By Mike Martin for Epilepsy USA
A new system that may help to “automate” epilepsy diagnosis has sparked an updated version of an old debate: Is medicine—in this case, clinical neurology— more art than science or more science than art?
Supplementing clinical expertise with artificial intelligence, researchers from Texas Tech University in Lubbock and China’s Jiangsu Provincial Hospital in Nanjing have used a computer algorithm to interpret so-called “interictal” or between-seizure electroencephalogram (EEG) data.
In a paper for the 2009 IEEE Engineering in Medicine and Biology Society conference, the research team—which includes computer scientists, electrical engineers, a neurologist, and two neurosurgeons—claims the algorithm correctly identified epileptiform EEG data with about 94 percent accuracy.
Though automated epilepsy diagnosis is not a new concept, the study also claims an important advance: the use of so-called “artificial probabilistic neural networks” to interpret scalp—rather than intracranial (inside the skull)—EEG data.