User login
BALTIMORE – An algorithm that analyzes recordings from a scalp, rather than an intracranial, EEG has been demonstrated to predict seizures with odds significantly greater than chance.
Dr. J. Chris Sackellares, chief scientific officer of Optima Neuroscience Inc., Gainesville, Fla., presented these findings at the annual meeting of the American Neurological Association.
At regular intervals, the algorithm sequentially calculates a pattern match regularity statistic–a measure of how ordered an electrical signal is–in four different channel groups located on patients' scalps.
Comparisons of the electrical activity in the four channels are used to predict seizures by detecting when the electrical activity of certain channels begins to converge on specific pattern match regularity statistics over time, indicating that a seizure is imminent.
Dr. Sackellares and his colleagues tested their algorithm in 51 patients with temporal lobe epilepsy. They captured 159 seizures and analyzed 93 segments of scalp EEG recordings. Each segment recording lasted a mean of 26 hours, with a total length of 48 hours per patient.
The researchers' automated prediction algorithm detected seizures with 95% sensitivity, generating one false-positive result every 8.6 hours, compared with a random predictor model that had an overall sensitivity of 83% and a rate of one false-positive result every 2.5 hours in individual patients.
The algorithm could detect seizures with nearly 70% sensitivity and a false-positive rate of about 0.22 per hour. The prediction was not sensitive enough for use in inpatient monitoring units or ambulatory EEC recording.
The study was supported by a grant from the National Institute of Neurological Disorders and Stroke.
BALTIMORE – An algorithm that analyzes recordings from a scalp, rather than an intracranial, EEG has been demonstrated to predict seizures with odds significantly greater than chance.
Dr. J. Chris Sackellares, chief scientific officer of Optima Neuroscience Inc., Gainesville, Fla., presented these findings at the annual meeting of the American Neurological Association.
At regular intervals, the algorithm sequentially calculates a pattern match regularity statistic–a measure of how ordered an electrical signal is–in four different channel groups located on patients' scalps.
Comparisons of the electrical activity in the four channels are used to predict seizures by detecting when the electrical activity of certain channels begins to converge on specific pattern match regularity statistics over time, indicating that a seizure is imminent.
Dr. Sackellares and his colleagues tested their algorithm in 51 patients with temporal lobe epilepsy. They captured 159 seizures and analyzed 93 segments of scalp EEG recordings. Each segment recording lasted a mean of 26 hours, with a total length of 48 hours per patient.
The researchers' automated prediction algorithm detected seizures with 95% sensitivity, generating one false-positive result every 8.6 hours, compared with a random predictor model that had an overall sensitivity of 83% and a rate of one false-positive result every 2.5 hours in individual patients.
The algorithm could detect seizures with nearly 70% sensitivity and a false-positive rate of about 0.22 per hour. The prediction was not sensitive enough for use in inpatient monitoring units or ambulatory EEC recording.
The study was supported by a grant from the National Institute of Neurological Disorders and Stroke.
BALTIMORE – An algorithm that analyzes recordings from a scalp, rather than an intracranial, EEG has been demonstrated to predict seizures with odds significantly greater than chance.
Dr. J. Chris Sackellares, chief scientific officer of Optima Neuroscience Inc., Gainesville, Fla., presented these findings at the annual meeting of the American Neurological Association.
At regular intervals, the algorithm sequentially calculates a pattern match regularity statistic–a measure of how ordered an electrical signal is–in four different channel groups located on patients' scalps.
Comparisons of the electrical activity in the four channels are used to predict seizures by detecting when the electrical activity of certain channels begins to converge on specific pattern match regularity statistics over time, indicating that a seizure is imminent.
Dr. Sackellares and his colleagues tested their algorithm in 51 patients with temporal lobe epilepsy. They captured 159 seizures and analyzed 93 segments of scalp EEG recordings. Each segment recording lasted a mean of 26 hours, with a total length of 48 hours per patient.
The researchers' automated prediction algorithm detected seizures with 95% sensitivity, generating one false-positive result every 8.6 hours, compared with a random predictor model that had an overall sensitivity of 83% and a rate of one false-positive result every 2.5 hours in individual patients.
The algorithm could detect seizures with nearly 70% sensitivity and a false-positive rate of about 0.22 per hour. The prediction was not sensitive enough for use in inpatient monitoring units or ambulatory EEC recording.
The study was supported by a grant from the National Institute of Neurological Disorders and Stroke.