Localization value of seizure semiology analyzed by the conditional inference tree method

Abstract

ObjectiveAlthough accurate interpretation of seizures is important for the management of patients with epilepsy, studies on the localizing value of seizure semiology and the reliability of the semiology descriptions are scarce. The objective of our study is to investigate the accuracy of video-recorded seizure semiology in the classification and localization of epileptic seizures. We also evaluated the reliability of the semiology descriptions provided by the patients or their caregivers. MethodsVideo-recorded clinical seizures from 831 consecutive patients (391 females; 31.7±11.6 years) were analyzed retrospectively. Epileptic seizures were classified as generalized and partial seizures, and patients with partial seizures were further divided into five ictal onset areas. In order to analyze the diagnostic value of individual semiologic features for clinical diagnosis, we used the conditional inference tree method. ResultsGeneralized and partial seizures were differentiated with high accuracy (97.1%), but the accuracy of localization among the five ictal onset areas was relatively low (56.1%), which was largely attributed to the difficulty in the discrimination between mesial and lateral temporal onset seizures. Lateralization of the ictal onset area in partial seizures was possible in 427 (55.1%) patients based on video analysis, nevertheless it was possible in only 158 (20.4%) patients based on historical semiology descriptions. ConclusionsThe results of our study suggest that careful observation of seizure semiology may be useful for the differentiation of ictal onset areas. However, the semiologic differentiation between mesial and lateral temporal onset seizures is difficult, and historical semiologic descriptions should be interpreted carefully because of their low reliability.