People using an iPad

Okazaki EM, et al. Epilepsy & Behavior, 82(2018):140-43

The diagnosis of epilepsy can present challenges for both neurologists and non-neurologists. Epilepsy can present with many different symptoms. Other medical conditions can have very similar symptoms, making the diagnosis of epilepsy challenging in some cases. Delays in diagnosis and treatment sometimes result from the difficulty in establishing the correct diagnosis based on these symptoms. Clinical decision support tools are being developed to help medical caregivers improve the accuracy of a diagnosis and shorten the time it takes for a person to receive the correct treatment.


The goal of this study was to test the usefulness and accuracy of a clinical decision support tool called EpiFinder. EpiFinder is a form of artificial intelligence based on pattern recognition. Specifically, this tool (application or “app”) is designed to recognize a person’s symptoms that may help lead to a correct diagnosis of epilepsy.

Editor’s Note: EpiFinder was a finalist in the Epilepsy Foundation’s 2017 Shark Tank Competition.

Description of Study

  • Data were collected in 57 adult patients being admitted to an epilepsy monitoring unit (EMU) for diagnosis of their neurological symptoms.
  • People in the study entered their symptoms into the EpiFinder application (“app”), using an iPad, with the guidance of a trained epilepsy neurologist.
  • EpiFinder was used to predict a diagnosis for each person when they were admitted to the EMU, based on the reported symptoms entered into the app.
  • The EpiFinder diagnosis was compared to the final diagnosis a person received after completing their video EEG (electroencephalogram) testing in the EMU.
  • Results were considered positive if the EpiFinder app was able to predict a diagnosis of a focal or generalized epilepsy syndrome that matched the seizure data recorded on video EEG testing during their stay in the EMU.

Summary of Study Findings

  • 30 women and 27 men, ages 18-78 (mean age 42), were enrolled in the study.
  • In the 53 adults studied, EpiFinder accurately predicted the presence of epilepsy versus a different diagnosis (non-epilepsy) in 86.8% of the people. This means most of the time EpiFinder was able to correctly identify if a person had an epilepsy syndrome based on the symptoms and descriptions entered into the app.
  • EpiFinder led to an incorrect diagnosis in 7 out of 53 people (13.2%).

What does this mean?

  • Testing in a group of adults with epilepsy found that EpiFinder was able to predict the presence of epilepsy versus another non-epilepsy diagnosis in 86% of cases.
  • Results showed that Epifinder may be a useful tool in screening people whose symptoms may or may not be related to epilepsy.
  • This study of EpiFinder was limited to only being able to demonstrate the ability to predict a diagnosis of epilepsy versus non-epilepsy.
  • EpiFinder is a tool that shows promise. Further development of the application may allow for broader use and more detailed clinical decision support of epilepsy syndromes and other non-epilepsy diagnosis.
  • Development of EpiFinder and other applications may help to diagnose epilepsy sooner and help people find appropriate medical care sooner.
  • EpiFinder will also need to be tested by non-neurologists to see if a larger number of health care providers can use this tool for support when making clinical decisions.

Article published in Epilepsy & Behavior, May 2018.

we recommend
Authored By: 
Elaine Kiriakopoulos MD, MSc
Authored Date: 
Reviewed By: 
Patty Obsorne Shafer RN, MN
Tuesday, August 7, 2018