Time Frequency Mapping of the Rhythmic Limb Movements Reliably Distinguishes Convulsive Epileptic from Psychogenic Non-epileptic Seizures

  • Ms Jade Bayly, The Department of Neurology and Swinburne University of Technology, Australia
  • Mrs Michelle Smit, The Departments of Neurology, Australia
  • Ms Erin Fogg, Swinburne University of Technology, Australia
  • Professor Terence O'Brien, The Departments of Neurology and Medicine, Australia
  • Dr Anita Vinton, The Departments of Neurology and Medicine, Australia
  • Mr John Carino, The Departments of Neurology, Australia
  • Dr Rob Griffiths, The Florey Neurosciences Institute, Australia
  • Professor Malcolm Horne, Centre for Neurosciences, The Florey Neurosciences Institute, St Vincent's Hospital, Australia

Aim: To evaluate the diagnostic sensitivity and specificity of a wrist-strapped movement monitor device (accelerometer) to differentiate between convulsive epileptic seizures and convulsive psychogenic non-epileptic seizures (PNES).
Objective: The diagnosis of PNES requires in-patient video-EEG monitoring, which is resource intensive and stressful for patients. Research shows convulsive PNES have a characteristic pattern of rhythmic movement artefact on the EEG that maintains a stable frequency over time, whereas during convulsive epileptic seizures it evolves. We examined the potential for time-frequency mapping of data from a wrist-strapped accelerometer to be used as a diagnostic tool.
Method: Time-frequency mapping was obtained during 61 convulsive seizure-like events from 32 patients during in-patient video-EEG monitoring. 23 had PNES, 8 had epileptic seizures, and one had both. The time-frequency maps were derived using NEUROSCAN™ software via fast fourier transformations.
Results: PNES displayed stable dominant frequencies of movement during the course of a seizure, while epileptic seizures showed a more variable, evolving frequency. The coefficient of variation (CV) of limb movement frequency during PNES was significantly less than during epileptic seizures (median, 14.9%; range, 1.8%-39.8% vs. 52.2%, 31.3%-75.6%; p<0.001). Blinded analysis of time frequency maps correctly diagnosed 48/50 (96%) non-epileptic and 7/11 (63.6%%) epileptic seizures. Using a CV of <32 as a diagnostic cut-off provided sensitivity=94%, specificity=91%, positive predictive value =98%, negative predictive value=77%
Conclusion: This study indicates that time-frequency analysis of data from an accelerometer can potentially be utilized as a diagnostic tool for PNES which is applicable for outpatient use.