffusion maps and anatomical images were compared. Pairwise t-tests were used to compare features among the 3 time points to examine WM changes over time. A total of 24 GBM patients (age: 26-77 years; 11 males:13 females) were included in this analysis. The mean FA values of t0-2GTV were statistically lower than those of t0-1GTV and rGTV (t0) (p=0.04 and p=0.02, respectively). The mean FA values of t0-1GTV and rGTV were not statistically different. The mean values of MD across all time points were not statistically different. Qualitatively, structural changes were visible on the FA maps in 70% of cases prior to visible changes on the T1w images. In this retrospective study, FA changes were observed 1-3 months before clinically detected tumor recurrence on anatomical imaging. The FA values were similar between recurrent and immediate pre-recurrent DTI images, suggesting that underlying progressive micro-structure changes may have been present prior to the appearance of enhancing tumor. In our study, FA appeared more sensitive than MD at detecting subclinical tumor presence. Future investigation of a larger cohort and more detailed analysis is planned to evaluate the role of DTI in detection of subclinical GBM in normal appearing brain to potentially refine clinical target volume definition for RT planning and prediction of GBM recurrence.