Researchers combine MRI and mathematics to predict future migraines in concussion patients

An information theory model known as Shannon entropy, used in conjunction with diffusion tensor MRI, may help predict the likelihood of migraines in patients with concussions, according to a study published Feb. 2 in the journal Radiology.

Nearly 4 million people a year suffer injuries resulting in concussions, with about 90 percent of those patients experiencing post-traumatic headache symptoms. When those symptoms escalate into post-traumatic migraines, it could signal a larger problem inside the brain, said lead author Joseph Delic, MD, and his colleagues from the University of Pittsburgh Medical Center.

“[Post-traumatic migraines] are associated with lower neurocognitive test scores, aggravation of other posttraumatic symptoms, and a protracted recovery,” Delic wrote. “Although the pathophysiology of [post-traumatic migraines] remains unclear, migraines outside of trauma have been associated with focal white matter abnormalities.”

The researchers set out to determine the performance of Shannon entropy—a mathematical model based in information theory, or the mathematical laws surrounding the behavior of data—as a diagnostic tool in concussion patients with and without post-traumatic migraines.

Delic and his team obtained fractional anistropy (FA) maps and conducted neurocognitive testing on 74 concussion patients: 57 who had post-traumatic migraines and 17 who did not. FA maps were also obtained in 22 control subjects and in 20 control patients who suffered from migraine headaches. They calculated both mean FA and Shannon entropy in the study participants using total brain histograms, then compared the results between concussion patients and control subjects as well as between patients with and without post-traumatic migraines.

Their analysis showed that patients with concussions who experienced posttraumatic migraines had significantly lower Shannon entropy—but not mean FA—than did concussion patients with no migraine symptoms, with Shannon entropy also providing better discrimination between concussion patients and control subjects. Specifically, patients with Shannon entropy of less than 0.75 were 16 times more likely to have experienced concussion and three times more likely to develop post-traumatic migraines.

“Our study results suggest that [Shannon entropy] analysis of FA histograms may better reflect the white matter pathologic conditions underlying [concussions] and postconcussive symptoms,” the researchers concluded, “and may have a role as a diagnostic tool and prognostic biomarker in individual patients with concussion.”