Abstracts

RATIONALE: Advanced warning is one of several valuable applications of real-time automated seizure detection. The generic mode of an algorithm (Epilepsia 1998 39(6):615-627) accurately predicts (mean 15.5 s) clinical seizure onset (CSO), defined as the "aura" or an observable typical change, whichever occurs first. Since this definition of CSO may not accurately reflect time of loss of function (LOF), we examine the latency from seizure detection to LOF in patients with intractable temporal lobe epilepsy undergoing invasive monitoring. METHODS: A complex reaction time test (36 stimuli, ~70 s duration) was administered between 8am and 8pm randomly and upon automated seizure detection to 17 patients, starting on the second post-operative day. Tests were classed as occurring during ictal or interictal periods, and corresponding ECoG segments were verified via expert visual analysis. Analysis of test performance included both correctness of reponse and reaction time. LOF was defined as the last point in time for which performance was statistically indistinguishable from random (interictal) tests. RESULTS: Fifteen out of 17 patients took tests during seizures. LOF occurred in at least one seizure test in 14 of 15 patients. Mean latency from seizure detection to LOF was 31 s. One case with a history of complex partial seizures retained function during all seizure tests, and another had no LOF during 6 of 12 seizure tests. CONCLUSIONS: Ability to function after electrographic onset of partial seizures, as measured in this study, is longer than estimates reported in the literature, where CSO is based on the "aura" or first observable change.