A Hidden Signal in Brain Surgery Data Could Make Tumor Removal Safer

Every few years, a surgical technique that has worked reasonably well for decades turns out to be hiding data that nobody thought to look for. Awake brain mapping - the procedure in which neurosurgeons electrically stimulate the brain while a conscious patient names pictures or reads aloud - has been used for roughly 30 years to identify language regions before removing tumors. What a new study from Carnegie Mellon University suggests is that those procedures have been generating richer data than anyone realized.

The study, to be published in Science Advances, examined not just whether patients made errors during brain stimulation but how fast they responded even when they answered correctly, and precisely when during a task the stimulation was applied. Those two variables, largely ignored in standard clinical practice, turned out to predict language organization with measurably greater accuracy than a simple error-or-no-error framework.

How Awake Brain Mapping Works

As a brain tumor grows, cancerous cells rarely stay neatly contained. They infiltrate adjacent tissue that may appear healthy, forcing surgeons to decide, region by region, how much to remove. Remove too little and cancer returns faster; remove too much and the patient may lose the ability to speak, move, or recognize faces.

During awake craniotomy, surgeons deliver brief electrical pulses to the exposed brain surface while the patient names objects or reads words. If stimulation disrupts a response - causing hesitation, slurring, or complete failure to respond - that area is flagged as functionally important and left alone. If behavior is unaffected, the tissue is considered safer to remove. The patient feels nothing; the brain has no pain receptors.

"Stimulation has traditionally been treated as a binary test - either it causes an error, or it doesn't," said Raouf Belkhir, lead author and a psycholinguist completing the University of Pittsburgh-Carnegie Mellon Medical Scientist Training Program. "But in reality, these effects are often more continuous than binary."

The Signal Hidden in Response Time

Bradford Mahon, a cognitive neuroscientist at CMU's Neuroscience Institute and senior author of the study, and his team analyzed data accumulated over a decade of awake surgeries. They found that physical parameters of electrical stimulation - its duration, and precisely when it began relative to the patient's task - predicted small but consistent changes in response speed, even when no overt error occurred.

"We found that if you measure both the types of errors that patients make, as well as how fast they respond even when they do not make errors, more granular inferences can be drawn about language organization from an awake brain mapping procedure," Mahon said.

A brain region that slows a patient's response by 80 milliseconds without triggering an error is different from one that causes complete speech arrest. Those gradations, the team argues, contain clinically meaningful information about how centrally a given region is wired into the language network - information that a binary pass-fail system discards entirely.

What This Means for Surgical Decision-Making

The practical application is the ability to build predictive models personalized to each patient before the surgeon makes a final cut. If such models can simulate the cognitive consequences of removing a given piece of tissue, patients can make more informed decisions about acceptable trade-offs.

Tyler Schmidt, MD, a neurosurgeon at the University of Rochester who has used a software platform spun out of Mahon's laboratory - called MindTrace - in more than a dozen awake surgeries, described the shift in thinking: "In the beginning of brain tumor surgery, it used to be, 'Can we remove any of this tumor safely?' But now in some brain tumor cases it's, 'Can we get you back to work potentially?'"

MindTrace is currently working with a consortium of six U.S. hospitals to build a longitudinal dataset of patient outcomes. The goal is to train models that, given a patient's pre-surgical profile and real-time stimulation data, can forecast the cognitive consequences of different surgical boundaries with quantifiable confidence.

Limitations and Next Steps

The study's findings are based on existing clinical data, not a prospective randomized trial. The optimal parameters for maximizing stimulation effects - duration, timing, intensity - remain unknown, and translating the laboratory findings into standardized clinical protocols will require validation across multiple surgical teams and patient populations. Awake brain mapping is itself highly variable: some patients tolerate the procedure poorly, and some tumors sit in regions too close to critical structures for extensive stimulation.

The team also does not yet know whether the granular response-time data will consistently translate into better oncological outcomes - longer survival, less recurrence - or primarily into improved cognitive preservation. Those are related but distinct questions, and answering them will take years of prospective data collection. What the study establishes is that the data being generated in awake surgeries right now is more informative than current practice extracts from it.