Introduction
Pre-operative task-based functional MRI (tfMRI) is often included in the battery of imaging studies for decision-making and operative planning to aid surgical resection of brain neoplasms, balancing long-term survival by maximizing the extent of resection while preserving the patient’s functional status^1-4. However, dependence on patient’s task-compliance, which may be impaired due to disability, and the limited number of areas that can be mapped due to clinical time constraints impact the clinical utility of task-based fMRI (tfMRI). Resting state functional MRI (rsfMRI) is a promising task-free approach to delineate eloquent cortex^5-11, which may complement and potentially replace tfMRI in patients who have difficulties performing required tasks^12,13. However, rsfMRI is highly sensitive to head movement and physiological noise, and validation relative to tfMRI and intra-operative electrocortical mapping is still necessary^14,15. In this study, we investigated the feasibility of presurgical mapping in patients with brain tumors comparing high-speed multi-band EPI (MB-EPI), which enables real-time rsfMRI analysis with online monitoring of data quality¹⁶, and recently developed high-speed multi-slab and multi-band echo-volumar imaging (MS-EVI, MB-EVI), which increases volume coverage and/or temporal resolution compared with multiband EPI. The objective was to map sensorimotor and language resting state networks in the vicinity of brain tumors and to validate this approach in comparison with tfMRI and intra-operative electrocortical-stimulation (ECS).
Methods
RsfMRI (eyes open, 5-10min) and tfMRI (3min per task) data were acquired in 9 patients (4F,5M) with brain tumors (2 glioblastomas, 2 gangliogliomas, 1 anaplastic astrocytoma, 3 oligodendrogliomas, 1 low grade glioma) as part of a multi-modal MRI protocol that included DTI and high-speed 3D proton-echo-planar-spectroscopic imaging using a 3T scanner equipped with 32-channel head coil. Informed written consent was obtained. Tasks were tailored to tumor location: motor (finger-tapping/fist-clenching), visual (eyes open/close), auditory (listening to syllables) and language (word/verb generation). Real-time fMRI was performed using MB-EPI (TR/TE: 400/35ms, multiband-acceleration:8, flip-angle:42o, voxel-size:(3mm)3, 32slices) and MS-EVI (TR/TE:246/30 ms, no.slabs:4, flip-angle:20o, voxel-size:(4mm)3, 29slices). MB-EVI (TR/TE:400/34ms, no.slabs:8, multiband-acceleration:2, flip-angle:40o, voxel-size:(3mm)3, 57slices) was reconstructed offline. Real-time and offline seed-based connectivity analysis (SCA) of 2 RSNs and task-based correlation analysis were performed using the TurboFIRE tool^16,17. SCA was performed with 8s moving average temporal low pass filter, 15s sliding window correlation analysis with running mean, and sliding window regression of 6 rigid body motion parameters and signal time courses from WM and CSF ROIs. RSNs were mapped using unilateral Brodmann area (BA) based seed regions (sensorimotor: SMN–BA01-03, language: LAN-BA44,45 (Broca’s) and LAN-BA22,39,40 (Wernicke’s). Motion parameters, true/false positive connectivity, and task-activation were monitored online. Image guidance (Stealth Station Treon and S8; Medtronic Navigation, CO, USA) for the surgical approach was based on clinical task-based presurgical fMRI obtained separately. Intraoperative brain mapping was performed using electro-cortical stimulation (ECS) and continuous electrocorticography. ECS locations were mapped onto stereotactic MRI. The Euclidean distances between the peaks and edges of rsfMRI connectivity and task-activations with respect to intra-operative ECS were measured in motor and language areas.
Results
Real-time monitoring of data quality enabled rapid optimization of scan protocols and seed selection, early identification of task non-compliance, and head movement related false-positive connectivity in white matter and at slice edges in all patients (Figure 1). Sensorimotor and language resting-state networks stabilized within 3-4 mins. A unilateral seed in BA1-3 contra-lateral to the tumor showed bilateral connectivity in primary motor areas with correlations ranging from 0.5 to 0.7 and in supplementary motor area (SMA) with correlations ranging from 0.3 to 0.5 (Figure 2). Resting-state language connectivity in Broca’s area was detected in all patients, predominantly left-lateralized and usually weaker than motor connectivity, necessitating a lower threshold of 0.3. Connectivity in Wernicke’s area was not detected in all patients, which may in part reflect cortical reorganization and/or impaired neurovascular coupling in these patients due to the proximity of the tumor. In patient 3, unexpected additional language connectivity in anterior insula was detected, which was retrospectively associated with a post-operative language deficit after resection. The Euclidean distance between the peaks of intra-operative ECS and rsfMRI connectivity and the peaks of ECS and task-activation in motor cortex, Broca’s and Wernicke’s areas in the first 5 patients was 5-13mm, except for discordant rsfMRI localization of Wernicke’s area in patient 3 due to possible altered neurovascular coupling and/or cortical reorganization (Table 1). Data analysis in patients 6-9 is in progress. Preliminary results show increased BOLD sensitivity of MB-EVI compared with MB-EPI (Figure 3).
Discussion
The data in patients show a high degree of consistency between resting state connectivity, task-based activation and ECS localization of motor cortex and Broca’s area. Localization of Wernicke’s area was more variable, both in rsfMRI and tfMRI, and in one case discordant with intra-operative ECS localization, which may in part reflect cortical reorganization. Limitations of this study include small number of patients studied and atlas-based seed selection, which needs to be adapted to possible cortical reorganization due to neuroplasticity. Future work will include integration of rsfMRI with tfMRI, DTI-based localization of peritumoral fiber tracts and MR spectroscopic imaging biomarkers of tumor borders to predict functional and oncological outcomes (Figure 4).

Acknowledgements

We gratefully acknowledge our patients for their time and effort participating in this study. Howard Yonas (Neurosurgery) graciously supported the initial phase of this project and provided guidance during the implementation of the research protocol. Dr. Brad Cushnyr conducted clinical fMRI scans. We thank Mona Chaney (Neurology) and Erin Semler (Neurosurgery) for their assistance with patient recruitment and Stacy Steadman (Neurosurgery) for facilitating data collection in the operating room. Kimball Malherbe (Medtronic Navigation, Inc) provided support for the Stealth workstation. We thank Catherine Smith and Diana South (Mind Research Network) for their expert support of MRI scanner operations.