Background: 65 million people worldwide experience some form of epilepsy or seizures in their lifetime, with around 33% of those people undergoing a psychogenic non-epileptic seizure (PNES). There is a distinguishable difference between an epileptic attack and a non-epileptic attack, as an epileptic seizure is visible on an electroencephalogram, rather than PNES which does not present with abnormal brain activity. Instead of presenting with this type of brain activity, it is triggered by emotional or mental distress along with other mental health disorders such as post traumatic stress disorder and anxiety. A large problem relating to PNES reoccurs with the diagnostic stage of the disease, as it can take seven to ten years to diagnose. Many of these patients are put on anti-epilepsy drugs due to the difficulty to diagnose PNES. This can lead to larger problems along the way, as the patient is resistant to the drug. They also face the possibility of addiction and the loss of independence and self-reliance.
Goal: To utilize the machine learning system IBM Watson to analyze the situation and history of the patient presenting with seizures and use personalized artificial intelligence to determine the diagnosis of psychogenic non-epileptic seizures without putting the patient on anti-epilepsy drugs. This can provide an accurate diagnosis that is based off of a personalized perspective of the patient as well as the analyzation of the patient’s reaction to their surroundings and situations.Methods: When a patient comes to their primary care physician/neurologist presenting with what the physician believes to be epilepsy, however the electrical brain activity appears to be normal, instead of putting them on an AED or ignoring the situation all together, the physician would introduce the IBM Watson tool to the patient. The unsupervised machine learning device would evaluate the patient and their reactions to certain situations over the duration of thirty days. After the time has exhausted, the system would analyze the results, specifically searching for the physical and emotional environmental patterns that triggered the psychogenic non-epileptic seizure to emerge. Following this process, the advanced machine learning would then make a connection (if applicable) to the family and personal history of the patient, for example if they were previously involved in an emotionally-scarring experience. After completed, the system would analyze those results and make an evidence-based and personalized decision about the diagnosis of the patient, then relaying it back to the neurologist or psychologist to decide where to direct their treatment from there.