Project Summary
Functional localization of eloquent brain areas for patients undergoing surgery for brain tumors, epilepsy, or other
neurological diseases is crucial to prevent post-surgical deficits and reduce morbidity. Task-based (tb) functional
MRI (fMRI), which detects blood oxygenation level–dependent (BOLD) signal changes while a patient performs
task paradigms, is a standard-of-care clinical procedure for presurgical mapping of eloquent cortices. Two major
limitations of clinical tb-fMRI are a patient's inability to perform the task and lesion-induced impairment of
neurovascular coupling (which drives the BOLD signal). Resting-state (rs) fMRI, which measures synchronized
BOLD signal oscillations during rest, can be used to map brain networks with minimal requirements for patient
compliance and has been demonstrated to accurately localize motor and language areas for presurgical planning.
Cerebrovascular reactivity (CVR) mapping, accessed by dynamic BOLD imaging during a hypercapnia task such
as breath-holding, can be used to identify areas with potential false-negative fMRI results due to neurovascular
uncoupling (NVU) and has been suggested as an emerging standard to be used with clinical fMRI. Currently,
there are no commercially available FDA-cleared software tools for localizing the resting-state networks (RSNs)
or CVR. Clinical investigators have relied on research software packages that are either not clinically integrated
or not yet optimized and validated in large patient populations. Thus, a vetted software solution is urgently
needed to enable these state-of-the-art fMRI methods to benefit patients beyond the limitations of tb-fMRI. We
hypothesize that enhancing, optimizing, and validating our preliminary software and integrating it with an
established commercial fMRI platform will create robust solutions for clinical mapping of RSN and CVR. Through
three specific aims, the software solutions will be optimized and tested with rs-fMRI and CVR datasets from
approximately 350 patients with brain tumors or epilepsy at three institutions. Aim 1 is to create the software for
mapping RSNs and determine optimized workflows for localizing eloquent areas including primary visual, motor
(hands, tongue, and feet), and language (primary and secondary) areas. Both seed-based correlation and
independent component analysis will be incorporated. Aim 2 is to create the software for mapping CVR and
determine the optimized workflow for identifying and visualizing brain areas with potential false-negative fMRI
results. The software will include a multiple-latency general linear model and a unique graphical user interface
to visualize the NVU. Aim 3 is to test and validate the software with presurgical fMRI datasets. The results will
be compared against those obtained from (1) processed using widely used research software packages, (2) tb-
fMRI, and (3) intraoperative direct cortical stimulation. This research is anticipated to create robust and clinically
available software that will greatly increase the patient population who can benefit from presurgical fMRI and will
improve confidence in functional localization for surgical planning. This will directly benefit patients by preserving
their post-surgical functions while allowing surgeons to safely maximize the resection of brain lesions.