Three-dimensional (3D) advanced imaging and spectroscopy techniques acquire volumetric anatomical and physiological data, yet most common visualization techniques are limited by projecting these 3D data into multiple 2D projections. 2D visualization can be especially limiting in the context of neurosurgery, where patient orientation may deviate from conventional orthogonal projections. In addition to accurately representing the extent of a brain tumor, spatially co-registered 3D visualization can be used to guide and monitor location-sensitive treatments, such as the Tumor Treating Fields (TTFields). TTFields is a novel technique that employs a worn transducer array to deliver alternating electrical fields aimed to disrupt mitosis, with field intensity optimized for tumor size and location. By co-registering volumetric data across multiple time-points and modalities, particularly physiological MRI sequences like perfusion and spectroscopy, a more comprehensive picture of a tumor’s true spatial extent and long-term response to therapy can emerge. This study proposes the development of a 3D image visualization, co-registration and quantification platform to improve spatial characterization of brain tumors and assist with both the planning of treatment and long-term response assessment.

High-resolution, volumetric datasets were obtained from two patient cohorts undergoing separate treatments for brain tumor [surgery (n=2) and TTFields (n=4)]. Two patients were scanned one day prior to neurosurgery while a separate group of four patients undergoing TTFields therapy were scanned prior to the first use of the transducer array and then at one-month intervals for a maximum of six months. Both cohorts were scanned using an advanced imaging protocol consisting of standard of care contrast-enhanced T1-weighted imaging (TE=3.1ms, TR=1760 ms, 250x188mm² field-of-view (FOV), 192 slices, 3:10min), T2 fluid attenuated inversion recovery (FLAIR) (TE=141 ms, TR=9420ms, 240x180mm² FOV, 60 slices, 3:10min), diffusion tensor imaging (DTI) (30 directions, TE=86 ms, TR=5000 ms, 3 avg, 220x220mm² FOV, 40 slices, 8:00min) and perfusion weighted imaging (PWI) (TE=54 ms, TR=2000 ms, 220x220mm² FOV, 20 slices, 1:38min) along with whole-brain echo-planar spectroscopic imaging (EPSI) (TE=17.6ms, TR=1550ms, 280x280x180mm³ FOV, 64x64x32 array size, 15:00min). All scans were performed using a 12-channel head coil on a Siemens Tim-Trio scanner. The full protocol’s total scan time was 45 minutes.

A Matlab-based 3D visualization platform was developed to read, display, co-register and quantify data from multiple modalities. The program reads DICOM or analyze format data and performs semi-automated co-registration by matching spatial parameters from data headers and anatomical landmarks. Segmented regions-of-interest (ROIs) were drawn semi-automatically by manually circumscribing the region of T2-FLAIR abnormality using MRIcro and thresholding T1-weighted data to identify contrast enhancement as intensity three standard deviations above that of healthy white matter. Regions of high choline to N-acetylaspartate (Cho/NAA) and choline-to-creatine (Cho/Cr) in the EPSI data were identified using threshold values of 0.6 and 1.2 (ratios of absolute concentration), respectively. Custom and native imaging macros were used in Matlab to project the data in three dimensions with a GUI-based window to freely rotate/zoom the projections in 3D. Datasets were represented as segmented 3D ROIs placed within individual imaging slices, which were then projected within the skull or brain (see Figures). A frame-capturing macro facilitated recording the 3D rotation of the datasets into individual images or high-resolution animations. Because each dataset was co-registered, including across multiple time points, the ROI volumes and slice data could be taken from any of the scanned imaging modalities.

The Matlab-based visualization platform successfully facilitated co-registration and visualization of datasets across multiple modalities and time-points. Figure 1 shows T1-weighted imaging highlighting the contrast-enhancing volume in red. Figures 2 and 3 show co-registered T2-FLAIR and cerebral blood volume (CBV) map taken from the same brain tumor patient (glioblastoma, WHO Grade IV) after 3 months of TTFields therapy. Data quantification showed a slight decrease in enhancing volume (5.8ml follow-up vs. 8.4ml baseline) and median rCBV (1.59 vs. 1.76 baseline). Spectroscopic data showed slight increase vs. baseline in Cho/NAA (1.07 vs 0.97) and Cho/Cr (0.83 vs. 0.66) in co-registered contrast enhancing regions. Figure 4 shows an NAA/Cho map generated from an EPSI study of a brain tumor patient (high grade glioma) prior to surgery. Co-registration with anatomical imaging data revealed a region of elevated Cho/NAA and Cho/Cr ratios corresponding to the region of T1-contrast enhancement and FLAIR signal abnormality. Co-registered PWI showed areas of elevated CBV within the T2-FLAIR abnormality as shown in Figure 5.We believe that additional functional and metabolic information as obtained from this study can be helpful for neuro-navigation, guided biopsies, radiation planning, and predicting recurrence.