Abstracts

Rationale: Epilepsy diagnosis is largely based on seizure semiology. However, the accuracy of visual diagnosis for different seizure type may be different. In addition, whether evaluating more seizures helps to improve the diagnosis is unknown. Methods: Totally 150 seizure video from 50 patients were clipped from our video EEG database. For each patient, we selected 3 typical seizure of the same seizure type. The videos were reviewed by 30 blinded physicians, which included 10 neurological graduate students, 10 neurological fellows, 4 neurologists and 5 epileptologists. The 50 seizure types include 37 epileptic seizures and 13 nonepileptic seizures (5 psychogenic non-epileptic seizure, 6 parosysmal movement disorders and 2 sleep disorder). The 37 epileptic seizures were classified by Luder s semiology classification (complex motor seizure, simple motor seizure and dialeptic seizure). The reviewers were informed that all 3 seizures from one patient are of the same type. The seizures from one patient were presented to the reviewers consecutively. They wrote down the answer as epileptic seizure or non-epileptic seizure for the patients after each seizure was presented. They were not allowed to change previous answers.Results: Epileptologists had a significant higher accuracy in visual diagnosis of seizures than other groups (Fig. 1), and no group difference was found between other groups. For all reviewers, only a slight increase of accuracy was found between the time when the first video was reviewed (0.626) and the third video was reviewed (0.656, p=.0008). However, such increase was not found for epileptologists. After all the 3 seizure in one patients were presented, the average accuracy of identifying complex motor seizures was 0.588, which was significantly lower than simple motor seizure (0.684) and dialeptic seizures (0.810). In simple motor seizures, the accuracy of identification for generalized tonic-clonic seizures was the highest. In complex motor seizures, the lowest accuracy of seizure identification was found for hypermotor seizure (0.483), gelastic seizures (0.414) and dycrastic seizure (0.517). Conclusions: The accuracy of visual diagnosis of seizure depends on seizure type as well as the reviewer s experience. It is difficult to identify the epileptic seizures if they present as more nature movements . Our data show reviewing more seizure does not greatly improve in accuracy in seizure diagnosis.