Real-time MRI scanning mid-craniotomy to correct for brain shift and confirm extent of tumor resection before closing
Every intraoperative MRI (iMRI) case begins with a meticulous preoperative study on a diagnostic 3T scanner: high-resolution anatomical sequences to delineate tumor margins, functional MRI (fMRI) to localize motor, language, and visual cortex, and diffusion tensor imaging (DTI) tractography to trace white-matter highways like the corticospinal tract and arcuate fasciculus. This baseline becomes the surgical roadmap loaded into the navigation system — and, critically, the reference point against which all later intraoperative distortion will be measured.
The preoperative workup for an eloquent-area tumor typically layers three imaging modalities onto a single 3D anatomical volume. First, a volumetric T1-weighted gadolinium-enhanced sequence (MPRAGE or SPGR, 0.9–1.0mm isotropic) defines the enhancing tumor core, while T2-FLAIR captures the surrounding non-enhancing infiltrative margin — critical for low-grade gliomas, which rarely enhance but diffusely infiltrate white matter well beyond visible T1 abnormality. Second, blood-oxygen-level-dependent (BOLD) functional MRI, acquired while the awake patient performs finger-tapping, verb-generation, or naming tasks, localizes eloquent cortex within a few millimeters — though fMRI activation maps are known to shift by 5–10mm relative to direct cortical stimulation "gold standard" mapping, a limitation surgeons must respect with margin.
Third, diffusion tensor imaging reconstructs white-matter tractography: deterministic or probabilistic fiber-tracking algorithms follow the direction of water diffusion anisotropy voxel-by-voxel to render 3D streamlines of the corticospinal tract, optic radiations, and arcuate fasciculus. A tumor sitting within 8mm of the corticospinal tract carries substantially elevated risk of postoperative hemiparesis; tractography lets the surgical team quantify that margin numerically rather than by visual estimation alone. All three datasets — anatomy, function, connectivity — are then fused into a single navigation volume loaded onto the intraoperative neuronavigation workstation (StealthStation, Brainlab, or similar), registered to the patient's head using either a rigid skull frame or surface-matching with a laser/optical wand after induction.
The entire baseline dataset, however, is acquired with the patient's head fixed, dura closed, and cerebrospinal fluid at physiological pressure and volume — conditions that will begin changing within minutes of skin incision.
The instant the dura is opened, the brain is no longer a fixed structure sealed inside a rigid, fluid-filled cranium — it is a soft, gravity-responsive organ suddenly exposed to atmospheric pressure. Cerebrospinal fluid egress, gravitational sag, mannitol-induced brain relaxation, and progressive tissue removal during debulking combine to physically displace the brain relative to its preoperative position, a phenomenon called "brain shift" that silently erodes the accuracy of neuronavigation as surgery proceeds.
Brain shift is driven by a combination of mechanical and physiological factors that begin acting the moment the dura is incised. CSF loss is the dominant early driver: as fluid drains from the subarachnoid space and ventricles, the brain loses its buoyant support and settles under gravity, an effect amplified by patient positioning — a lateral or prone craniotomy produces more pronounced shift than supine. Nimsky and colleagues' foundational intraoperative MRI studies (Neurosurgery, 2000–2001) quantified cortical surface displacement of up to 24mm during glioma surgery, while deeper structures near the tumor bed and ventricular system shift comparatively little (1–3mm) because they are tethered by falx, tentorium, and vascular pedicles — a gradient that makes shift correction geometrically nonuniform and impossible to compensate with a simple rigid transform.
Additional contributors compound the effect: osmotic diuretics (mannitol, hypertonic saline) given to relax the brain for exposure reduce total brain volume by up to 10%; tumor debulking itself removes mass and collapses the resection cavity, causing surrounding parenchyma to shift inward to fill the void; and gravity continues to act throughout the case, so shift is not a one-time event but a continuously evolving process — studies show measurable progression in cortical position even 60–90 minutes into resection. Within 20–30 minutes of dural opening, image-guidance error from brain shift alone can already exceed 5mm; by the time a large tumor has been substantially debulked, navigation accuracy referenced to the preoperative scan can be off by 1cm or more — an error large enough to make the surgeon either dangerously over-confident near eloquent tissue or unnecessarily conservative, leaving tumor behind.
This is precisely the problem intraoperative MRI was engineered to solve: rather than trusting an increasingly stale map, the surgical team can pause and re-image the brain in its actual, current, shifted configuration.
When the surgeon needs an updated picture, the operating room itself becomes an MRI suite. Three major architectural approaches have emerged over two decades of iMRI engineering: fixed high-field magnets built directly into a dedicated OR (BrainSUITE-style 1.5T/3T systems), low-field portable units that dock beside the table (PoleStar N-series, 0.15–0.5T), and mobile high-field magnets on ceiling rails or floor tracks that travel between adjoining operating rooms (IMRIS-type systems).
The earliest clinical iMRI systems, such as the PoleStar N20/N30 (Odin Medical/Medtronic), used compact 0.12–0.15T permanent magnets that could be wheeled beside the operating table and raised around the patient's head mid-procedure. Their low field strength sacrifices image resolution and contrast compared to diagnostic scanners, but the tradeoff buys speed, low cost, and minimal room modification — a scan can be acquired in a few minutes without moving the patient at all, making PoleStar-class systems attractive for smaller centers doing biopsy confirmation or coarse residual-tumor checks.
At the opposite end, fixed high-field suites such as the BrainSUITE (BrainLAB/Siemens, 1.5T or 3T) integrate a full diagnostic-grade cylindrical magnet into a dedicated, specially shielded operating room. The patient is either transferred on a compatible non-ferromagnetic table into the bore, or — in most modern designs — the magnet itself rotates or translates on a ceiling-mounted rail system (IMRIS/Deerfield Imaging architecture) to travel over the sterile field and scan the patient in place, then retract back into a shielded parking bay for the next case. Because the magnet can serve two or three adjoining operating rooms on the same rail, capital cost per case is amortized across higher throughput. These fixed high-field systems deliver full diagnostic image quality — including gadolinium-enhanced sequences, DTI reacquisition, and functional confirmation — at the cost of a 20–40 minute workflow interruption per scan cycle: the surgical field must be covered with a sterile, MR-safe drape, all ferromagnetic instruments cleared from the 5-Gauss line, the patient repositioned relative to the coil, and the team steps back for acquisition before resuming surgery.
A typical glioma case acquires one to three intraoperative scans: an early "sanity check" after initial debulking, and a final confirmatory scan before presumed closure. Each reacquisition costs real time and expense — total added OR time for iMRI-equipped cases runs 30–90 minutes longer than conventional navigation alone — a cost surgical teams accept because the alternative is closing on unrecognized residual tumor.
A freshly acquired intraoperative scan is only useful once it is co-registered with the surgical navigation system and interpreted correctly. Fusing the new iMRI volume onto the live patient reference frame lets the surgeon re-localize instruments in real anatomical space — but the harder task is visual: distinguishing genuine residual tumor from the blood products, edema, and resection-cavity artifact that flood the surgical bed after debulking, a discrimination problem that intraoperative MRI, and complementary tools like 5-ALA fluorescence, were built to solve.
Once the updated volume is acquired, it must replace — or better, deformably warp — the stale preoperative dataset inside the navigation system. Simple rigid re-registration (matching skull-fixed fiducials or surface contours) corrects for whole-head repositioning but cannot account for the nonuniform brain shift described earlier; modern platforms increasingly apply deformable (non-rigid) registration algorithms that model the brain as an elastic or biomechanical volume, warping the preoperative tractography and functional maps onto the new, distorted anatomy so that eloquent-region overlays remain accurate even after significant tissue shift. Achieved fusion accuracy is typically 1–2mm, versus the 5–24mm of uncorrected error the scan was acquired to fix.
Interpreting the new images requires experience: the resection cavity is filled with blood products, saline irrigation, hemostatic material (Surgicel, thrombin-soaked gelfoam), and reactive edema — all of which can enhance or restrict diffusion in ways that superficially resemble residual tumor on standard sequences. Neuroradiologists cross-reference T1 gadolinium enhancement patterns (thin linear rim enhancement favors normal postsurgical change; thick nodular or irregular enhancement favors tumor), diffusion-weighted imaging, and comparison against the pre-resection tumor geometry to make the call. Studies comparing iMRI detection against final histopathology report sensitivity of 85–95% for clinically significant residual tumor, compared with roughly 50–60% for the operating surgeon's naked-eye assessment through the microscope alone — meaning iMRI catches genuine residual disease that even experienced neurosurgeons cannot see or palpate, but it also generates a false-positive rate of 5–15% where reactive change is mistaken for tumor, occasionally prompting resection of normal tissue.
A widely used complementary technology is 5-aminolevulinic acid (5-ALA) fluorescence-guided resection: patients drink a 5-ALA solution 2–4 hours preoperatively, which is metabolized preferentially by malignant glioma cells into fluorescent protoporphyrin IX, visible as pink-red fluorescence under a modified blue-light microscope. The landmark Stummer et al. randomized trial (Lancet Oncology, 2006) showed 5-ALA fluorescence-guided surgery increased complete resection of contrast-enhancing tumor from 36% to 65% and improved 6-month progression-free survival from 21% to 41% in malignant glioma. 5-ALA and iMRI are complementary rather than competing: fluorescence gives instantaneous, high-resolution optical feedback at the resection margin in real time, while iMRI provides volumetric, whole-brain confirmation and corrects for brain shift that fluorescence cannot detect — many contemporary iMRI suites use both technologies together.
The entire iMRI workflow exists to answer one question before the surgeon closes: is there more tumor left that can be safely removed? Across two decades of clinical experience and multiple randomized and prospective cohort studies, the answer to "does checking matter" has become unambiguous — intraoperative MRI-guided resection meaningfully increases gross-total-resection rates in glioma surgery, and greater extent of resection is one of the most consistently reproduced prognostic factors for both progression-free and overall survival.
The strongest evidence for iMRI comes from Senft and colleagues' randomized controlled trial (Lancet Oncology, 2011): 58 patients with contrast-enhancing glioma were randomized to iMRI-guided resection versus conventional neuronavigation-guided resection alone. The iMRI arm achieved complete tumor resection in 96% of patients versus just 68% in the control arm — a striking absolute difference — and iMRI-guided patients also showed significantly longer progression-free survival, with no increase in new postoperative neurological deficits between groups, directly rebutting the concern that more aggressive image-guided resection might come at the cost of greater morbidity.
Kubben and colleagues' systematic review and meta-analysis (Lancet Oncology, 2011) pooling multiple iMRI cohort and randomized studies confirmed the pattern at scale: intraoperative MRI use was consistently associated with higher rates of complete resection compared with conventional surgery, particularly for low-grade gliomas where tumor margins are indistinct from normal brain on visual and tactile inspection alone. In roughly 29% of cases across published series, the intraoperative scan reveals unsuspected residual tumor that the surgeon believed had already been completely removed — prompting a second, targeted resection pass before final closure. This "second look" capability is the single largest driver of iMRI's benefit: it converts what would otherwise be an unrecognized subtotal resection into a genuine gross-total resection.
Extent of resection matters because it is one of the few surgically modifiable prognostic factors in glioma treatment. Multiple large retrospective cohorts (Sanai & Berger, McGirt et al.) demonstrate that each additional percentage point of tumor removed above roughly 78–90% resection thresholds is associated with incrementally longer survival, and that patients achieving gross-total resection gain a median overall survival benefit on the order of several months to just under a year compared with those left with residual enhancing tumor — a magnitude of benefit comparable to, and additive with, adjuvant chemoradiation. Because iMRI increases the rate at which true GTR is achieved without raising the rate of new permanent neurological deficit, it has become standard of care at high-volume neuro-oncology centers for surgically accessible, eloquent-adjacent gliomas.
In the Senft et al. 2011 randomized trial, gross-total resection was achieved in 96% of patients in the intraoperative MRI arm versus only 68% in the conventional-navigation control arm — and this near-30-point jump in complete resection came with no increase in new neurological deficits, directly translating into significantly longer progression-free survival for the iMRI group.