MRI scans predict recovery from spinal cord injury

Making a prognosis for spinal cord injury has been a guessing game, but a neuroimaging study by Stanford Medicine scientists and collaborators finds answers hiding in plain sight.

- By Nina Bai

The width of spared spinal cord tissue at the injury site allowed Stanford Medicine researchers and their colleagues to predict a patient’s improvement on motor and sensory function tests.
Dario Pfyffer

Neck-level spinal cord injuries are among the most devastating, and recovery is difficult to predict. Some patients live with lasting paralysis of all four limbs, while others regain the ability to walk.

A new, international study by Stanford Medicine researchers and collaborators has found that a simple measurement taken from standard MRI images can more accurately predict a patient’s recovery.

A single metric — the width of spared spinal cord tissue at the injury site — allowed researchers to project a patient’s improvement on tests of motor and sensory function.

“Our model can tell you, depending on how many millimeters of tissue that’s spared, how much function they’re likely to gain during a follow-up period of three months or 12 months,” said Dario Pfyffer, PhD, postdoctoral scholar in anesthesiology, perioperative and pain medicine and lead author of the new study published June 27 in The Lancet Neurology.

Such insights will not only aid clinicians in counselling patients and choosing treatments and rehabilitation approaches, but also help researchers sort patients into groups with similar recovery potential — a critical first step in testing experimental therapies for spinal cord injuries.

“When you’re testing a new drug, you want to make sure that you’ll be able to distinguish spontaneous recovery from recovery induced by the drug,” Pfyffer said.

The senior author of the study is Patrick Freund, MD, PhD, professor of experimental imaging of the spinal cord at the University of Zurich.

A flawed standard

Soon after a patient with spinal cord injury is admitted to the hospital, clinicians assess the patient’s sensorimotor function — testing muscle movement and touch sensation throughout the body. These initial test scores are the gold standard for forecasting recovery, but they often miss the mark. Scoring can be subjective and affected by other conditions — a fractured arm, for example — and doesn’t capture nuances of the injury.

Dario Pfyffer

“We’ve always been limited in predicting recovery in these patients,” Pfyffer said.

The research team realized that better clues to recovery could be hiding in plain sight — in the MRI scans commonly performed in these cases.

“MRIs are often assessed after spinal cord injury, but it’s mostly to look at the damage and to guide surgery, not for outcome prediction,” Pfyffer said.

The researchers knew that MRIs not only visualized the damage, but also the remaining spinal cord tissue around the injury site, known as tissue bridges.

“We can get a measure of how much spinal cord tissue is spared and functional and still transmitting information from the spinal cord to the brain and back,” Pfyffer said. “The more tissue spared, the better we would expect the recovery to be.”

Predicting recovery

In the new study, researchers retrospectively analyzed data from 227 patients admitted to three hospitals in Denver, Colorado; Murnau, Germany; and Zurich, Switzerland. The patients had cervical spinal cord injuries and no other neurological or psychiatric conditions.

Because the researchers wanted to look at spontaneous recovery, they included only patients who were not enrolled in any clinical trials and receiving only conventional care.

They measured the width of tissue bridges from MRI scans taken three to four weeks after injury — when the spinal cord has stabilized and any initial swelling has subsided.

When they compared the width of tissue bridges with the patients’ sensorimotor test scores at hospital admittance and at discharge three months later, they found a clear pattern. From these data, they developed a statistical model that could predict how many points a patient was likely to gain on specific sensorimotor tests. The model accounts for the patient’s age, sex, injury location and starting test scores, as well as differences between hospitals.

Tissue bridges added so much predictive power to our models — in some ways outperforming models that include only clinical factors.

For every millimeter of preserved tissue bridge width, for example, a patient was likely to gain 5.9 points (out of 100) on their motor score and 6.4 points (out of 112 points) on their light touch score after three months.

The patients in Germany and Switzerland were followed for 12 months, allowing researchers to examine more long-term trends. Overall, they found that patients experienced most of their recovery within the first three months.

Sorting patients

Promising experimental therapies for spinal cord injuries include electrical stimulation and antibody treatments that encourage nerve regeneration, but a pressing problem in getting clinical trials off the ground is not knowing whether and to what extent some patients will recover on their own. Grouping patients with similar recovery potential would increase the efficiency of clinical trials, requiring fewer patients to reach statistical validity.

The new study found that tissue bridges greatly improved the ability to sort patients into groups with similar prognoses.

“What was most surprising is that tissue bridges added so much predictive power to our models — in some ways outperforming models that include only clinical factors,” Pfyffer said.

Models that combined tissue bridge measurements with clinical assessments achieved the most consistent sorting results at all three centers. And consistent results from different datasets is the goal, Pfyffer said.

“You would not want to see differences in prediction because it would mean that the study findings cannot be replicated across centers,” he said.

Nothing fancy

The researchers are now working on ways to further improve prognostication with tissue bridges. They are applying machine learning to take more objective measurements from MRI scans and to use additional scans to measure all dimensions of the spared tissue. That would allow them to identify the specific tracts of the spinal cord, which innervate distinct parts of the body, that remain viable.

But the benefits of tissue bridge measurements are already accessible in most hospitals that treat spinal cord injuries. “We retrieved these measures from conventional MRI scans that are done in a lot of centers, so it’s not fancy MRI scans that take a lot of time,” Pfyffer said. “It’s quite easy to retrieve that information.”

Researchers from Craig Hospital, the University of Colorado School of Medicine, BG Trauma Center Murnau, Balgrist University Hospital, University of College London and the Ohio State University contributed to the work.

The study received funding from Wings for Life, Austria; the International Foundation for Research in Paraplegia; the EU project Horizon 2020; and the framework of ERA-NET NEURON.

About Stanford Medicine

Stanford Medicine is an integrated academic health system comprising the Stanford School of Medicine and adult and pediatric health care delivery systems. Together, they harness the full potential of biomedicine through collaborative research, education and clinical care for patients. For more information, please visit med.stanford.edu.