We exist to improve the lives
of everyone affected by epilepsy

Mathematical analysis of eeg may lead to quicker diagnosis

19 May, 2005

Researchers claim
to have found a mathematical way to translate brain wave readings into
simple terms to help doctors and nurses more easily identify babies
with epilepsy.

Scientists at the Evelyn F and William L McKnight Brain Institute
at the University of Florida (UF) commented that it is difficult to
tell whether babies have epilepsy because they are often asleep. Even
when awake, they cannot provide clues through their speech, nor do
abnormal movements necessarily indicate a seizure.

One
method for diagnosing epilepsy is to have an electroencephalogram (EEG)
which monitors electrical activity through electrodes placed on a
patient's scalp. However, the researchers point out that the test is
expensive, requires a high level of training to interpret and often
isn't readily available in hospitals.

The
researchers acknowledge that experienced paediatric neurologists and
EEG technicians can identify babies with epilepsy by reviewing their
EEGs. However, they claim for babies with less severe epilepsy,
abnormal brain wave patterns may only be obvious in a few segments in
the entire recording.

Dr Paul Carney, professor at the McKnight Brain Institute, said:

"An
EEG provides a squiggly line readout of brain activity. Our goal is to
take our findings and develop a tool that can run in real time right
next to the blood pressure and other monitoring devices in a hospital.
If successful it would be one of the first brain function monitors for
clinical use in the neonatal intensive care unit."

The
researchers say they can convert an EEG readout into a numeric value.
For example, a reading of "20" would indicate normal brain activity and
a reading of "10" would indicate a seizure. They tested their idea by
reviewing the EEGs of 35 babies up to a month old, 23 of whom had
normal brain function. They were able to pinpoint which babies had
epilepsy through differences in statistical values.