Clinicians traditionally use fractional anistropy (FA) to evaluate concussions on MR scans, but a mathematical tool called Shannon entropy may be more effective at determining which concussion patients will suffer migraines. A study investigating the tool was recently published online in the journal Radiology.

Post-traumatic migraines are prevalent among concussion patients. To detect them, clinicians assess damage to the brain’s signal-transmitting white matter and look for symptoms such as headaches.

MR is used to generate a frequency distribution graph of the whole brain, which is called a histogram. A mean FA, which is the measure of how easily water can move through the brain, can then be derived to determine if there is white matter injury.

But mean FA has its drawbacks since it represents and average. If a patient has a higher FA to start with and then loses white matter integrity from trauma, their mean FA may still average out to be considered normal.

"This is often what happens with concussion patients where swelling of the white matter can cause increases in FA and breakage of the white matter can cause decreases in FA," Dr. Lea M. Alhilali from the University of Pittsburgh Medical Center, told HCB News. "If you are only looking at the mean, these opposite directional changes will average out to no change."


Researchers at the University of Pittsburgh Medical Center took a new approach and analyzed the MR images using an information theory model called Shannon entropy. The model is useful in evaluating areas of disorder in a complex system such as the brain.

A healthy brain has a high amount of entropy, but people with injuries that affect the white matter might lose some of that entropy.

The researchers used FA maps and neurocognitive testing results from 74 concussion patients — 57 with post-traumatic migraines and 17 without. FA maps were also obtained from 22 healthy controls and 20 control patients with migraine headaches.

The mean FA and Shannon entropy were taken from the total brain FA histograms and compared between the concussion and control groups and then between the groups with and without post-traumatic migraine.

They found that the Shannon entropy analysis of FA histograms was a better at differentiating between concussion and control patients compared to mean FA. It also was more effective at differentiating between those with and without post-traumatic migraine.

The concussion patients had much lower Shannon entropy than the control patients and the patients with post-traumatic migraines had significantly lower Shannon entropy than other concussion patients. They also found that the patients with lower entropy look a longer amount of time to recover.

If this approach continues to show promise, it may have the potential to change how concussion patients are assessed with imaging. Alhilali believes that Shannon entropy can succeed where conventional measures of FA have failed — accurately characterizing the multidimensional white matter changes in concussion injuries.

"Shannon entropy can completely change the way neuroimaging approaches the evaluation of concussion because it can be performed simply, in an automated fashion without additional labor, is easily implemented as part of all MRI vendor platforms, is quantitative, and has the potential to identify white matter injuries that may go undetected by conventional analysis methods," said Alhilali.

But more research is required to determine if there are other potential applications for Shannon entropy including predicting future cognitive performance in concussion patients.