Resting-State Versus Breath-Hold for Testing Functional Magnetic Resonance Imaging Integrity in Pre-Surgical Planning

Abstract

The neuro-vascular coupling mechanism responsible for blood oxygenation level dependent (BOLD) functional magnetic resonance imaging (fMRI) signals can be focally disrupted by various pathological factors (such as brain tumors) while leaving the underlying neurons functionally intact. Such neuro-vascular uncoupling (NVU) can pose serious complications for clinical use of fMRI. Mapping of cerebrovascular reactivity (CVR), which is a measure of the dilatory function of cerebral vasculature, can be a useful approach for detecting potential NVU. The widely-accepted approach for non-invasive CVR mapping requires the patient to perform a breath-hold challenge, which may have practical disadvantages for many patient populations. Thus, a clinical need for CVR mapping with fewer (if no) practical limitations could benefit surgeon’s decision-making. Resting-state fMRI has great potential to fill this need. To assess the feasibility for the use of resting-state CVR maps as an alternative to breath-hold CVR maps for identifying regions of likely NVU, we investigated their three normative characteristics at a voxel level of resolution: (1) quality of spatial labeling of the critical brain areas, (2) continuity and uniformity, and (3) validity for clinical use. In all three phases of this study, we first thresholded the CVR maps by optimizing coverage of gray matter while minimizing false responses in white matter. When so optimized, the resting-state vs. breath-hold CVR maps had moderately better gray matter coverage and specificity but our ability to statistically specify a hole as a site of potential NVU was comparable across approaches. We also assessed the voxel-wise spatial correspondence between the two maps across a wide range of thresholds. Optimal spatial correspondence was strongly dependent on threshold settings, which if improperly set, tended to produce statistically biased maps. While the two CVR maps were not quantitatively identical at a voxel level of resolution when optimized, they did consistently have moderately good correspondence with each other. To test the validity of our approach, we retrospectively evaluated the resting-state CVR maps of a cohort of 13 patients with brain tumors and assessed how reliably these maps would reveal foci of potential NVU in pre-surgical task-fMRI activation maps. The impaired resting-state CVR maps spatially corresponded to NVU-induced impairment of task-fMRI activation patterns in four patients with suspected NVU. While NVU-related false negatives on task-based fMRI activation maps were characterized with holes in resting-state CVR maps, our results also provided evidence that tumor-induced NVU could occur with no apparent effects on vasoreactivity in one patient. From the standpoint of clinical utility, the use of resting-state CVR maps with significantly more specific labeling of critical brain areas may provide a comparative advantage over the breath-hold CVR maps for testing the integrity of task-fMRI activation. A statistically comparable degree of sensitivity between the two CVR maps in detecting potential NVU may also offer the use of resting-state approach due to its advantages such as the ease of use and computation. In sum, our results take a step towards using the resting-state-based CVR mapping as an imaging marker in the clinical setting to identify fMRI signal dropouts caused likely by NVU, for which the clinicians should be warned to take extra care when interpreting the fMRI brain maps, particularly in surgical scenarios.