Authors :
Akshaya Rathin Sivaji, MD – Epilepsy Center, Neurological Institute, University Hospitals Cleveland Medical Center, Cleveland, OH, USA
Saeideh Salahizadeh, MD – Epilepsy Center, Neurological Institute, University Hospitals Cleveland Medical Center, Cleveland, OH, USA
Presenting Author: Suraj Thyagaraj, PhD – Epilepsy Center, Neurological Institute, University Hospitals Cleveland Medical Center, Cleveland, OH, USA
Jun Park, MD – Epilepsy Center, Neurological Institute, University Hospitals Cleveland Medical Center, Cleveland, OH, USA
Rationale:
EEG waveforms associated with different conditions often display a uniform pattern making them difficult to distinguish. In these scenarios, review of relevant clinical information along with careful analysis of subtle morphological features are necessary for accurate diagnosis. This study aims to assess the diagnostic reliability of four distinct conditions with often indistinguishable waveform morphology, in the absence of relevant clinical history.
Methods:
This prospective analysis involved 86 participants from multiple centers, including attending physicians, general neurologists, residents, fellows and others group, engaged in EEG interpretation. We selected four different clinical disorders—triphasic waves, slow spike-and-waves(Lennox-Gastaut syndrome), benign focal epileptiform discharges, and generalized spike-and-wave during sleep(Landau-Kleffner syndrome) due to strikingly similar EEG patterns. This similarity makes distinguishing them particularly challenging, even for experienced clinicians, especially when the clinical context is unknown. We created a QR code-enabled online survey featuring EEG epochs from these disorders. Participants were asked to provide EEG interpretation or clinical diagnosis. These EEGs were performed for clinical indications at University Hospitals, Cleveland Medical Center.
Results:
The analysis of survey responses revealed varying levels of agreement in identifying among participants(Table 1). For triphasic waves, resulted in significant agreement of 47%, with 53% of discordance rate. In contrast, slow spike and wave pattern demonstrated significant, low agreement rate of 24% and discordance rate of 76%. Similarly, BFEDs indicated an 36% of agreement rate and a discordance rate of 64%. Generalized slow spike and wave mirrored the variability seen with slow-spike and wave pattern, with only 24% of agreement, while 76% were discordant.
We used the non-parametric Kruskal-Wallis test to determine statistical significance across the response from the five independent groups. Upon running with 0.05 as significance level, the differences were found to be not statistically significant (p >0.05). Testing for significance in response to each question by each group as overall response to each question, we found none of the four questions were answered correctly by more than half of responders. We then tried to find out if any of the individual groups were able to get an accuracy significantly greater than 50% . We ran a one-sided t-test on the modified response data that was binarized, with the null hypothesis that the data comes from a population that has a mean greater than 0.5 ( >50% accuracy). None of the five groups were able to score an accuracy greater than 50% at the default significance level of 0.05, on any of the four questions.
Conclusions:
The analysis shows that EEG indications of triphasic waves, slow spike and wave, benign focal epileptiform discharges, and generalized spike and wave complexes, by themselves are difficult to distinguish from each other, irrespective of operator training level. Relevant clinical history has to be considered while performing EEG interpretation, for accurate identification and clinical planning.
Funding: None